{"id":1053711,"date":"2024-07-03T05:21:39","date_gmt":"2024-07-03T12:21:39","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-project&#038;p=1053711"},"modified":"2026-03-24T16:28:04","modified_gmt":"2026-03-24T23:28:04","slug":"tools-for-thought","status":"publish","type":"msr-project","link":"https:\/\/www.microsoft.com\/en-us\/research\/project\/tools-for-thought\/","title":{"rendered":"Tools for Thought"},"content":{"rendered":"<section class=\"mb-3 moray-highlight\">\n\t<div class=\"card-img-overlay mx-lg-0\">\n\t\t<div class=\"card-background  has-background- card-background--full-bleed\">\n\t\t\t<img loading=\"lazy\" decoding=\"async\" width=\"1920\" height=\"610\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/Tools4Thought_header_1920x720-1.jpg\" class=\"attachment-full size-full\" alt=\"An illustration of the top of the statue of The Thinker by Rodin. There is a nest in his head, with three chicks sticking out, and their parent bird flying above.\" style=\"object-position: 55% 54%\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/Tools4Thought_header_1920x720-1.jpg 1920w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/Tools4Thought_header_1920x720-1-300x95.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/Tools4Thought_header_1920x720-1-1024x325.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/Tools4Thought_header_1920x720-1-768x244.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/Tools4Thought_header_1920x720-1-1536x488.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/Tools4Thought_header_1920x720-1-240x76.jpg 240w\" sizes=\"auto, (max-width: 1920px) 100vw, 1920px\" \/>\t\t<\/div>\n\t\t<!-- Foreground -->\n\t\t<div class=\"card-foreground d-flex mt-md-n5 my-lg-5 px-g px-lg-0\">\n\t\t\t<!-- Container -->\n\t\t\t<div class=\"container d-flex mt-md-n5 my-lg-5 \">\n\t\t\t\t<!-- Card wrapper -->\n\t\t\t\t<div class=\"w-100 w-lg-col-5\">\n\t\t\t\t\t<!-- Card -->\n\t\t\t\t\t<div class=\"card material-md-card py-5 px-md-5\">\n\t\t\t\t\t\t<div class=\"card-body \">\n\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\n\n<h1 class=\"wp-block-heading is-style-l\" id=\"tools-for-thought\"><strong>Tools for Thought<\/strong><\/h1>\n\n\n\n<p><strong>Better thinking through AI<\/strong><br><em>A <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/theme\/people-centric-ai\/\" target=\"_blank\" rel=\"noreferrer noopener\">People-Centric AI<\/a> Project<\/em><\/p>\n\n\n\n<div class=\"wp-block-buttons is-layout-flex wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"mailto:\/\/tools4thought@microsoft.com\" target=\"_blank\" rel=\"noreferrer noopener\">Contact Us<\/a><\/div>\n<\/div>\n\n\t\t\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div>\n\t<\/div>\n<\/section>\n\n\n\n\n\n<div class=\"wp-block-group is-layout-constrained wp-block-group-is-layout-constrained\">\n<div class=\"wp-block-group is-layout-constrained wp-block-group-is-layout-constrained\">\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<h2 class=\"wp-block-heading\" id=\"news-1\">Advait Sarkar at <strong>TED<\/strong>AI Vienna<\/h2>\n\n\n\n<p><em>&#8220;What would you rather have: a tool that thinks for you, or a tool that makes you think?&#8221;<\/em><\/p>\n\n\n\n<figure class=\"wp-block-embed is-type-video is-provider-youtube wp-block-embed-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"wp-block-embed__wrapper\">\n<iframe loading=\"lazy\" title=\"How to Stop AI from Killing Your Critical Thinking | Advait Sarkar | TED\" width=\"500\" height=\"281\" src=\"https:\/\/www.youtube-nocookie.com\/embed\/3lPnN8omdPA?feature=oembed&rel=0\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" referrerpolicy=\"strict-origin-when-cross-origin\" allowfullscreen><\/iframe>\n<\/div><figcaption class=\"wp-element-caption\">Advait at <strong>TED<\/strong>AI in September<\/figcaption><\/figure>\n\n\n\n<p>Senior Researcher <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/advait\/\" target=\"_blank\" rel=\"noreferrer noopener\">Advait Sarkar<\/a> presented at <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.ted.com\/talks\/164483\" target=\"_blank\" rel=\"noopener noreferrer\"><strong>TED<\/strong>AI<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> in Vienna on September 26<sup>th<\/sup> 2025.<\/p>\n\n\n\n<p id=\"caption-attachment-1155766\">\u00a9 TEDAI Vienna \/ Robert Leslie<\/p>\n\n\n\n<p>Many AI tools focus on solving specific tasks like content generation or process automation. Though useful and powerful, these systems may affect how we think, learn, build skills, and develop expertise. In this talk, Advait introduces a prototype that shows a new kind of AI tool, one that leverages the value of AI generated content, but keeps the user cognitively engaged in looking for critiques, alternatives, and lateral moves to improve the outcome.<\/p>\n\n\n\n<div class=\"wp-block-buttons is-layout-flex wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/www.ted.com\/talks\/164483\">Watch the<strong style=\"font-size: 1rem\">TED<\/strong>AI Video<\/a><\/div>\n\n\n\n<div class=\"wp-block-button\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/tedai-2025-artificial-intelligence-as-a-tool-for-thought\/\" target=\"_blank\" rel=\"noreferrer noopener\">Transcript with citations<\/a><\/div>\n\n\n\n<div class=\"wp-block-button\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/articles\/rethinking-ai-in-knowledge-work-from-assistant-to-tool-for-thought\/\" target=\"_blank\" rel=\"noreferrer noopener\">Blog<\/a><\/div>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-vertically-aligned-top is-layout-flow wp-block-column-is-layout-flow\">\n<h2 class=\"wp-block-heading\" id=\"news-1\">Promptions<\/h2>\n\n\n\n<p>Promptions (prompts + options) is an open source project on <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/labs.ai.azure.com\/projects\/promptions\/\" target=\"_blank\" rel=\"noopener noreferrer\"><strong>Microsoft Foundry Labs<\/strong><span class=\"sr-only\"> (opens in new tab)<\/span><\/a>.<\/p>\n\n\n\n<figure class=\"wp-block-embed is-type-video is-provider-youtube wp-block-embed-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"wp-block-embed__wrapper\">\n<iframe loading=\"lazy\" title=\"Promptions\" width=\"500\" height=\"281\" src=\"https:\/\/www.youtube-nocookie.com\/embed\/vr3fZpkKy8Q?feature=oembed&rel=0\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" referrerpolicy=\"strict-origin-when-cross-origin\" allowfullscreen><\/iframe>\n<\/div><figcaption class=\"wp-element-caption\">The Promptions open-source repo has two sample applications, one chatbot and one image generator.<\/figcaption><\/figure>\n\n\n\n<p>Promptions helps users steer AI responses more effectively. Instead of manually refining prompts, Promptions generates contextual UI elements, like radio buttons, checkboxes, or toggles.<\/p>\n\n\n\n<p><strong>Blog Post:<\/strong> <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/project\/tools-for-thought\/promptions-repo\/\" target=\"_blank\" rel=\"noreferrer noopener\">Promptions Open Source Repository: If You\u2019re Building UI for AI, Give Users the Power of Choice<\/a><\/p>\n\n\n\n<div class=\"wp-block-buttons is-layout-flex wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/github.com\/microsoft\/Promptions\" target=\"_blank\" rel=\"noreferrer noopener\">GitHub Repo<\/a><\/div>\n\n\n\n<div class=\"wp-block-button\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/aka.ms\/promptionspaper\">Paper<\/a><\/div>\n\n\n\n<div class=\"wp-block-button\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/labs.ai.azure.com\/projects\/promptions\/\" target=\"_blank\" rel=\"noreferrer noopener\">Microsoft Foundry Labs<\/a><\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"about-the-tools-for-thought-project-1\"><strong>About the Tools for Thought Project<\/strong><\/h3>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:66%\">\n<p>The Tools for Thought (T4T) project explores&nbsp;how AI might help people to <em>think better<\/em>, so that:&nbsp;<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>as well as getting the job done, it helps us better understand and figure out the job.<\/li>\n\n\n\n<li>as well as creating content, it helps us think more critically and with more insight throughout an entire workflow.<\/li>\n\n\n\n<li>as well as automating known processes, it helps organisations predict and explore the unknown.<\/li>\n<\/ul>\n\n\n\n<p>We explore:<\/p>\n\n\n\n<p><strong>Working With Purpose<\/strong>: Focused on <em>work over time<\/em>, how can AI can surface people&#8217;s goal-driven reasoning, capture it as an AI input, and harness it to support and explore workflows.<\/p>\n\n\n\n<p><strong>Thinking By Doing<\/strong>: Focused on <em>work in the moment<\/em>, how can AI can deepen the quality of critical and creative thinking as people consume and create?<\/p>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:34%\">\n<figure class=\"wp-block-image aligncenter\"><img loading=\"lazy\" decoding=\"async\" width=\"826\" height=\"720\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/t4t.png\" alt=\"diagram, venn diagram\" class=\"wp-image-1146609\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/t4t.png 826w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/t4t-300x262.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/t4t-768x669.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/t4t-207x180.png 207w\" sizes=\"auto, (max-width: 826px) 100vw, 826px\" \/><\/figure>\n<\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"what-we-know-an-overview-of-our-research\"><strong>What We Know<\/strong> | An Overview of our Research<\/h3>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div class=\"wp-block-msr-show-more\">\n\t<div class=\"bg-neutral-100 p-5\">\n\t\t<div class=\"show-more-show-less\">\n\t\t\t<div>\n\t\t\t\t<span>\n\t\t\t\t\t\n\n<h4 class=\"wp-block-heading\" id=\"ai-should-support-critical-thinking-and-metacognition-in-knowledge-work\">AI should support critical thinking and metacognition in knowledge work<\/h4>\n\n\n\n\t\t\t\t<\/span>\n\t\t\t\t<span id=\"show-more-show-less-toggle-1\" class=\"show-more-show-less-toggleable-content\">\n\t\t\t\t\t\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-should-challenge-not-obey\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/ai-should-challenge-not-obey\/\">AI should challenge, not obey<\/a><\/h5>\n\n\n\n<p>Rather than optimizing solely for productivity or user satisfaction, AI systems should trigger deeper reflection and more rigorous decision-making in people \u2014exposing flawed reasoning, asking probing questions, and offering alternative perspectives. Strategies such as argument mapping, adversarial prompting, and adaptive friction levels should balance engagement with user comfort\u2014too much friction will frustrate users, while too little will be ignored. Ultimately, this vision recasts AI as not just efficient tools but as catalysts for cultivating essential human skills like critical reflection and reasoning.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-should-help-users-monitor-evaluate-and-control-their-thinking\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/the-metacognitive-demands-and-opportunities-of-generative-ai\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI should help users monitor, evaluate, and control their thinking<\/a><\/h5>\n\n\n\n<p>AI tools not only alter tasks but also transform how users think about their own thinking, imposing new metacognitive demands. Novice users often over-rely on AI, for example, resulting in misaligned confidence and limited reflection on errors. We propose two design strategies. Systems should boost user metacognition through planning aids, feedback cues, and support for reflective questioning. Second, systems should reduce metacognitive demand by creating interfaces\u2014such as task-specific prompting wizards and uncertainty-aware outputs\u2014that lighten the burden. Metacognition offers a coherent framework for understanding the usability challenges posed by GenAI, enabling us to offer research and design directions to advance human-GenAI interaction.<\/p>\n\n\t\t\t\t<\/span>\n\t\t\t<\/div>\n\t\t\t<button\n\t\t\t\tclass=\"action-trigger glyph-prepend mt-2 mb-0 show-more-show-less-toggle\"\n\t\t\t\taria-expanded=\"false\"\n\t\t\t\tdata-show-less-text=\"Show less\"\n\t\t\t\ttype=\"button\"\n\t\t\t\taria-controls=\"show-more-show-less-toggle-1\"\n\t\t\t\taria-label=\"Show more content\"\n\t\t\t\tdata-alternate-aria-label=\"Show less content\">\n\t\t\t\tShow more\t\t\t<\/button>\n\t\t<\/div>\n\t<\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n<div class=\"wp-block-msr-show-more\">\n\t<div class=\"bg-neutral-100 p-5\">\n\t\t<div class=\"show-more-show-less\">\n\t\t\t<div>\n\t\t\t\t<span>\n\t\t\t\t\t\n\n<h4 class=\"wp-block-heading\" id=\"ai-reflection-on-meeting-goals-before-during-and-between-meetings-improves-effectiveness-1\">AI reflection on meeting goals before, during, and between meetings improves effectiveness<\/h4>\n\n\n\n\t\t\t\t<\/span>\n\t\t\t\t<span id=\"show-more-show-less-toggle-2\" class=\"show-more-show-less-toggleable-content\">\n\t\t\t\t\t\n\n\n\n<h5 class=\"wp-block-heading\" id=\"inefficiencies-arise-from-misaligned-mental-models-of-meeting-goals-1\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/mental-models-of-meeting-goals-supporting-intentionality-in-meeting-technologies\/\">Inefficiencies arise from misaligned mental models of meeting goals<\/a><\/h5>\n\n\n\n<p>The formative study for this series of papers explored meeting inefficiencies as a key blocker to productivity (WTI2024). Interviews with knowledge workers (N=21) found that two dominant models of meetings emerge: meetings as means to achieve explicit, outcome-driven goals versus meetings as ends in themselves for fostering discussion and connection. This goal ambiguity affects scheduling, participation, and evaluation, and existing meeting tools lack the structure to capture dynamic intentions. We recommend embedding explicit goal fields, facilitating dynamic goal negotiation, and post-meeting checks to support co-constructed meeting purposes.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-can-generate-goal-driven-interfaces-to-improve-meeting-effectiveness-1\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/the-coexplorer-technology-probe-a-generative-ai-powered-adaptive-interface-to-support-intentionality-in-planning-and-running-video-meetings\/\">AI can generate goal-driven interfaces to improve meeting effectiveness<\/a><\/h5>\n\n\n\n<p>CoExplorer is a generative-AI meeting system that turns the plain text of a calendar invite into a concrete goal, sequenced phases, and phase-specific workspaces. Before the call, it asks invitees to vote on what matters, then refines agendas and resources accordingly. During the call, it monitors speech, proposes phase transitions, and automatically tiles relevant apps, keeping discussion, reference, and task spaces synchronized. Professionals (N=26) reported that this scaffolding could reduce set-up effort and sharpen shared focus. Participants liked the adaptive layouts and the ability to spotlight disputed topics, but warned that hidden automation risks eroding trust and agency and that abrupt layout shifts can jar social norms.&nbsp; Recommendations include exploration of tools for helping choose what needs to be discussed synchronously versus asynchronously, making system logic legible, and requiring lightweight human-on-the-loop confirmation for critical changes.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-assisted-reflection-before-meetings-improves-goal-setting-and-effectiveness-1\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/what-does-success-look-like-catalyzing-meeting-intentionality-with-ai-assisted-prospective-reflection\/\">AI-assisted reflection before meetings improves goal setting and effectiveness<\/a><\/h5>\n\n\n\n<p>We explore users\u2019 reactions to a Meeting Purpose Assistant (MPA), an AI tool that prompts organisers and attendees to articulate intended goals, challenges, and success criteria before meetings, then returns a concise, shareable summary. Participants (N=18) using the MPA uncovered overlooked concerns, helped prioritize topics, and sometimes led to meeting restructuring or cancellations, while reducing anxiety and increasing ownership. Recommendations include integrating brief, role-sensitive reflection prompts into existing calendar and meeting tools to enhance intentionality and alignment.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-assisted-reflection-on-goals-during-meetings-can-keep-meetings-on-track-2\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/are-we-on-track-ai-assisted-active-and-passive-goal-reflection-during-meetings\/\">AI-assisted reflection on goals during meetings can keep meetings on track<\/a><\/h5>\n\n\n\n<p>We explored how AI nudges can keep real-world meetings aligned with their goals. Professionals (N=15) compared two probes: an Ambient Visualization that passively maps emerging topics to stated goals, and an Interactive Questioning agent that actively interrupts when no goal is set or discussion drifts. Participants said clarifying goals first enabled them to judge when talk was off-track; they welcomed the unobtrusive visual cues of the passive probe, but found active pop-ups effective yet socially disruptive. Recommendations include tuning intervention strength, balancing democratic input with efficiency, and preserving user control to sustain intentional, goal-oriented meetings.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-can-surface-temporal-context-to-maintain-goal-continuity-1\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/designing-interfaces-that-support-temporal-work-across-meetings-with-generative-ai\/\">AI can surface temporal context to maintain goal continuity<\/a><\/h5>\n\n\n\n<p>We explored how AI can strengthen the goal reflection between past and future meetings. By mapping sets of meetings, we found that recurring meetings served as hubs that knit multiple projects together across varied timescales. From this insight we developed a two-axis framework that visualises \u201cobjective\u201d time against \u201csubjective\u201d importance, revealing under-supported regions where tools could help. Using AI to fuse transcripts, recordings and slides, we present three prototypes: Instant Recaps (lightweight reflection moments), Adaptive Meeting Handoff (contextual summaries between back-to-back meetings), and Project Browsers (long-horizon overviews). Together they illustrate design principles of aligning support to both available time and cognitive need, blending AI outputs with human annotations, and maintaining visibility of information to carry goals forward.<\/p>\n\n\t\t\t\t<\/span>\n\t\t\t<\/div>\n\t\t\t<button\n\t\t\t\tclass=\"action-trigger glyph-prepend mt-2 mb-0 show-more-show-less-toggle\"\n\t\t\t\taria-expanded=\"false\"\n\t\t\t\tdata-show-less-text=\"Show less\"\n\t\t\t\ttype=\"button\"\n\t\t\t\taria-controls=\"show-more-show-less-toggle-2\"\n\t\t\t\taria-label=\"Show more content\"\n\t\t\t\tdata-alternate-aria-label=\"Show less content\">\n\t\t\t\tShow more\t\t\t<\/button>\n\t\t<\/div>\n\t<\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n<div class=\"wp-block-msr-show-more\">\n\t<div class=\"bg-neutral-100 p-5\">\n\t\t<div class=\"show-more-show-less\">\n\t\t\t<div>\n\t\t\t\t<span>\n\t\t\t\t\t\n\n<h4 class=\"wp-block-heading\" id=\"creativity-with-ai-should-focus-on-thoughtful-curation-and-orchestration\">Creativity with AI should focus on thoughtful curation and orchestration<\/h4>\n\n\n\n\t\t\t\t<\/span>\n\t\t\t\t<span id=\"show-more-show-less-toggle-3\" class=\"show-more-show-less-toggleable-content\">\n\t\t\t\t\t\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-shifts-human-creative-labour-from-mere-production-to-curation-of-meaning\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/exploring-perspectives-on-the-impact-of-artificial-intelligence-on-the-creativity-of-knowledge-work-beyond-mechanised-plagiarism-and-stochastic-parrots\/\">AI shifts human creative labour from mere production to curation of meaning<\/a><\/h5>\n\n\n\n<p>While current debates on generative AI often reduce creativity to issues of originality, plagiarism, and uncredited remixing, this paper contends that true creativity arises from the human acts of selection, framing, and integration of machine-generated material\u2014thus shifting creative labor from mere production of artefacts to thoughtful curation of meaning, raising essential questions of authorship and credit, and urging the design of interfaces and policies that foreground human judgment in meaning-making.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"creatives-orchestrate-not-automate-with-ai\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"http:\/\/Evolving Roles and Workflows of Creative Practitioners in the Age of Generative AI\">Creatives orchestrate, not automate, with AI<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/h5>\n\n\n\n<p>We explored how creative professionals integrate AI into their work and how this reshapes their roles, workflows, and expectations. Practitioners (N=31) increasingly see themselves as orchestrators\u2014managing tasks, tools, and ideas\u2014rather than simply executing discrete steps. They adopt AI not to automate creativity, but to support processes like brainstorming, goal refinement, and iteration. While they value the efficiency and inspiration GenAI offers, they also face challenges: articulating goals, maintaining coherence across fragmented tools, and aligning outputs with intent. Creatives want to retain agency throughout and, in some cases, desire emotionally aware systems that respond empathetically in high-stakes phases. Future Creativity Support Tools (CSTs) should preserve human control while enhancing creative flow, context management, and co-creation.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"conversations-with-multiple-agents-can-boost-ideation-diversity\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/yes-and-a-generative-ai-multi-agent-framework-for-enhancing-diversity-of-thought-in-individual-ideation-for-problem-solving-through-confidence-based-agent-turn-taking\/\">Conversations with multiple agents can boost ideation diversity<\/a><\/h5>\n\n\n\n<p>Diversity of thought is critical in creative problem-solving, yet solo workers often lack access to alternative perspectives. We introduce YES AND, a multi-agent GenAI framework simulating diverse expert perspectives to enrich individual ideation. The confidence-based turn-taking model allows for the user and the agents to interact using natural conversation, which in turn, allows for organic development of ideas using diverse viewpoints. We highlight several design trade-offs: maintaining coherence across agent contributions, balancing agent autonomy with user control, and avoiding response verbosity.<\/p>\n\n\t\t\t\t<\/span>\n\t\t\t<\/div>\n\t\t\t<button\n\t\t\t\tclass=\"action-trigger glyph-prepend mt-2 mb-0 show-more-show-less-toggle\"\n\t\t\t\taria-expanded=\"false\"\n\t\t\t\tdata-show-less-text=\"Show less\"\n\t\t\t\ttype=\"button\"\n\t\t\t\taria-controls=\"show-more-show-less-toggle-3\"\n\t\t\t\taria-label=\"Show more content\"\n\t\t\t\tdata-alternate-aria-label=\"Show less content\">\n\t\t\t\tShow more\t\t\t<\/button>\n\t\t<\/div>\n\t<\/div>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\"><div class=\"wp-block-msr-show-more\">\n\t<div class=\"bg-neutral-100 p-5\">\n\t\t<div class=\"show-more-show-less\">\n\t\t\t<div>\n\t\t\t\t<span>\n\t\t\t\t\t\n\n<h4 class=\"wp-block-heading\" id=\"integrating-ai-into-knowledge-work-requires-a-balance-of-design-and-change-management\">Integrating AI into Knowledge Work&nbsp;requires a balance of design and change management<\/h4>\n\n\n\n\t\t\t\t<\/span>\n\t\t\t\t<span id=\"show-more-show-less-toggle-4\" class=\"show-more-show-less-toggleable-content\">\n\t\t\t\t\t\n\n\n\n<h5 class=\"wp-block-heading\" id=\"knowledge-workers-balance-ai-efficiency-with-critical-oversight\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/the-impact-of-generative-ai-on-critical-thinking-self-reported-reductions-in-cognitive-effort-and-confidence-effects-from-a-survey-of-knowledge-workers\/\">Knowledge Workers balance AI efficiency with critical oversight<\/a><\/h5>\n\n\n\n<p>We surveyed 319 professionals across industries showing that workflows are shifting from direct task execution to overseeing and integrating AI-generated content. This change improves efficiency and reduces perceived effort but risks diminishing independent judgment\u2014especially when users overly trust AI due to low self-confidence. Conversely, those confident in their abilities scrutinize and refine AI outputs more critically. We identify obstacles such as time pressure, limited evaluative skills, and low awareness of when deeper analysis is needed. Design enhancements like prompts for justification, skill-building feedback, and structured critical reflection tools should ensure that gains in efficiency do not come at the cost of independent, critical thought.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-may-introduce-inefficiencies-unless-it-is-integrated-deliberately\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/ironies-of-generative-ai-understanding-and-mitigating-productivity-loss-in-human-ai-interactions\/\">AI may introduce inefficiencies unless it is integrated deliberately<\/a><\/h5>\n\n\n\n<p>AI is frequently touted as a productivity enhancer, yet it can introduce inefficiencies, confusion, and added cognitive load\u2014reflecting the \u201cironies of automation\u201d from human factors research. We identify four main challenges: a shift from creative production to increased supervisory demands; workflow disruptions breaking established rhythms; frequent task interruptions from AI suggestions; and a polarization effect, where simple tasks become easier while complex ones grow more challenging. These problems stem largely from interface design flaws and skewed user expectations, rather than from the models themselves. These issues may be remedied by principles such as continuous feedback, personalization, ecological interface design, and clear task allocation to support situational awareness and reduce unintended delegation, urging a more deliberate integration of AI into workflows.<\/p>\n\n\t\t\t\t<\/span>\n\t\t\t<\/div>\n\t\t\t<button\n\t\t\t\tclass=\"action-trigger glyph-prepend mt-2 mb-0 show-more-show-less-toggle\"\n\t\t\t\taria-expanded=\"false\"\n\t\t\t\tdata-show-less-text=\"Show less\"\n\t\t\t\ttype=\"button\"\n\t\t\t\taria-controls=\"show-more-show-less-toggle-4\"\n\t\t\t\taria-label=\"Show more content\"\n\t\t\t\tdata-alternate-aria-label=\"Show less content\">\n\t\t\t\tShow more\t\t\t<\/button>\n\t\t<\/div>\n\t<\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n<div class=\"wp-block-msr-show-more\">\n\t<div class=\"bg-neutral-100 p-5\">\n\t\t<div class=\"show-more-show-less\">\n\t\t\t<div>\n\t\t\t\t<span>\n\t\t\t\t\t\n\n<h4 class=\"wp-block-heading\" id=\"ai-can-scaffold-decision-making-processes\">AI can scaffold decision-making processes<\/h4>\n\n\n\n\t\t\t\t<\/span>\n\t\t\t\t<span id=\"show-more-show-less-toggle-5\" class=\"show-more-show-less-toggleable-content\">\n\t\t\t\t\t\n\n\n\n<h5 class=\"wp-block-heading\" id=\"provocations-drive-diverse-ai-shortlisting-decisions\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/arxiv.org\/abs\/2501.17247\">Provocations drive diverse AI shortlisting decisions<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/h5>\n\n\n\n<p>Shortlisting tasks in knowledge work demand structured yet subjective critical thinking, but relying solely on AI risks uncritical \u201cmechanised convergence\u201d of rankings. We explored whether brief textual provocations attached to AI suggestions can enhance reflective thinking. In a between-subjects study, participants (n=24) exposed to these cues engaged more in evaluation, modification, and questioning\u2014resulting in more varied shortlists. Provocations function best as \u201cmicroboundaries\u201d\u2014small interruptions that nudge users toward reflective cognition without derailing progress.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-decision-support-should-balance-delivering-solutions-with-supporting-reasoning\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/ai-help-me-think-but-for-myself-assisting-people-in-complex-decision-making-by-providing-different-kinds-of-cognitive-support\/\">AI decision-support should balance delivering solutions with supporting reasoning<\/a>&nbsp;<\/h5>\n\n\n\n<p>We compared two paradigms in AI-assisted decision-making: RecommendAI, which offers direct suggestions akin to common recommendation-based decision-support tools, and ExtendAI, which prompts users to articulate their decision rationale before providing reflective feedback. In a simulated investment scenario (N=21), ExtendAI led to more diversified portfolios and was perceived as more supportive of users&#8217; reasoning, but demanded greater cognitive effort. Conversely, RecommendAI provided more novel ideas but was often less aligned with users&#8217; thought processes. We recommend that AI systems should follow a balanced approach that dynamically alternates between inspiring users with new ideas and scaffolding their reasoning, reframing AI decision support as a collaborative partnership.<\/p>\n\n\t\t\t\t<\/span>\n\t\t\t<\/div>\n\t\t\t<button\n\t\t\t\tclass=\"action-trigger glyph-prepend mt-2 mb-0 show-more-show-less-toggle\"\n\t\t\t\taria-expanded=\"false\"\n\t\t\t\tdata-show-less-text=\"Show less\"\n\t\t\t\ttype=\"button\"\n\t\t\t\taria-controls=\"show-more-show-less-toggle-5\"\n\t\t\t\taria-label=\"Show more content\"\n\t\t\t\tdata-alternate-aria-label=\"Show less content\">\n\t\t\t\tShow more\t\t\t<\/button>\n\t\t<\/div>\n\t<\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n<div class=\"wp-block-msr-show-more\">\n\t<div class=\"bg-neutral-100 p-5\">\n\t\t<div class=\"show-more-show-less\">\n\t\t\t<div>\n\t\t\t\t<span>\n\t\t\t\t\t\n\n<h4 class=\"wp-block-heading\" id=\"prompting-and-evaluating-ai-outputs-is-a-new-skill-and-a-ui-opportunity\">Prompting and evaluating AI outputs is a new skill and a UI opportunity<\/h4>\n\n\n\n\t\t\t\t<\/span>\n\t\t\t\t<span id=\"show-more-show-less-toggle-6\" class=\"show-more-show-less-toggleable-content\">\n\t\t\t\t\t\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-should-enhance-human-auditing-for-reliable-oversight\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/co-audit-tools-to-help-humans-double-check-ai-generated-content\/\">AI should enhance human auditing for reliable oversight<\/a><\/h5>\n\n\n\n<p>Large language models now generate complex outputs, posing challenges for accurate error detection. We introduce the concept of&nbsp; \u201cco-audit,\u201d where specialized interfaces support people assessing AI-generated content. We explore the cognitive and contextual requirements for effective co-auditing, identifying key factors like error visibility, mistake cost, audit frequency, and solution ambiguity. We stress that co-audit tools must balance detail with cognitive load and provide uncertainty signals, adjustable rationales, and timely human intervention. Their framework guides the responsible design of AI auditing systems for high-stakes domains.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-enhances-sensemaking-in-data-analysis-but-demands-better-prompt-guidance\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/its-like-a-rubber-duck-that-talks-back-understanding-generative-ai-assisted-data-analysis-workflows-through-a-participatory-prompting-study\/\">AI enhances sensemaking in data analysis but demands better prompt guidance<\/a><\/h5>\n\n\n\n<p>We investigated how everyday spreadsheet users use AI to make sense of data. Working with 15 volunteers (students, analysts, engineers and others), we found that AI sped up the information-foraging loop by quickly surfacing relevant datasets, code snippets and step-by-step instructions, and enriched the sense-making loop by suggesting fresh hypotheses or analysis tactics. Yet progress often stalled when users struggled to phrase precise prompts, supply sufficient context or judge whether long, sometimes vague answers were correct. People coped by iterating queries, chasing citations, or testing AI-generated code, but they worried about errors and \u201challucinations.\u201d Future AI tools should help us craft richer prompts, flag uncertain outputs, teach verification skills and integrate smoothly with familiar apps like Excel so we can keep control while enjoying AI\u2019s speed and creativity.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"dynamic-ui-for-ai-lowers-the-barrier-for-steering-ai-outputs\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/dynamic-prompt-middleware-contextual-prompt-refinement-controls-for-comprehension-tasks\/\">Dynamic UI for AI lowers the barrier for steering AI outputs<\/a><\/h5>\n\n\n\n<p>Effective prompting remains a major barrier for AI users, particularly when outputs must be contextual, accurate, and controllable. We explore how UI controls that focus on task decomposition can facilitate better AI outputs. We compare two middleware prototypes: one offering static, generic prompt options, and another providing dynamic, context-refined controls generated from the user\u2019s initial prompt. In a study with 16 participants, the dynamic controls lowered barriers to providing context, improved user perception of control, and provided guidance for steering the AI. This reduces prompt iteration, but at some cost of increased cognitive load and reduced predictability. Dynamically generated UI that can be customised to context is feasible, and can empower users to shape AI behaviour outside of the chat paradigm.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-cannot-explain-itself\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/large-language-models-cannot-explain-themselves\/\">AI cannot explain itself<\/a><\/h5>\n\n\n\n<p>Large language models produce fluent text but cannot introspectively explain their outputs. Instead, they generate \u201cexoplanations\u201d\u2014post hoc, plausible justifications that do not reflect genuine reasoning. This misleading behavior risks users overtrusting AI outputs, especially in sensitive contexts. The paper distinguishes between mechanistic explanations and these synthetic rationales, urging designers to incorporate features like uncertainty signals, structured responses, and external auditing to foster genuine critical reflection rather than illusionary transparency.<\/p>\n\n\t\t\t\t<\/span>\n\t\t\t<\/div>\n\t\t\t<button\n\t\t\t\tclass=\"action-trigger glyph-prepend mt-2 mb-0 show-more-show-less-toggle\"\n\t\t\t\taria-expanded=\"false\"\n\t\t\t\tdata-show-less-text=\"Show less\"\n\t\t\t\ttype=\"button\"\n\t\t\t\taria-controls=\"show-more-show-less-toggle-6\"\n\t\t\t\taria-label=\"Show more content\"\n\t\t\t\tdata-alternate-aria-label=\"Show less content\">\n\t\t\t\tShow more\t\t\t<\/button>\n\t\t<\/div>\n\t<\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n<div class=\"wp-block-msr-show-more\">\n\t<div class=\"bg-neutral-100 p-5\">\n\t\t<div class=\"show-more-show-less\">\n\t\t\t<div>\n\t\t\t\t<span>\n\t\t\t\t\t\n\n<h4 class=\"wp-block-heading\" id=\"ai-for-education-and-learning-requires-active-integration\">AI for education and learning requires active integration<\/h4>\n\n\n\n\t\t\t\t<\/span>\n\t\t\t\t<span id=\"show-more-show-less-toggle-7\" class=\"show-more-show-less-toggleable-content\">\n\t\t\t\t\t\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-in-higher-ed-needs-policies-designed-around-balanced-usage\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/the-new-calculator-practices-norms-and-implications-of-generative-ai-in-higher-education\/\">AI in Higher Ed needs policies designed around balanced usage<\/a><\/h5>\n\n\n\n<p>Generative AI\u2019s rapid uptake in universities has outpaced clear institutional guidance, creating uncertainty for both students and educators. Interviews with 26 students and 11 educators from two UK universities reveal that students employ AI in varied, context-sensitive ways\u2014such as summarizing material, generating examples, and exploring topics\u2014to boost efficiency, despite concerns over plagiarism, dependency, and skill erosion. Educators, meanwhile, see both promise and risk yet lack consensus on how to adapt policies and assessments. The study highlights a misalignment between student practices and educator expectations, calling for explicit training, revised assessment formats, and policies that align with real-world AI use while upholding academic integrity.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"note-taking-plus-llm-use-balances-learning-and-engagement\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/papers.ssrn.com\/sol3\/papers.cfm?abstract_id=5095149\">Note-taking plus LLM use balances learning and engagement<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/h5>\n\n\n\n<p>As students increasingly use large language models (LLMs) for learning, educators face urgent questions about how such tools affect core outcomes like comprehension and retention. We conducted experiments comparing LLMs, traditional note-taking, and combined approaches on student comprehension and retention. We found note-taking improved long-term retention and comprehension compared to LLM use alone, while combined note-taking and LLM use balanced comprehension benefits with student engagement. We identified \u201cprompting archetypes,\u201d capturing diverse strategies students used when interacting with the LLM. These behaviours correlated with learning outcomes, highlighting the importance of metacognitive prompting skills. This highlights the importance of actively involving learners in processing information rather than passively receiving AI-generated summaries.<\/p>\n\n\n\n<h5 class=\"wp-block-heading\" id=\"ai-may-make-programmers-naming-choices-more-predictable\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/predictability-of-identifier-naming-with-copilot-a-case-study-for-mixed-initiative-programming-tools\/\">AI may make programmers&#8217; naming choices more predictable<\/a><\/h5>\n\n\n\n<p>We looked at how GitHub Copilot, an AI coding assistant, influences the names programmers choose for things like classes and methods in their code. Naming is important in programming because it helps others understand what the code does. Participants (N=12) completed coding tasks with Copilot turned on, turned off, or showing suggestions without letting them auto-accept. We found that when Copilot was active\u2014even if its suggestions had to be typed out\u2014people chose more predictable, less unique names. When Copilot was off, names were more varied and sometimes more descriptive. AI tools like Copilot can nudge programmers toward using common names, which might save time but reduce creativity or clarity. Future AI tools should offer more distinctive suggestions and help refine names over time. This could help balance AI\u2019s speed with the human need for meaningful, understandable code.<\/p>\n\n\t\t\t\t<\/span>\n\t\t\t<\/div>\n\t\t\t<button\n\t\t\t\tclass=\"action-trigger glyph-prepend mt-2 mb-0 show-more-show-less-toggle\"\n\t\t\t\taria-expanded=\"false\"\n\t\t\t\tdata-show-less-text=\"Show less\"\n\t\t\t\ttype=\"button\"\n\t\t\t\taria-controls=\"show-more-show-less-toggle-7\"\n\t\t\t\taria-label=\"Show more content\"\n\t\t\t\tdata-alternate-aria-label=\"Show less content\">\n\t\t\t\tShow more\t\t\t<\/button>\n\t\t<\/div>\n\t<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n\n\n<h2 class=\"wp-block-heading\" id=\"what-we-know-tools-for-thought-research-overview-volume-1\">What We Know: Tools for Thought Research Overview Volume 1&nbsp;<\/h2>\n\n\n\n<p>The Tools for Thought project began in June 2024 with the goal of exploring how to protect and augment human thought in knowledge work. This page documents the first volume of our research, spanning theoretical perspectives, empirical studies, and design prototypes.<\/p>\n\n\n\n<p>We begin with our core belief that AI must support critical thinking and metacognition. We then move to three areas of exploration: <em>AI and Knowledge Work<\/em> (covering AI in knowledge work, decision-making, meetings, creativity, and programming); <em>AI and Education<\/em>, and <em>Prompting AI and Evaluating AI Outputs<\/em>.<\/p>\n\n\n\n<p>We conclude with synthesized early design implications.&nbsp;<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"ai-must-support-critical-thinking-and-metacognition\">AI Must Support Critical Thinking and Metacognition&nbsp;<\/h3>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"critical-thinking-ai-should-challenge-not-obey-13\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/ai-should-challenge-not-obey\/\" target=\"_blank\" rel=\"noreferrer noopener\">Critical Thinking: AI should challenge, not obey<\/a> [13]&nbsp;<\/h4>\n\n\n\n<p>Rather than optimizing solely for productivity or user satisfaction, AI systems should trigger deeper reflection and more rigorous decision-making in people \u2014exposing flawed reasoning, asking probing questions, and offering alternative perspectives. Strategies such as argument mapping, adversarial prompting, and adaptive friction levels should balance engagement with user comfort\u2014too much friction will frustrate users, while too little will be ignored. Ultimately, this vision recasts AI as not just efficient tools but as catalysts for cultivating essential human skills like critical reflection and reasoning.&nbsp;&nbsp;&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"metacognition-ai-should-help-users-monitor-evaluate-and-control-their-thinking-20\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/the-metacognitive-demands-and-opportunities-of-generative-ai\/\" target=\"_blank\" rel=\"noreferrer noopener\">Metacognition: AI should help users monitor, evaluate, and control their thinking<\/a> [20]<\/h4>\n\n\n\n<p>AI tools not only alter tasks but also transform how users think about their own thinking, imposing new metacognitive demands. Novice users often over-rely on AI, for example, resulting in misaligned confidence and limited reflection on errors. We propose two design strategies. Systems should boost user metacognition through planning aids, feedback cues, and support for reflective questioning. Second, systems should reduce metacognitive demand by creating interfaces\u2014such as task-specific prompting wizards and uncertainty-aware outputs\u2014that lighten the burden. Metacognition offers a coherent framework for understanding the usability challenges posed by GenAI, enabling us to offer research and design directions to advance human-GenAI interaction.<strong>&nbsp;<\/strong>&nbsp;<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"ai-and-knowledge-work\">AI and Knowledge Work&nbsp;<\/h3>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"integrating-ai-into-knowledge-work\">Integrating AI into Knowledge Work&nbsp;<\/h4>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"knowledge-workers-balance-ai-efficiency-with-critical-oversight-8\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/the-impact-of-generative-ai-on-critical-thinking-self-reported-reductions-in-cognitive-effort-and-confidence-effects-from-a-survey-of-knowledge-workers\/\" target=\"_blank\" rel=\"noreferrer noopener\">Knowledge Workers balance AI efficiency with critical oversight<\/a> [8]<\/h4>\n\n\n\n<p>We surveyed 319 professionals across industries showing that workflows are shifting from direct task execution to overseeing and integrating AI-generated content. This change improves efficiency and reduces perceived effort but risks diminishing independent judgment\u2014especially when users overly trust AI due to low self-confidence. Conversely, those confident in their abilities scrutinize and refine AI outputs more critically. We identify obstacles such as time pressure, limited evaluative skills, and low awareness of when deeper analysis is needed. Design enhancements like prompts for justification, skill-building feedback, and structured critical reflection tools should ensure that gains in efficiency do not come at the cost of independent, critical thought.&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"the-irony-of-ai-is-that-it-can-introduce-inefficiencies-unless-it-is-integrated-deliberately-19\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/ironies-of-generative-ai-understanding-and-mitigating-productivity-loss-in-human-ai-interactions\/\" target=\"_blank\" rel=\"noreferrer noopener\">The irony of AI is that it can introduce inefficiencies unless it is integrated deliberately<\/a> [19]&nbsp;<\/h4>\n\n\n\n<p>AI is frequently touted as a productivity enhancer, yet it can introduce inefficiencies, confusion, and added cognitive load\u2014reflecting the \u201cironies of automation\u201d from human factors research. We identify four main challenges: a shift from creative production to increased supervisory demands; workflow disruptions breaking established rhythms; frequent task interruptions from AI suggestions; and a polarization effect, where simple tasks become easier while complex ones grow more challenging. These problems stem largely from interface design flaws and skewed user expectations, rather than from the models themselves. These issues may be remedied by principles such as continuous feedback, personalization, ecological interface design, and clear task allocation to support situational awareness and reduce unintended delegation, urging a more deliberate integration of AI into workflows.&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"decision-making\">Decision-Making&nbsp;<\/h4>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"provocations-drive-reflective-diverse-ai-shortlisting-decisions-4\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/arxiv.org\/abs\/2501.17247\" target=\"_blank\" rel=\"noopener noreferrer\">Provocations drive reflective, diverse AI shortlisting decisions<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> [4]&nbsp;&nbsp;<\/h4>\n\n\n\n<p>Shortlisting tasks in knowledge work demand structured yet subjective critical thinking, but relying solely on AI risks uncritical \u201cmechanised convergence\u201d of rankings. We explored whether brief textual provocations attached to AI suggestions can enhance reflective thinking. In a between-subjects study, participants (n=24) exposed to these cues engaged more in evaluation, modification, and questioning\u2014resulting in more varied shortlists. Provocations function best as \u201cmicroboundaries\u201d\u2014small interruptions that nudge users toward reflective cognition without derailing progress.&nbsp;&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"ai-decision-support-should-balance-delivering-solutions-with-supporting-reasoning-12\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/ai-help-me-think-but-for-myself-assisting-people-in-complex-decision-making-by-providing-different-kinds-of-cognitive-support\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI decision-support should balance delivering solutions with supporting reasoning<\/a> [12]&nbsp;<\/h4>\n\n\n\n<p>We compared two paradigms in AI-assisted decision-making: RecommendAI, which offers direct suggestions akin to common recommendation-based decision-support tools, and ExtendAI, which prompts users to articulate their decision rationale before providing reflective feedback. In a simulated investment scenario (N=21), ExtendAI led to more diversified portfolios and was perceived as more supportive of users&#8217; reasoning, but demanded greater cognitive effort. Conversely, RecommendAI provided more novel ideas but was often less aligned with users&#8217; thought processes. We recommend that AI systems should follow a balanced approach that dynamically alternates between inspiring users with new ideas and scaffolding their reasoning, reframing AI decision support as a collaborative partnership.&nbsp;<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"intentional-meetings\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/project\/intentional-meetings\/\" target=\"_blank\" rel=\"noreferrer noopener\">Intentional Meetings<\/a>&nbsp;<\/h3>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"inefficiencies-arise-from-misaligned-mental-models-of-meeting-goals-16\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/01\/chi24-774-authorcameraready.pdf\" target=\"_blank\" rel=\"noreferrer noopener\">Inefficiencies arise from misaligned mental models of meeting goals<\/a> [16]&nbsp;<\/h4>\n\n\n\n<p>The formative study for this series of papers explored meeting inefficiencies as a key blocker to productivity (WTI2024). Interviews with knowledge workers (N=21) found that two dominant models of meetings emerge: meetings as means to achieve explicit, outcome-driven goals versus meetings as ends in themselves for fostering discussion and connection. This goal ambiguity affects scheduling, participation, and evaluation, and existing meeting tools lack the structure to capture dynamic intentions. We recommend embedding explicit goal fields, facilitating dynamic goal negotiation, and post-meeting checks to support co-constructed meeting purposes.&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"ai-can-generate-goal-driven-interfaces-to-improve-meeting-effectiveness-11\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/the-coexplorer-technology-probe-a-generative-ai-powered-adaptive-interface-to-support-intentionality-in-planning-and-running-video-meetings\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI can generate goal-driven interfaces to improve meeting effectiveness<\/a> [11]&nbsp;&nbsp;<\/h4>\n\n\n\n<p>CoExplorer is a generative-AI meeting system that turns the plain text of a calendar invite into a concrete goal, sequenced phases, and phase-specific workspaces. Before the call, it asks invitees to vote on what matters, then refines agendas and resources accordingly. During the call, it monitors speech, proposes phase transitions, and automatically tiles relevant apps, keeping discussion, reference, and task spaces synchronized. Professionals (N=26) reported that this scaffolding could reduce set-up effort and sharpen shared focus. Participants liked the adaptive layouts and the ability to spotlight disputed topics, but warned that hidden automation risks eroding trust and agency and that abrupt layout shifts can jar social norms.&nbsp; Recommendations include exploration of tools for helping choose what needs to be discussed synchronously versus asynchronously, making system logic legible, and requiring lightweight human-on-the-loop confirmation for critical changes.&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"ai-assisted-reflection-before-meetings-improves-goal-setting-and-effectiveness-17\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/what-does-success-look-like-catalyzing-meeting-intentionality-with-ai-assisted-prospective-reflection\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI-assisted reflection before meetings improves goal setting and effectiveness<\/a> [17]&nbsp;<\/h4>\n\n\n\n<p>We explore users\u2019 reactions to a Meeting Purpose Assistant (MPA), an AI tool that prompts organisers and attendees to articulate intended goals, challenges, and success criteria before meetings, then returns a concise, shareable summary. Participants (N=18) using the MPA uncovered overlooked concerns, helped prioritize topics, and sometimes led to meeting restructuring or cancellations, while reducing anxiety and increasing ownership. Recommendations include integrating brief, role-sensitive reflection prompts into existing calendar and meeting tools to enhance intentionality and alignment.&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"ai-assisted-reflection-on-goals-during-meetings-can-keep-meetings-on-track-1\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/are-we-on-track-ai-assisted-active-and-passive-goal-reflection-during-meetings\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI-assisted reflection on goals during meetings can keep meetings on track<\/a> [1]&nbsp;&nbsp;<\/h4>\n\n\n\n<p>We explored how AI nudges can keep real-world meetings aligned with their goals. Professionals (N=15) compared two probes: an Ambient Visualization that passively maps emerging topics to stated goals, and an Interactive Questioning agent that actively interrupts when no goal is set or discussion drifts. Participants said clarifying goals first enabled them to judge when talk was off-track; they welcomed the unobtrusive visual cues of the passive probe, but found active pop-ups effective yet socially disruptive. Recommendations include tuning intervention strength, balancing democratic input with efficiency, and preserving user control to sustain intentional, goal-oriented meetings.&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"ai-can-surface-temporal-context-to-maintain-goal-continuity-21\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/designing-interfaces-that-support-temporal-work-across-meetings-with-generative-ai\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI can surface temporal context to maintain goal continuity<\/a> [21]&nbsp;&nbsp;<\/h4>\n\n\n\n<p>We explored how AI can strengthen the goal reflection between past and future meetings. By mapping sets of meetings, we found that recurring meetings served as hubs that knit multiple projects together across varied timescales. From this insight we developed a two-axis framework that visualises \u201cobjective\u201d time against \u201csubjective\u201d importance, revealing under-supported regions where tools could help. Using AI to fuse transcripts, recordings and slides, we present three prototypes: Instant Recaps (lightweight reflection moments), Adaptive Meeting Handoff (contextual summaries between back-to-back meetings), and Project Browsers (long-horizon overviews). Together they illustrate design principles of aligning support to both available time and cognitive need, blending AI outputs with human annotations, and maintaining visibility of information to carry goals forward.&nbsp;&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"creativity\">Creativity&nbsp;<\/h4>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"ai-shifts-human-creative-labour-from-mere-production-to-curation-of-meaning-15\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/exploring-perspectives-on-the-impact-of-artificial-intelligence-on-the-creativity-of-knowledge-work-beyond-mechanised-plagiarism-and-stochastic-parrots\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI shifts human creative labour from mere production to curation of meaning<\/a> [15]&nbsp;<\/h4>\n\n\n\n<p>While current debates on generative AI often reduce creativity to issues of originality, plagiarism, and uncredited remixing, this paper contends that true creativity arises from the human acts of selection, framing, and integration of machine-generated material\u2014thus shifting creative labor from mere production of artefacts to thoughtful curation of meaning, raising essential questions of authorship and credit, and urging the design of interfaces and policies that foreground human judgment in meaning-making.&nbsp;&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"creatives-orchestrate-not-automate-with-ai-10\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/Evolving Roles and Workflows of Creative Practitioners in the Age of Generative AI\" target=\"_blank\" rel=\"noopener noreferrer\">Creatives orchestrate, not automate, with AI<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> [10]&nbsp;<\/h4>\n\n\n\n<p>We explored how creative professionals integrate AI into their work and how this reshapes their roles, workflows, and expectations. Practitioners (N=31) increasingly see themselves as orchestrators\u2014managing tasks, tools, and ideas\u2014rather than simply executing discrete steps. They adopt AI not to automate creativity, but to support processes like brainstorming, goal refinement, and iteration. While they value the efficiency and inspiration GenAI offers, they also face challenges: articulating goals, maintaining coherence across fragmented tools, and aligning outputs with intent. Creatives want to retain agency throughout and, in some cases, desire emotionally aware systems that respond empathetically in high-stakes phases. Future Creativity Support Tools (CSTs) should preserve human control while enhancing creative flow, context management, and co-creation.&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"conversations-with-multiple-agents-can-boost-ideation-diversity-5\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/yes-and-a-generative-ai-multi-agent-framework-for-enhancing-diversity-of-thought-in-individual-ideation-for-problem-solving-through-confidence-based-agent-turn-taking\/\" target=\"_blank\" rel=\"noreferrer noopener\">Conversations with multiple agents can boost ideation diversity<\/a> [5]&nbsp;<\/h4>\n\n\n\n<p>Diversity of thought is critical in creative problem-solving, yet solo workers often lack access to alternative perspectives. We introduce YES AND, a multi-agent GenAI framework simulating diverse expert perspectives to enrich individual ideation. The confidence-based turn-taking model allows for the user and the agents to interact using natural conversation, which in turn, allows for organic development of ideas using diverse viewpoints. We highlight several design trade-offs: maintaining coherence across agent contributions, balancing agent autonomy with user control, and avoiding response verbosity.&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"programming\">Programming&nbsp;<\/h4>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"ai-makes-programmers-naming-choices-more-predictable-9\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/predictability-of-identifier-naming-with-copilot-a-case-study-for-mixed-initiative-programming-tools\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI makes programmers&#8217; naming choices more predictable<\/a> [9]&nbsp;<\/h4>\n\n\n\n<p>We looked at how GitHub Copilot, an AI coding assistant, influences the names programmers choose for things like classes and methods in their code. Naming is important in programming because it helps others understand what the code does. Participants (N=12) completed coding tasks with Copilot turned on, turned off, or showing suggestions without letting them auto-accept. We found that when Copilot was active\u2014even if its suggestions had to be typed out\u2014people chose more predictable, less unique names. When Copilot was off, names were more varied and sometimes more descriptive. AI tools like Copilot can nudge programmers toward using common names, which might save time but reduce creativity or clarity. Future AI tools should offer more distinctive suggestions and help refine names over time. This could help balance AI\u2019s speed with the human need for meaningful, understandable code.&nbsp;<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"ai-in-education-and-learning\">AI in Education and Learning&nbsp;<\/h3>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"ai-in-higher-ed-needs-policies-designed-around-balanced-usage-18\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/the-new-calculator-practices-norms-and-implications-of-generative-ai-in-higher-education\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI in Higher Ed needs policies designed around balanced usage<\/a> [18]&nbsp;<\/h4>\n\n\n\n<p>Generative AI\u2019s rapid uptake in universities has outpaced clear institutional guidance, creating uncertainty for both students and educators. Interviews with 26 students and 11 educators from two UK universities reveal that students employ AI in varied, context-sensitive ways\u2014such as summarizing material, generating examples, and exploring topics\u2014to boost efficiency, despite concerns over plagiarism, dependency, and skill erosion. Educators, meanwhile, see both promise and risk yet lack consensus on how to adapt policies and assessments. The study highlights a misalignment between student practices and educator expectations, calling for explicit training, revised assessment formats, and policies that align with real-world AI use while upholding academic integrity.&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"note-taking-plus-llm-use-balances-learning-and-engagement-7\"><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/papers.ssrn.com\/sol3\/papers.cfm?abstract_id=5095149\" target=\"_blank\" rel=\"noopener noreferrer\">Note-taking plus LLM use balances learning and engagement<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> [7]&nbsp;&nbsp;<\/h4>\n\n\n\n<p>As students increasingly use large language models (LLMs) for learning, educators face urgent questions about how such tools affect core outcomes like comprehension and retention. We conducted experiments comparing LLMs, traditional note-taking, and combined approaches on student comprehension and retention. We found note-taking improved long-term retention and comprehension compared to LLM use alone, while combined note-taking and LLM use balanced comprehension benefits with student engagement. We identified \u201cprompting archetypes,\u201d capturing diverse strategies students used when interacting with the LLM. These behaviours correlated with learning outcomes, highlighting the importance of metacognitive prompting skills. This highlights the importance of actively involving learners in processing information rather than passively receiving AI-generated summaries.&nbsp;<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"prompting-ai-and-evaluating-ai-outputs\">Prompting AI and Evaluating AI Outputs&nbsp;<\/h3>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"ai-should-enhance-human-auditing-for-reliable-oversight-6\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/co-audit-tools-to-help-humans-double-check-ai-generated-content\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI should enhance human auditing for reliable oversight<\/a> [6]&nbsp;<\/h4>\n\n\n\n<p>Large language models now generate complex outputs, posing challenges for accurate error detection. We introduce the concept of&nbsp; \u201cco-audit,\u201d where specialized interfaces support people assessing AI-generated content. We explore the cognitive and contextual requirements for effective co-auditing, identifying key factors like error visibility, mistake cost, audit frequency, and solution ambiguity. We stress that co-audit tools must balance detail with cognitive load and provide uncertainty signals, adjustable rationales, and timely human intervention. Their framework guides the responsible design of AI auditing systems for high-stakes domains.&nbsp;&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"ai-enhances-sensemaking-in-data-analysis-but-demands-better-prompt-guidance-2\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/its-like-a-rubber-duck-that-talks-back-understanding-generative-ai-assisted-data-analysis-workflows-through-a-participatory-prompting-study\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI enhances sensemaking in data analysis but demands better prompt guidance<\/a> [2]&nbsp;<\/h4>\n\n\n\n<p>We investigated how everyday spreadsheet users use AI to make sense of data. Working with 15 volunteers (students, analysts, engineers and others), we found that AI sped up the information-foraging loop by quickly surfacing relevant datasets, code snippets and step-by-step instructions, and enriched the sense-making loop by suggesting fresh hypotheses or analysis tactics. Yet progress often stalled when users struggled to phrase precise prompts, supply sufficient context or judge whether long, sometimes vague answers were correct. People coped by iterating queries, chasing citations, or testing AI-generated code, but they worried about errors and \u201challucinations.\u201d Future AI tools should help us craft richer prompts, flag uncertain outputs, teach verification skills and integrate smoothly with familiar apps like Excel so we can keep control while enjoying AI\u2019s speed and creativity.&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"dynamic-ui-for-ai-lowers-the-barrier-for-steering-ai-outputs-3\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/dynamic-prompt-middleware-contextual-prompt-refinement-controls-for-comprehension-tasks\/\" target=\"_blank\" rel=\"noreferrer noopener\">Dynamic UI for AI lowers the barrier for steering AI outputs<\/a> [3]<\/h4>\n\n\n\n<p>Effective prompting remains a major barrier for AI users, particularly when outputs must be contextual, accurate, and controllable. We explore how UI controls that focus on task decomposition can facilitate better AI outputs. We compare two middleware prototypes: one offering static, generic prompt options, and another providing dynamic, context-refined controls generated from the user\u2019s initial prompt. In a study with 16 participants, the dynamic controls lowered barriers to providing context, improved user perception of control, and provided guidance for steering the AI. This reduces prompt iteration, but at some cost of increased cognitive load and reduced predictability. Dynamically generated UI that can be customised to context is feasible, and can empower users to shape AI behaviour outside of the chat paradigm.&nbsp;&nbsp;<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"ai-cannot-explain-itself-14\"><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/large-language-models-cannot-explain-themselves\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI cannot explain itself<\/a> [14]&nbsp;<\/h4>\n\n\n\n<p>Large language models produce fluent text but cannot introspectively explain their outputs. Instead, they generate \u201cexoplanations\u201d\u2014post hoc, plausible justifications that do not reflect genuine reasoning. This misleading behavior risks users overtrusting AI outputs, especially in sensitive contexts. The paper distinguishes between mechanistic explanations and these synthetic rationales, urging designers to incorporate features like uncertainty signals, structured responses, and external auditing to foster genuine critical reflection rather than illusionary transparency.&nbsp;<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"design-implications\">Design Implications&nbsp;<\/h3>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Support Active Oversight, Not Passive Acceptance. <\/strong>Generative AI tools should keep users actively engaged in evaluating and integrating AI output, rather than simply accepting it.&nbsp;<\/li>\n\n\n\n<li><strong>Make Cognitive Processes Visible. <\/strong>Expose the reasoning, assumptions, or intermediate steps in AI-assisted workflows to help users understand, verify, and correct them.&nbsp;<\/li>\n\n\n\n<li><strong>Prompt and Scaffold Reflection. <\/strong>AI systems can nudge goal setting, re-evaluation, and other forms of metacognitive reflection, especially around meetings and learning.&nbsp;<\/li>\n\n\n\n<li><strong>Design for Temporal Continuity of Thought. <\/strong>Support longitudinal workflows by helping users reconnect with past contexts, decisions, and goals across time.&nbsp;<\/li>\n\n\n\n<li><strong>Adapt to Goals and Roles. <\/strong>AI systems should recognize that different users have different goals, roles, and levels of authority, and adapt interactions accordingly.&nbsp;<\/li>\n\n\n\n<li><strong>Build Generative Interfaces That Afford Goal-Directed Steering.<\/strong> AI interfaces should be easy for users to refine and steer (AI) outcomes using dynamic controls customized to their needs.&nbsp;<\/li>\n\n\n\n<li><strong>Clarify the Nature and Limits of AI Output. <\/strong>Users should be supported in interpreting AI output appropriately, especially avoiding the illusion of understanding.&nbsp;<\/li>\n\n\n\n<li><strong>Foster Meaningful Diversity of Thought, then Convergence. <\/strong>Generative AI systems should enable and encourage divergent thinking rather than narrowing options too early, but then enable convergence on key criteria.&nbsp;<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"references\">References&nbsp;<\/h3>\n\n\n\n<ol class=\"wp-block-list\">\n<li>Xinyue<strong> Chen<\/strong>, Lev Tankelevitch, Rishi Vanukuru, Ava Elizabeth Scott, Payod Panda, and Sean Rintel. 2025. Are We On Track? AI-Assisted Active and Passive Goal Reflection During Meetings. In CHI Conference on Human Factors in Computing Systems (CHI \u201925), April 26-May 1, 2025, Yokohama, Japan. ACM, New York, NY, USA, 22 pages. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3706598.3714052\">https:\/\/doi.org\/10.1145\/3706598.3714052<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Ian<strong> Drosos<\/strong>, Advait Sarkar, Xiaotong Xu, Carina Negreanu, Sean Rintel, and Lev Tankelevitch. 2024. &#8220;It&#8217;s like a rubber duck that talks back&#8221;: Understanding Generative AI-Assisted Data Analysis Workflows through a Participatory Prompting Study. In Proceedings of the 3rd Annual Meeting of the Symposium on Human-Computer Interaction for Work (CHIWORK &#8217;24). Association for Computing Machinery, New York, NY, USA, Article 16, 1\u201321. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3663384.3663389\">https:\/\/doi.org\/10.1145\/3663384.3663389<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Ian<strong> Drosos<\/strong>, Jack Williams, Advait Sarkar, Nicholas Wilson, Sean Rintel, and Payod Panda. 2025. Dynamic Prompt Middleware: Contextual Prompt Refinement Controls for Comprehension Tasks. In CHIWORK \u201925: Proceedings of the 4th Annual Symposium on Human-Computer Interaction for Work (CHIWORK \u201925), June 23\u201325, 2025, Amsterdam, Netherlands. ACM, New York, NY, USA, 37 pages. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3729176.3729203\">https:\/\/doi.org\/10.1145\/3729176.3729203<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Ian<strong> Drosos<\/strong>, Advait Sarkar, Xiaotong (Tone) Xu, and Neil Toronto 2025. \u201cIt makes you think\u201d: Provocations Help Restore Critical Thinking to AI-Assisted Knowledge Work. Preprint. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/arxiv.org\/abs\/2501.17247\">https:\/\/arxiv.org\/abs\/2501.17247<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Pratik<strong> Ghosh<\/strong> and Sean Rintel. 2025. YES AND: A Generative AI Multi-Agent Framework for Enhancing Diversity of Thought in Individual Ideation for Problem-Solving Through Confidence-Based Agent Turn-Taking. In Extended Abstracts of the CHI Conference on Human Factors in Computing Systems (CHI EA \u201925), April 26-May 1, 2025, Yokohama, Japan. ACM, New York, NY, USA, 13 pages. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3706599.3720142\">https:\/\/doi.org\/10.1145\/3706599.3720142<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Andrew D. <strong>Gordon<\/strong>, Carina Negreanu, Jos\u00e9 Cambronero, Rasika Chakravarthy, Ian Drosos, Hao Fang, Bhaskar Mitra, Hannah Richardson, Advait Sarkar, Stephanie Simmons, Jack Williams, and Ben Zorn. 2023. Co-audit: tools to help humans double-check AI-generated content. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.48550\/arXiv.2310.01297\">https:\/\/doi.org\/10.48550\/arXiv.2310.01297<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Pia<strong> Kreijkes<\/strong>, Viktor Kewenig, Martina Kuvalja, Mina Lee, Sylvia Vitello, Jake M. Hofman, Abigail Sellen, Sean Rintel, Daniel G. Goldstein, David M. Rothschild, Lev Tankelevitch, and Tim Oates. 2025. Effects of LLM Use and Note-Taking On Reading Comprehension and Memory: A Randomised Experiment in Secondary Schools. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.2139\/ssrn.5095149\">https:\/\/doi.org\/10.2139\/ssrn.5095149<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Hao-Ping (Hank) <strong>Lee<\/strong>, Advait Sarkar, Lev Tankelevitch, Ian Drosos, Sean Rintel, Richard Banks, and Nicholas Wilson. 2025. The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI &#8217;25). Association for Computing Machinery, New York, NY, USA, Article 1121, 1\u201322. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3706598.3713778\">https:\/\/doi.org\/10.1145\/3706598.3713778<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Michael Jing Long<strong> Lee<\/strong>, Advait Sarkar, Alan F. Blackwell. 2024. Predictability of identifier naming with Copilot: A case study for mixed-initiative programming tools. Proceedings of the 35th Annual Conference of the Psychology of Programming Interest Group (PPIG 2024)<\/li>\n\n\n\n<li>Srishti<strong> Palani<\/strong> and Gonzalo Ramos. 2024. Evolving Roles and Workflows of Creative Practitioners in the Age of Generative AI. In Proceedings of the 16th Conference on Creativity & Cognition (C&C &#8217;24). Association for Computing Machinery, New York, NY, USA, 170\u2013184. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3635636.3656190\">https:\/\/doi.org\/10.1145\/3635636.3656190<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Gun Woo (Warren) <strong>Park<\/strong>, Payod Panda, Lev Tankelevitch, and Sean Rintel. 2024. The CoExplorer Technology Probe: A Generative AI-Powered Adaptive Interface to Support Intentionality in Planning and Running Video Meetings. In Proceedings of the 2024 ACM Designing Interactive Systems Conference (DIS &#8217;24). Association for Computing Machinery, New York, NY, USA, 1638\u20131657. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3643834.3661507\">https:\/\/doi.org\/10.1145\/3643834.3661507<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Leon<strong> Reicherts<\/strong>, Zelun Tony Zhang, Elisabeth von Oswald, Yuanting Liu, Yvonne Rogers, and Mariam Hassib. 2025. AI, Help Me Think\u2014but for Myself: Assisting People in Complex Decision-Making by Providing Different Kinds of Cognitive Support. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI &#8217;25). Association for Computing Machinery, New York, NY, USA, Article 255, 1\u201319. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3706598.3713295\">https:\/\/doi.org\/10.1145\/3706598.3713295<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Advait<strong> Sarkar<\/strong>. 2024. AI Should Challenge, Not Obey. Commun. ACM 67, 10 (October 2024), 18\u201321. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3649404\">https:\/\/doi.org\/10.1145\/3649404<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Advait<strong> Sarkar<\/strong>. 2024. Large Language Models Cannot Explain Themselves. In Proceedings of the ACM CHI 2024 Workshop on Human-Centered Explainable AI, HCXAI at CHI &#8217;24, 2024. Honolulu, HI, USA. Available online: <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/arxiv.org\/abs\/2405.04382\">https:\/\/arxiv.org\/abs\/2405.04382<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Advait<strong> Sarkar<\/strong>. 2023. Exploring Perspectives on the Impact of Artificial Intelligence on the Creativity of Knowledge Work: Beyond Mechanised Plagiarism and Stochastic Parrots. In Proceedings of the 2nd Annual Meeting of the Symposium on Human-Computer Interaction for Work (CHIWORK &#8217;23). Association for Computing Machinery, New York, NY, USA, Article 13, 1\u201317. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3596671.3597650\">https:\/\/doi.org\/10.1145\/3596671.3597650<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Ava Elizabeth<strong> Scott<\/strong>, Lev Tankelevitch, Payod Panda, Rishi Vanukuru, Xinyue Chen, and Sean Rintel. 2025. What Does Success Look Like? Catalyzing Meeting Intentionality with AI-Assisted Prospective Reflection. In CHIWORK \u201925: Proceedings of the 4th Annual Symposium on Human-Computer Interaction for Work (CHIWORK \u201925), June 23\u201325, 2025, Amsterdam, Netherlands. ACM, New York, NY, USA, 33 pages. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3729176.3729204\">https:\/\/doi.org\/10.1145\/3729176.3729204<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Ava Elizabeth<strong> Scott<\/strong>, Lev Tankelevitch, and Sean Rintel. 2024. Mental Models of Meeting Goals: Supporting Intentionality in Meeting Technologies. In Proceedings of the CHI Conference on Human Factors in Computing Systems (CHI \u201924), May 11\u201316, 2024, Honolulu, HI, USA. ACM, New York, NY, USA, 17 pages. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3613904.3642670\">https:\/\/doi.org\/10.1145\/3613904.3642670<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Auste<strong> Simkute<\/strong>, Viktor Kewenig, Abigail Sellen, Sean Rintel, and Lev Tankelevitch. 2025. The New Calculator? Practices, Norms, and Implications of Generative AI in Higher Education. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.48550\/arXiv.2501.08864\">https:\/\/doi.org\/10.48550\/arXiv.2501.08864<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Auste<strong> Simkute<\/strong>, Lev Tankelevitch, Viktor Kewenig, Ava Elizabeth Scott, Abigail Sellen, and Sean Rintel. 2024. \u201cIronies of Generative AI: Understanding and Mitigating Productivity Loss in Human-AI Interaction.\u201d International Journal of Human\u2013Computer Interaction, October, 1\u201322. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1080\/10447318.2024.2405782\">https:\/\/doi.org\/10.1080\/10447318.2024.2405782<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Lev<strong> Tankelevitch<\/strong>, Viktor Kewenig, Auste Simkute, Ava Elizabeth Scott, Advait Sarkar, Abigail Sellen, and Sean Rintel. 2024. The Metacognitive Demands and Opportunities of Generative AI. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (CHI &#8217;24). Association for Computing Machinery, New York, NY, USA, Article 680, 1\u201324. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3613904.3642902\">https:\/\/doi.org\/10.1145\/3613904.3642902<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li>Rishi<strong> Vanukuru<\/strong>, Payod Panda, Xinyue Chen, Ava Elizabeth Scott, Lev Tankelevitch, and Sean Rintel. 2025. Designing Interfaces that Support Temporal Work Across Meetings with Generative AI. In Designing Interactive Systems Conference (DIS \u201925), July 5\u20139, 2025, Funchal, Portugal. ACM, NewYork, NY, USA, 21 pages. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/doi.org\/10.1145\/3715336.3735833\">https:\/\/doi.org\/10.1145\/3715336.3735833<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n<\/ol>\n\n\n\n\n\n<h1 class=\"wp-block-heading\" id=\"promptions-open-source-repository\">Promptions Open Source Repository<\/h1>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"if-you-re-building-ui-for-ai-give-users-the-power-of-choice\">If You\u2019re Building UI for AI, Give Users the Power of Choice<\/h2>\n\n\n\n<div class=\"wp-block-buttons is-layout-flex wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/labs.ai.azure.com\/projects\/promptions\/\" target=\"_blank\" rel=\"noreferrer noopener\">AI Foundry Labs<\/a><\/div>\n\n\n\n<div class=\"wp-block-button\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/github.com\/microsoft\/Promptions\" target=\"_blank\" rel=\"noreferrer noopener\">GitHub<\/a><\/div>\n<\/div>\n\n\n\n<p>Dynamic UI for prompting can transform users\u2019 AI experiences. You can use Promptions whether you\u2019re just starting out or building advanced systems.<\/p>\n\n\n\n<figure class=\"wp-block-embed is-type-video is-provider-youtube wp-block-embed-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"wp-block-embed__wrapper\">\n<iframe loading=\"lazy\" title=\"Promptions\" width=\"500\" height=\"281\" src=\"https:\/\/www.youtube-nocookie.com\/embed\/vr3fZpkKy8Q?feature=oembed&rel=0\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" referrerpolicy=\"strict-origin-when-cross-origin\" allowfullscreen><\/iframe>\n<\/div><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"the-challenge-of-prompting-ai\">The Challenge of Prompting AI<\/h3>\n\n\n\n<p>Prompting an AI might seem simple\u2014just type what you want and wait for the magic. But for many users, especially developers trying to build engaging AI-powered tools, the reality is more complex.Getting the AI to respond the way you want often requires nuanced prompt engineering, which can be:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Hard to learn: Knowing what context the AI needs isn\u2019t always obvious.<\/li>\n\n\n\n<li>Tedious to write: Even when you know what to say, crafting the right prompt takes time.<\/li>\n<\/ul>\n\n\n\n<figure class=\"wp-block-image aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"419\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/03-Promptions-compared-to-standard-Hi-Res-1024x419.png\" alt=\"graphical user interface, application\" class=\"wp-image-1156258\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/03-Promptions-compared-to-standard-Hi-Res-1024x419.png 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/03-Promptions-compared-to-standard-Hi-Res-300x123.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/03-Promptions-compared-to-standard-Hi-Res-768x314.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/03-Promptions-compared-to-standard-Hi-Res-1536x629.png 1536w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/03-Promptions-compared-to-standard-Hi-Res-2048x838.png 2048w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/03-Promptions-compared-to-standard-Hi-Res-240x98.png 240w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><em>A Promptions chatbot interface (left) contains UI elements to steer the AI response in ways that would take much more effort with a traditional AI chatbot interface (right).<\/em><\/figcaption><\/figure>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u201cI have to do so much [work] around my prompt to give it the context to make it specific, point it to the tone to do, and a lot of times I\u2019m repeating \u2026 those same things. I felt like [this] would really help me shortcut what I\u2019m doing manually today.\u201d<\/p>\n\n\n\n<p>(<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/aka.ms\/promptionspaper\" target=\"_blank\" rel=\"noopener noreferrer\">Dynamic Prompt Middleware study<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, Participant 15)<\/p>\n<\/blockquote>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"enter-promptions-ephemeral-ui-to-make-steering-easy\">Enter Promptions: Ephemeral UI to Make Steering Easy<\/h3>\n\n\n\n<p>Promptions (\u201cPrompt\u201d plus \u201coptions\u201d) helps users steer AI responses more effectively. Instead of manually refining prompts, Promptions generates contextual UI elements, like radio buttons, checkboxes, or toggles.<\/p>\n\n\n\n<p>As users interact with these elements, Promptions updates the AI\u2019s response in real time, making it more relevant and useful.<\/p>\n\n\n\n<p>This is an example of \u201cEphemeral UI\u201d, an emerging development paradigm in which user interfaces are created on-the-fly by AI systems and last just long enough to serve a specific purpose.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"what-can-promptions-do\">What Can Promptions Do<\/h3>\n\n\n\n<p>Promptions is best suited for any end-user interface where parameterizing prompts to add context can help steer AI output toward the user\u2019s preferences\u2014without requiring them to write or speak that context. It\u2019s simple, effective, and easily customizable, making it suitable for developers from individual vibe-coders to enterprise software engineers.<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><tbody><tr><td><strong>Domain<\/strong><\/td><td><strong>Description<\/strong><\/td><\/tr><tr><td>Customer support chatbots<\/td><td>Users refine support queries on the fly (e.g., specify tone or detail level) and see updated answers instantly, improving resolution speed and satisfaction.<\/td><\/tr><tr><td>Content creation platforms<\/td><td>Writers and marketers tweak style, length, or format parameters through GUI controls, iterating drafts faster while maintaining creative direction.<\/td><\/tr><tr><td>Data analytics and BI dashboards<\/td><td>Analysts adjust filters, aggregation levels, or visualization styles via checkboxes and sliders, regenerating AI-driven reports and insights instantly.<\/td><\/tr><tr><td>Educational tutoring systems<\/td><td>Students select difficulty, focus topics, or feedback style, prompting the AI tutor to adapt explanations and examples to individual learning needs.<\/td><\/tr><tr><td>Healthcare decision-support tools<\/td><td>Clinicians refine symptom context, risk factors, or treatment priorities through guided options, obtaining tailored diagnostic suggestions and care pathways.<\/td><\/tr><tr><td>Data annotation and curation<\/td><td>Promptions can parameterize labeling decisions into structured GUI inputs (e.g. sentiment sliders, style toggles), improving consistency, speed, and auditability in dataset creation.<\/td><\/tr><tr><td>Interactive explainability & auditing<\/td><td>Promptions allows users to explore how AI outputs shift with different refinement choices, offering a lightweight way to probe bias, model boundaries, or failure modes through UI interaction.<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"why-promptions-works\">Why Promptions Works<\/h3>\n\n\n\n<p>Promptions is grounded in our CHIWORK 2025 <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/aka.ms\/promptionspaper\" target=\"_blank\" rel=\"noopener noreferrer\">Dynamic Prompt Middleware research<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, which shows that users benefit from contextual UI elements that help them refine prompts without needing deep expertise in prompt engineering. This approach:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Lowers barriers to providing useful context<\/li>\n\n\n\n<li>Encourages task exploration and reflection<\/li>\n\n\n\n<li>Improves control over AI responses<\/li>\n\n\n\n<li>Enhances the overall user experience of generative AI workflows<\/li>\n<\/ul>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u201c[It] generated options that were things that I would maybe assume the system couldn\u2019t handle&#8230;.\u201d<\/p>\n\n\n\n<p>(<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/aka.ms\/promptionspaper\" target=\"_blank\" rel=\"noopener noreferrer\">Dynamic Prompt Middleware study<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, Participant 1)<\/p>\n<\/blockquote>\n\n\n\n<p>In the study, participants reported greater preferring dynamic elements, that it made them feel more successful with AI.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"600\" height=\"300\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/04-Promptions-Chart1.png\" alt=\"chart, box and whisker chart\" class=\"wp-image-1156260\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/04-Promptions-Chart1.png 600w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/04-Promptions-Chart1-300x150.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/04-Promptions-Chart1-240x120.png 240w\" sizes=\"auto, (max-width: 600px) 100vw, 600px\" \/><figcaption class=\"wp-element-caption\"><em>Comparison of participants\u2019 preferences of dynamic choices (Promptions) to static choices.<\/em><\/figcaption><\/figure>\n\n\n\n<p>Participants perceived the dynamic controls as more effective in managing AI output and expressed greater satisfaction with the degree of control provided, in contrast to the static condition, where many reported a continued need for additional control.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-full is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"600\" height=\"343\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/05-Promptions-Chart2-1.png\" alt=\"chart, box and whisker chart\" class=\"wp-image-1156262\" style=\"width:600px;height:auto\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/05-Promptions-Chart2-1.png 600w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/05-Promptions-Chart2-1-300x172.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/05-Promptions-Chart2-1-240x137.png 240w\" sizes=\"auto, (max-width: 600px) 100vw, 600px\" \/><figcaption class=\"wp-element-caption\"><em>Comparison of participants\u2019 reported effectiveness of dynamic choices (Promptions) to static choices<\/em>.<\/figcaption><\/figure>\n\n\n\n<p>But Promptions is more than just a usability improvement\u2014it\u2019s an example of the Microsoft Research <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/aka.ms\/toolsforthought\">Tools for Thought<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> approach to designing generative AI systems. The Promptions technique is a response to the <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/aka.ms\/metacognition\">metacognitive demand<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> of task decomposition:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>For novices, it reveals the scope of what\u2019s possible, helping them understand how a complex task can be broken down into manageable parts.<\/li>\n\n\n\n<li>For experts, it streamlines the process by surfacing relevant aspects of the task, allowing them to make precise refinements without tedious typing.<\/li>\n<\/ul>\n\n\n\n<p>This combination of cognitive support and interaction design makes Promptions a powerful tool for building AI systems that are not only more effective\u2014but also more empowering.<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u201c[It\u2019s giving me more of a prompt learning experience, I&#8217;m getting what I want out of the AI.&nbsp; And actually, it&#8217;s a better response too.\u201d<\/p>\n\n\n\n<p>(<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/aka.ms\/promptionspaper\" target=\"_blank\" rel=\"noopener noreferrer\">Dynamic Prompt Middleware study<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, Participant 11)<\/p>\n<\/blockquote>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"build-with-promptions\">Build with Promptions<\/h3>\n\n\n\n<p>Promptions is provided to developers as a TypeScript monorepo for AI-powered applications built with React, Fluent UI, and OpenAI integration. This project includes chat and image generation interfaces along with shared UI components and LLM utilities. The repo is available at <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/github.com\/microsoft\/Promptions\">https:\/\/github.com\/microsoft\/Promptions<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-large is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"982\" height=\"1024\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/06-Promptions-systems-model-HiRes-982x1024.png\" alt=\"diagram\" class=\"wp-image-1156263\" style=\"width:555px;height:auto\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/06-Promptions-systems-model-HiRes-982x1024.png 982w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/06-Promptions-systems-model-HiRes-288x300.png 288w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/06-Promptions-systems-model-HiRes-768x801.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/06-Promptions-systems-model-HiRes-173x180.png 173w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/06-Promptions-systems-model-HiRes.png 1103w\" sizes=\"auto, (max-width: 982px) 100vw, 982px\" \/><figcaption class=\"wp-element-caption\"><em>Promptions system model. (1) The Option Module ingests the user&#8217;s prompt input along with the conversation history. (2) It then outputs a set of prompt options, each initialized based on the content of the prompt. (3) These options are rendered inline via a dedicated rendering engine. (4) The Chat Module incorporates the refined options as grounding, alongside the original prompt and conversation history, to generate a chat response. (5) The user can modify the GUI controls, which updates the refinements and triggers the Chat Module to regenerate the current response accordingly.<\/em><\/figcaption><\/figure>\n\n\n\n<p>You can fork the repo and customize Promptions for your own applications. You can also contribute to Promptions. Whether you\u2019re building a chatbot, a coding assistant, or a creative writing tool, Promptions gives users the power of choice\u2014and better results.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"looking-ahead\">Looking Ahead<\/h3>\n\n\n\n<p>Promptions is part of a growing movement to make AI more user-centered, customizable, and accessible. By giving users intuitive control over how AI responds, we\u2019re building a future where everyone\u2014from novice coders to seasoned engineers\u2014can create smarter, more responsive AI experiences.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"credits\">Credits<\/h3>\n\n\n\n<p><em>Promptions was invented by <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.iandrosos.me\/\" target=\"_blank\" rel=\"noopener noreferrer\">Ian Drosos<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> (a past Microsoft Research Resident) with support from <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/johnwilliams\/\" target=\"_blank\" rel=\"noreferrer noopener\">Jack Williams<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/advait\/\" target=\"_blank\" rel=\"noreferrer noopener\">Advait Sarkar<\/a>, Nicholas Wilson, <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/payodpanda.com\/\" target=\"_blank\" rel=\"noopener noreferrer\">Payod Panda<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, and Sean Rintel. It is a collaboration between the&nbsp;<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/project\/tools-for-thought\/\">Tools for Thought<\/a>&nbsp;and ENCODE projects in the <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/theme\/people-centric-ai\/\" target=\"_blank\" rel=\"noreferrer noopener\">People-Centric AI<\/a> focus area at&nbsp;<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/lab\/microsoft-research-cambridge\/\" target=\"_blank\" rel=\"noreferrer noopener\">Microsoft Research Cambridge, UK<\/a>.<\/em><\/p>\n\n\n\n\n\n<h1 class=\"wp-block-heading\" id=\"workshops\">Workshops<\/h1>\n\n\n\n<div class=\"wp-block-group is-layout-constrained wp-block-group-is-layout-constrained\">\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<h3 class=\"wp-block-heading\" id=\"chi2026-barcelona-april-2026\">CHI2026 | Barcelona, April 2026<\/h3>\n\n\n\n<p><strong>Website:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/ai-tools-for-thought.github.io\/workshop\/\">CHI 2026 Workshop on Tools for Thought: Understanding, Protecting, and Augmenting Human Cognition with Generative AI \u2014 From Vision to Implementation<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p><strong>Description:<\/strong> GenAI radically widens the scope and capability of automation for work, learning, and creativity. While impactful, it also changes workflows, raising questions about its effects on cognition, including critical thinking and learning. Yet GenAI also offers opportunities for designing \u201ctools for thought\u201d (TfT) that protect and augment cognition. Such systems provoke critical thinking, provide personalized tutoring, or enable novel ways of sensemaking, among other approaches. How does GenAI change workflows and human cognition? What are opportunities and challenges for designing GenAI systems that protect and augment thinking? Which theories, perspectives, and methods are relevant? This workshop aims to develop a multidisciplinary community interested in exploring these questions to protect against the erosion, and fuel the augmentation, of human cognition using GenAI. This workshop is a follow-up to the CHI 2025 Tools for Thought workshop, which brought together 56 participants with 34 accepted submissions, culminating in a workshop synthesis and an HCI journal special issue on tools for thought. While last year\u2019s workshop focussed on mapping the field, this edition moves towards developing operational frameworks, principles, and tools.<\/p>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<h3 class=\"wp-block-heading\" id=\"chi2025-japan-april-2025\">CHI2025 | Japan, April 2025<\/h3>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<ul class=\"wp-block-list\">\n<li><strong>Website (including all papers):<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/aka.ms\/toolsforthoughtworkshop\" target=\"_blank\" rel=\"noopener noreferrer\">Tools for Thought: Research and Design for Understanding, Protecting, and Augmenting Human Cognition with Generative AI<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li>\n\n\n\n<li><strong>Synthesis report: <\/strong><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/understanding-protecting-and-augmenting-human-cognition-with-generative-ai-a-synthesis-of-the-chi-2025-tools-for-thought-workshop\/\" target=\"_blank\" rel=\"noreferrer noopener\">Understanding, Protecting, and Augmenting Human Cognition with Generative AI: A Synthesis of the CHI 2025 Tools for Thought Workshop<\/a><\/li>\n\n\n\n<li><strong>Special Issue:<\/strong> <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/ai-tools-for-thought.github.io\/workshop\/hci-special-issue\" target=\"_blank\" rel=\"noopener noreferrer\">HCI Journal Special Issue<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> (In progress)<\/li>\n<\/ul>\n\n\n\n<p><strong>Description:<\/strong> We invited researchers, designers, practitioners, and provocateurs to explore what it means to understand and shape the impact of Generative AI (GenAI) on human cognition. GenAI radically widens the scope and capability of automation for work, learning, and creativity. While impactful, it also changes workflows and the quality of thinking involved, raising questions about its effects on cognition, including critical thinking and learning. Yet, GenAI also offers opportunities for designing tools for thought that protect and augment cognition. Such systems provoke critical thinking, provide personalized tutoring, or enable novel ways of sensemaking, among other approaches. How does GenAI change workflows and human cognition? What are opportunities and challenges for designing GenAI systems that protect and augment human cognition? Which theories, perspectives, and methods are relevant? This workshop aimed to develop a multidisciplinary community interested in exploring these questions to protect against the erosion, and fuel the augmentation, of human cognition using GenAI.<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n\n\n\n<h1 class=\"wp-block-heading\" id=\"we-re-hiring\">We&#8217;re hiring!<\/h1>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"power-the-next-step-change-in-human-cognition-with-microsoft-research\">Power the next step-change in human cognition with Microsoft Research<\/h2>\n\n\n\n<p>Are you passionate about exploring how generative AI can transform the way we think, learn, and create? Do you have a PhD in Human-Computer Interaction (HCI) or a related field, and a talent for building innovative prototypes? If so, we have an exciting opportunity for you!<\/p>\n\n\n\n<p>The&nbsp;Microsoft Research <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/project\/t4t\/\" target=\"_blank\" rel=\"noreferrer noopener\">Tools for Thought&nbsp;(T4T)<\/a> project, based&nbsp;in Cambridge UK, is seeking a dynamic and creative researcher with engineering and design skills to join our team as a&nbsp;Post-Doctoral Resident. This two-year position offers a unique chance to work with a multi-disciplinary team of social scientists, designers, machine learning researchers, and engineers. Together, we&#8217;ll explore the cutting-edge of AI and its impact on human cognition.<\/p>\n\n\n\n<div class=\"wp-block-buttons is-layout-flex wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button\"><a data-bi-type=\"button\" class=\"wp-block-button__link wp-element-button\" href=\"https:\/\/aka.ms\/t4tpostdoc\">Apply now!<\/a><\/div>\n<\/div>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"what-you-ll-do\">What You&#8217;ll Do<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Collaborate and Innovate: Work closely with the T4T team to shape our research agenda and project work. Your ideas and insights will help turn foundational research into interactive AI system prototypes.<\/li>\n\n\n\n<li>Design and Develop: Create and test interactive prototypes that blend user experiences with machine learning technologies. Your work will help us understand and enhance critical thinking, metacognition, reflection, planning, sensemaking, expertise, learning, and creativity.&nbsp;The AI experiences should either address current work scenarios or explore future workflows made possible by AI, showing a deep understanding of their context and accessibility to diverse users.<\/li>\n\n\n\n<li>Learn and Lead: Take a role in existing research projects and develop your own seedling projects. Connect with partners across Microsoft Research, product teams, and academia to push the boundaries of what&#8217;s possible.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"what-we-re-looking-for\">What We&#8217;re Looking For<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Research Excellence: <\/strong>A PhD (or near completion) in HCI or a related field (e.g., Psychology, Sociology, Anthropology, Cognitive Science). Experience conducting research with people, preferably with high ecological validity. You should have experience in researching how AI affects thinking and have developed prototypes or systems that demonstrate this.<\/li>\n\n\n\n<li><strong>Demonstrated ability to build interactive prototypes:<\/strong> Experience creating interfaces with web frameworks, such as React (HTML, Javscript\/Typescript). Build backend systems that integrate generative AI with system prompt engineering, using technologies such as NodeJS, Python (e.g. Plotly, Django, Flask, FastAPI). Incorporating UX\/UI design tools such as Figma\/Sketch into your process. Not required but a plus &#8211; experience evaluating LLM\/SLM outputs and knowledge of model tuning methods to enhance AI performance, and machine learning frameworks such as TensorFlow and Pytorch.<\/li>\n\n\n\n<li><strong>Creative Vision: <\/strong>A passion for exploring new ways AI can augment human thinking. Experience designing systems that address real-world scenarios or envision future workflows.<\/li>\n\n\n\n<li><strong>Communication Skills: <\/strong>Excellent verbal and written communication skills. Ability to translate research findings into practical actions and communicate complex ideas to non-expert audiences.<\/li>\n\n\n\n<li><strong>Project Management: <\/strong>Strong project and time management skills. Experience working in multi-disciplinary teams and managing multiple projects simultaneously.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"why-join-us\">Why Join Us?<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Inclusive Environment: <\/strong>We welcome applications from individuals of all genders, ethnicities, socioeconomic backgrounds, disabilities, and other dimensions of diversity. We believe in the power of diverse perspectives and are committed to creating an inclusive environment for all employees.<\/li>\n\n\n\n<li><strong>Support for Women: <\/strong>We are particularly committed to supporting women in technology and research. Our lab includes many accomplished women researchers and engineers who are eager to mentor and support new team members. We offer flexible working arrangements to help balance work and personal life, and we actively promote a culture of respect and inclusion.<\/li>\n\n\n\n<li><strong>World-Class Research: <\/strong>Be part of a world-class research centre that prioritizes human-centred design and deploys technologies built on deep research in real-world settings.<\/li>\n\n\n\n<li><strong>Impactful Work: <\/strong>Your work will contribute to developing an AI ecosystem that fosters creativity and expertise while avoiding overreliance or skill degradation.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"additional-information\">Additional Information<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Contract Duration: <\/strong>2 years<\/li>\n\n\n\n<li>Start Date: Must be available to work before the end of 2025, preferably no later than October 2025.<\/li>\n\n\n\n<li><strong>Hybrid work: <\/strong>This position is based in Cambridge UK, and requires 50% time in office.<\/li>\n<\/ul>\n\n\n\n<p><strong>Ready to make a difference? <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/aka.ms\/t4tpostdoc\">Apply now<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> and join us in shaping the future of Tools for Thought.<\/strong><\/p>\n\n\n\n\n\n<figure class=\"wp-block-image alignleft size-large is-style-default\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"576\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/T4T-CHI-2025-BlogHeroFeature-1400x788-1-1920x1080-1-1024x576.jpg\" alt=\"Tools for Thought at CHI2025 logo\" class=\"wp-image-1137253\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/T4T-CHI-2025-BlogHeroFeature-1400x788-1-1920x1080-1-1024x576.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/T4T-CHI-2025-BlogHeroFeature-1400x788-1-1920x1080-1-300x169.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/T4T-CHI-2025-BlogHeroFeature-1400x788-1-1920x1080-1-768x432.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/T4T-CHI-2025-BlogHeroFeature-1400x788-1-1920x1080-1-1536x864.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/T4T-CHI-2025-BlogHeroFeature-1400x788-1-1920x1080-1-1066x600.jpg 1066w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/T4T-CHI-2025-BlogHeroFeature-1400x788-1-1920x1080-1-655x368.jpg 655w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/T4T-CHI-2025-BlogHeroFeature-1400x788-1-1920x1080-1-240x135.jpg 240w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/T4T-CHI-2025-BlogHeroFeature-1400x788-1-1920x1080-1-640x360.jpg 640w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/T4T-CHI-2025-BlogHeroFeature-1400x788-1-1920x1080-1-960x540.jpg 960w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/T4T-CHI-2025-BlogHeroFeature-1400x788-1-1920x1080-1-1280x720.jpg 1280w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/T4T-CHI-2025-BlogHeroFeature-1400x788-1-1920x1080-1.jpg 1920w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:100%\">\n<p>At <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/event\/chi-2025\/\">CHI2025<\/a>, we\u2019re presenting four new research papers and cohosting a <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/tools-for-thought-research-and-design-for-understanding-protecting-and-augmenting-human-cognition-with-generative-ai\/\">Tools for Thought workshop<\/a>, that dives deep into the intersection of AI and human cognition.<\/p>\n\n\n\n<p><strong>Read our blog post summary<\/strong><\/p>\n\n\n\n<p><strong><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/blog\/the-future-of-ai-in-knowledge-work-tools-for-thought-at-chi-2025\/\">The Future of AI in Knowledge Work: Tools for Thought at CHI 2025 &#8211; Microsoft Research<\/a><\/strong><\/p>\n\n\n\n<p><strong>CHI2025 Papers<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/the-impact-of-generative-ai-on-critical-thinking-self-reported-reductions-in-cognitive-effort-and-confidence-effects-from-a-survey-of-knowledge-workers\/\">The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers &#8211; Microsoft Research<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/ai-help-me-think-but-for-myself-assisting-people-in-complex-decision-making-by-providing-different-kinds-of-cognitive-support\/\">AI, Help Me Think\u2014but for Myself: Assisting People in Complex Decision-Making by Providing Different Kinds of Cognitive Support &#8211; Microsoft Research<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/are-we-on-track-ai-assisted-active-and-passive-goal-reflection-during-meetings\/\">Are We On Track? AI-Assisted Active and Passive Goal Reflection During Meetings &#8211; Microsoft Research<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/yes-and-a-generative-ai-multi-agent-framework-for-enhancing-diversity-of-thought-in-individual-ideation-for-problem-solving-through-confidence-based-agent-turn-taking\/\">YES AND: A Generative AI Multi-Agent Framework for Enhancing Diversity of Thought in Individual Ideation for Problem-Solving Through Confidence-Based Agent Turn-Taking &#8211; Microsoft Research<\/a><\/li>\n<\/ul>\n<\/div>\n<\/div>\n\n\n\n\n\n<h2 class=\"wp-block-heading\" id=\"2025-january\">2025 January<\/h2>\n\n\n\n<p>The January 16 edition of the Microsoft Education Blog report features two of our studies: <a href=\"https:\/\/www.microsoft.com\/en-us\/education\/blog\/2025\/01\/delivering-greater-impact-with-copilot-and-the-power-of-agents\/\">Delivering greater impact with Copilot and the power of agents<\/a><\/p>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<p><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/papers.ssrn.com\/sol3\/papers.cfm?abstract_id=5095149\"><strong>Effects of LLM use and note-taking on reading comprehension and memory: A randomised experiment in secondary schools<\/strong><span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>PDF Summary: <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/cdn-dynmedia-1.microsoft.com\/is\/content\/microsoftcorp\/microsoft\/final\/en-us\/microsoft-product-and-services\/microsoft-education\/downloadables\/AI-and-Reading-Study.pdf\">AI and Traditional Learning: Complementary Strategies for Deeper Learning<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>The rapid uptake of Generative AI, particularly large language models (LLMs), by students raises urgent questions about their effects on learning. We compared the impact of LLM use to that of traditional note-taking, or a combination of both, on secondary school students&#8217; reading comprehension and retention. We conducted a pre-registered, randomised controlled experiment with within-and between-participant design elements in schools. 405 students aged 14-15 studied two text passages and completed comprehension and retention tests three days later. Quantitative results demonstrated that both note-taking alone and combined with the LLM had significant positive effects on retention and comprehension compared to the LLM alone. Yet, most students preferred using the LLM over note-taking, and perceived it as more helpful. Qualitative results revealed that many students valued LLMs for making complex material more accessible and reducing cognitive load, while they appreciated note-taking for promoting deeper engagement and aiding memory. Additionally, we identified &#8220;archetypes&#8221; of prompting behaviour, offering insights into the different ways students interacted with the LLM. Overall, our findings suggest that, while note-taking promotes cognitive engagement and long-term comprehension and retention, LLMs may facilitate initial understanding and student interest. The study reveals the continued importance of traditional learning approaches, the benefits of combining AI use with traditional learning over using AI alone, and the AI skills that students need to maximise those benefits.<\/p>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<p><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/arxiv.org\/abs\/2501.08864\"><strong>The New Calculator? Practices, Norms, and Implications&nbsp;<\/strong><span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>PDF Summary: <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/cdn-dynmedia-1.microsoft.com\/is\/content\/microsoftcorp\/microsoft\/final\/en-us\/microsoft-product-and-services\/microsoft-education\/downloadables\/Microsoft-AI-in-HED.pdf\">How Students and Educators See AI in Higher Education<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<p>Generative AI (GenAI) has introduced myriad opportunities and challenges for higher education. Anticipating this potential transformation requires understanding students&#8217; contextualised practices and norms around GenAI. We conducted semi-structured interviews with 26 students and 11 educators from diverse departments across two universities. Grounded in Strong Structuration Theory, we find diversity in students&#8217; uses and motivations for GenAI. Occurring in the context of unclear university guidelines, institutional fixation on plagiarism, and inconsistent educator communication, students&#8217; practices are informed by unspoken rules around appropriate use, GenAI limitations and reliance strategies, and consideration of agency and skills. Perceived impacts include changes in confidence, and concerns about skill development, relationships with educators, and plagiarism. Both groups envision changes in universities&#8217; attitude to GenAI, responsible use training, assessments, and integration of GenAI into education. We discuss socio-technical implications in terms of current and anticipated changes in the external and internal structures that contextualise students&#8217; GenAI use.<\/p>\n<\/div>\n<\/div>\n\n\n\n<div style=\"height:30px\" aria-hidden=\"true\" class=\"wp-block-spacer\"><\/div>\n\n\n\n\n\n<h2 class=\"wp-block-heading\" id=\"10-strategies-for-thinking-better-with-ai\"><strong>10 Strategies For Thinking Better With AI<\/strong><\/h2>\n\n\n\n<p><em>A hands-on practical guide for knowledge workers<\/em><\/p>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<p><strong>01: AI Can Ask Questions<br><\/strong>Let AI ask you questions instead of giving you answers<\/p>\n\n\n\n<p><strong>02: AI Can Surface Best-Practices<br><\/strong>Let AI connect you to what works \u2014 before you reinvent the wheel<\/p>\n\n\n\n<p><strong>03: AI Can Spot Overlooked Things<br><\/strong>Let AI highlight gaps you didn&#8217;t know were there<\/p>\n\n\n\n<p><strong>04: AI Can Leverage Expertise<br><\/strong>Let AI amplify your experience, not replace it<\/p>\n\n\n\n<p><strong>05: AI Can Offer Alternatives<br><\/strong>Don&#8217;t just ask AI for an answer \u2014 ask for options<\/p>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<p><strong>06: AI Can Help Understand Outcomes<br><\/strong>Use AI to think forward, not just backward<\/p>\n\n\n\n<p><strong>07: AI Can Identify Patterns<br><\/strong>Use AI to plan what you should be looking for<\/p>\n\n\n\n<p><strong>08: AI Can Break Down Tasks<br><\/strong>Let AI turn big goals into manageable pieces<\/p>\n\n\n\n<p><strong>09: AI Can Surface Related Material<br><\/strong>Find connections you wouldn&#8217;t have thought to look for<\/p>\n\n\n\n<p><strong>10: AI Can Support Learning<br><\/strong>Use AI as a mentor, not just a machine<\/p>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<div data-wp-interactive=\"core\/file\" class=\"wp-block-file\"><object data-wp-bind--hidden=\"!state.hasPdfPreview\" hidden class=\"wp-block-file__embed\" data=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/2026-01-T4T-Principles-Printable.pdf\" type=\"application\/pdf\" style=\"width:100%;height:200px\" aria-label=\"Embed of Printable Principles Page.\"><\/object><a id=\"wp-block-file--media-001f0f91-0e2f-40df-9aff-00f85a6a570c\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/2026-01-T4T-Principles-Printable.pdf\">Printable Principles Page<\/a><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/07\/2026-01-T4T-Principles-Printable.pdf\" class=\"wp-block-file__button wp-element-button\" download aria-describedby=\"wp-block-file--media-001f0f91-0e2f-40df-9aff-00f85a6a570c\">Download<\/a><\/div>\n<\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n\n<p>The world is learning to AI as a writing tool, a code assistant, a meeting summarizer. But AI&#8217;s real power may lie somewhere more subtle \u2014 and more profound: it can help us think better.<\/p>\n\n\n\n<p>Thinking well is hard. It takes time, focus, and the kind of mental discipline that our busy workdays rarely allow. Much of the recent focus of AI has been on accelerating routine work, which should give us more time. It can do more than that, though. AI can be more than just a way to do more. You can use it to help think more clearly, more deeply, and more creatively.<\/p>\n\n\n\n<p>This guide is the product of the&nbsp;<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/aka.ms\/toolsforthought\" target=\"_blank\" rel=\"noopener noreferrer\">Tools for Thought<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>&nbsp;team at Microsoft Research, grounded in our research into how AI can enhance human cognition. Drawing on empirical studies, prototypes, and design principles, these strategies reflect our commitment to shaping AI that augments not just productivity, but the quality of thought itself.<\/p>\n\n\n\n<p>These strategies are hands-on. Open a Copilot window next to this window and copy prompts in to see their effect in real scenarios.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-9d6595d7 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<h3 class=\"wp-block-heading\" id=\"01-ai-can-ask-questions\">01: <strong>AI Can Ask Questions<\/strong><\/h3>\n\n\n\n<p>Let AI ask you questions instead of giving you answers. This helps you better understand and critique the problem.<\/p>\n\n\n\n<p><em>You&#8217;re preparing for a strategy offsite. Normally, you&#8217;d jump straight into planning slides. This time, you try something different: you ask the AI to challenge your assumptions. It replies with, &#8220;What evidence do you have that this is your most important customer segment?&#8221; Now you&#8217;re thinking about your data in a whole new way.<\/em><\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"related-research\"><strong>Related Research<\/strong><\/h4>\n\n\n\n<p>When AI asks people questions about what they want, rather than providing immediate answers, people engage more deeply with their underlying purposes and processes. AI-assisted reflection before meetings helps people articulate what success looks like and anticipate potential challenges, leading to better preparation (<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/what-does-success-look-like-catalyzing-meeting-intentionality-with-ai-assisted-prospective-reflection\/\" target=\"_blank\" rel=\"noreferrer noopener\">Scott, et al., 2025<\/a>).<\/p>\n\n\n\n<p>Asking people to &#8216;unpack&#8217; what they want is a good start, but AI-generated &#8216;provocations&#8217; \u2014 alternatives, counter-positions, and multiple perspectives \u2014 can be even more effective. This approach not only improves immediate outputs but also develops ongoing critical thinking skills that apply to all future work (<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/arxiv.org\/abs\/2501.17247\" target=\"_blank\" rel=\"noopener noreferrer\">Drosos, et al, 2025<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>,&nbsp;<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/its-like-a-rubber-duck-that-talks-back-understanding-generative-ai-assisted-data-analysis-workflows-through-a-participatory-prompting-study\/\" target=\"_blank\" rel=\"noreferrer noopener\">Drosos et al. 2024<\/a>,&nbsp;<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/ai-should-challenge-not-obey\/\" target=\"_blank\" rel=\"noreferrer noopener\">Sarkar, 2024<\/a>).<\/p>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<h4 class=\"wp-block-heading\" id=\"less-effective-approach\">\u274c Less Effective Approach<\/h4>\n\n\n\n<pre class=\"wp-block-code\"><code>Create a marketing plan for our new product.<\/code><\/pre>\n\n\n\n<p>This gives you answers, but doesn&#8217;t help you think through the underlying strategy.<\/p>\n\n\n\n<h4 class=\"wp-block-heading\" id=\"better-thinking-approach\">\u2705 Better Thinking Approach<\/h4>\n\n\n\n<pre class=\"wp-block-code\"><code>I'm preparing to present our new customer retention initiative to the board next week. Before I finalize my pitch, ask me five challenging questions that board members might raise about our approach, target metrics, and potential risks.<\/code><\/pre>\n\n\n\n<p>This helps you clearly express your intentions, and helps you discover insights you might have missed.<\/p>\n\n\n\n<p><strong>More advanced examples<\/strong><\/p>\n\n\n\n<p><em>Get multiple perspectives:<\/em><\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>Help me stress-test my project through a sequence of probing questions that adapt to my answers. In each round, shift perspective (e.g., customer, competitor, investor, team member) and then add a meta-question: 'What important question have we not asked yet?' Continue until no major gaps remain.<\/code><\/pre>\n\n\n\n<p><em>Unpack your assumptions:<\/em><\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>Guide me through an iterative questioning process. Begin with a broad probing question. For each of my answers, generate both a follow-up question and a higher-level meta-question about what's missing. Organize these into a branching 'question map' that shows how my assumptions connect, where the blind spots cluster, and which lines of inquiry are most critical to pursue.<\/code><\/pre>\n<\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" \/>\n\n\n\n\n\n<p><\/p>\n\n\n","protected":false},"excerpt":{"rendered":"<p>Better thinking through AIA People-Centric AI Project &#8220;What would you rather have: a tool that thinks for you, or a tool that makes you think?&#8221; Senior Researcher Advait Sarkar presented at TEDAI (opens in new tab) in Vienna on September 26th 2025. \u00a9 TEDAI Vienna \/ Robert Leslie Many AI tools focus on solving specific [&hellip;]<\/p>\n","protected":false},"featured_media":1155638,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","footnotes":""},"research-area":[13556,13554,13559],"msr-locale":[268875],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-1053711","msr-project","type-msr-project","status-publish","has-post-thumbnail","hentry","msr-research-area-artificial-intelligence","msr-research-area-human-computer-interaction","msr-research-area-social-sciences","msr-locale-en_us","msr-archive-status-active"],"msr_project_start":"","related-publications":[1149887,1134757,1135044,1135061,1136305,1137954,1139162,1139299,1139962,1143413,1134742,1155619,1162665,1163447,1163452,1163518,1163529,1163532,1163790,1163823,1040949,954237,971928,981003,1000449,1002048,1010463,1014123,1016154,1031829,950808,1050798,1050837,1077624,1089255,1095081,1095732,1109211,1110786],"related-downloads":[],"related-videos":[1013847,1125393,1156719],"related-groups":[1142579],"related-events":[1134700,1019022],"related-opportunities":[],"related-posts":[1136647,1157824,1167876],"related-articles":[],"tab-content":[],"slides":[],"related-researchers":[{"type":"user_nicename","display_name":"Sean Rintel","user_id":33579,"people_section":"Members","alias":"serintel"},{"type":"user_nicename","display_name":"Richard Banks","user_id":33361,"people_section":"Members","alias":"rbanks"},{"type":"user_nicename","display_name":"Advait Sarkar","user_id":37146,"people_section":"Members","alias":"advait"},{"type":"user_nicename","display_name":"Pratik Ghosh","user_id":38245,"people_section":"Members","alias":"prghos"},{"type":"user_nicename","display_name":"Martin Grayson","user_id":32893,"people_section":"Members","alias":"mgrayson"},{"type":"user_nicename","display_name":"Britta Burlin","user_id":39456,"people_section":"Members","alias":"brburlin"},{"type":"user_nicename","display_name":"Lev Tankelevitch","user_id":43209,"people_section":"Members","alias":"levt"},{"type":"guest","display_name":"Payod Panda","user_id":841492,"people_section":"Members","alias":""},{"type":"guest","display_name":"Viktor Kewenig","user_id":1146458,"people_section":"Members","alias":""},{"type":"user_nicename","display_name":"Abigail Sellen","user_id":31112,"people_section":"Collaborators","alias":"asellen"},{"type":"user_nicename","display_name":"Si\u00e2n Lindley","user_id":33651,"people_section":"Collaborators","alias":"sianl"},{"type":"user_nicename","display_name":"Jack Williams","user_id":40156,"people_section":"Collaborators","alias":"johnwilliams"},{"type":"user_nicename","display_name":"Christian Poelitz","user_id":42612,"people_section":"Collaborators","alias":"cpoelitz"},{"type":"user_nicename","display_name":"Jake Hofman","user_id":32340,"people_section":"Collaborators","alias":"jmh"},{"type":"user_nicename","display_name":"Jenn Wortman Vaughan","user_id":32235,"people_section":"Collaborators","alias":"jenn"},{"type":"guest","display_name":"Xinyue Chen","user_id":1150917,"people_section":"Past Contributors","alias":""},{"type":"guest","display_name":"Emily Doherty","user_id":1150914,"people_section":"Past Contributors","alias":""},{"type":"guest","display_name":"Ian Drosos","user_id":1150918,"people_section":"Past Contributors","alias":""},{"type":"guest","display_name":"Gonzalo Ramos","user_id":1150913,"people_section":"Past Contributors","alias":""},{"type":"guest","display_name":"Ava Scott","user_id":1150915,"people_section":"Past Contributors","alias":""},{"type":"guest","display_name":"Rishi Vanukuru","user_id":1150916,"people_section":"Past Contributors","alias":""}],"msr_research_lab":[199561,199565],"msr_impact_theme":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/1053711","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-project"}],"version-history":[{"count":165,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/1053711\/revisions"}],"predecessor-version":[{"id":1166903,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/1053711\/revisions\/1166903"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/1155638"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1053711"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1053711"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1053711"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1053711"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=1053711"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}