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.

Tools for Thought

Better thinking through AI

Much AI research focuses on solving specific tasks for people – generating content or automating processes. While such systems may be powerful, there are risks that this approach impact the way people think and therefore learn, build skills, and deploy expertise.

The Tools for Thought (T4T) team aims to help researchers and developers imagine how AI might help people to think better, so that:

as well as getting the job done, it helps us better understand and figure out the job.
as well as creating content, it helps us think more critically and with more insight throughout an entire workflow.
as well as seeking speed and efficiency, it helps us create outcomes that are more effective and of higher quality because they are the product of better answers from better questions.
as well as augmenting individual cognition and tasks, it augments collective cognition and workflows.
as well as automating known processes, it helps organisations predict and explore the unknown.

Outputs of T4T include principles and guidelines for supporting cognition in any user experience, as well as systems and new technologies that stand as practical instantiations of what it means to support better thinking using AI.

This is a venn diagram that shows 10 types of thinking that the Tools for Thought team are interested in.

Workstreams

Workstream	Description \| Owner
Critical Thinking with AI	How might we design interfaces to promote critical thinking about task and AI output when working with generative AI \| Advait Sarkar AI Should Challenge, Not Obey Exploring Perspectives on the Impact of Artificial Intelligence on the Creativity of Knowledge Work: Beyond Mechanised Plagiarism and Stochastic Parrots Co-audit: tools to help humans double-check AI-generated content
Diversity of Thought with AI	Can AI agents support conversations with users that are rich, meaningful and well rounded, and that give humans agency over the discussion? \| Pratik Ghosh
Explanations from AI	How might we enable better AI explanations for LLMs by providing user control over their characteristics and formation? \| Ian Drosos “It’s like a rubber duck that talks back”: Understanding Generative AI-Assisted Data Analysis Workflows through a Participatory Prompting Study
Intentional Meetings with AI	How might we leverage generative AI to encourage self reflection and self awareness around collaboration, in order to make meetings more intentional? \| Sean Rintel / Lev Tankelevitch Mental Models of Meeting Goals: Supporting Intentionality in Meeting Technologies The CoExplorer Technology Probe: A Generative AI-Powered Adaptive Interface to Support Intentionality in Planning and Running Video Meetings
Learning & Understanding with AI	What impact do generative AI tools have on people’s understanding and memory of what they are doing, and how can they help build core cognitive capabilities? \| Abigail Sellen / Leon Reicherts / Lev Tankelevitch / Sean Rintel Preprint: Effects of LLM use and note-taking on reading comprehension and memory: A randomised experiment in secondary schools Summary report: AI and Traditional Learning: Complementary Strategies for Deeper Learning Preprint: The New Calculator? Practices, Norms, and Implications of Generative AI in Higher Education Summaryh report: How Students and Educators See AI in Higher Education
Task decomposition with AI for data analysis	How might we create interfaces that support task decomposition in the context of data analysis, as well as provide affordances that allow for verification? \| Jack Williams Improving Steering and Verification in AI-Assisted Data Analysis with Interactive Task Decomposition
Metacognition & AI	How do generative AI systems impose metacognitive demands on users, and how might they instead incorporate metacognitive support strategies? \| Lev Tankelevitch The Metacognitive Demands and Opportunities of Generative AI Ironies of Generative AI: Understanding and mitigating productivity loss in human-AI interactions
Workflows & Collaborative Intent with AI	How are complex collaborative intents formed? How are they formed and how do they evolve over time? \| Britta Burlin
Creating with AI	How can we give people rich agency in the ways they participate in human-AI collaborative creativity? \| Gonzalo Ramos Evolving Roles and Workflows of Creative Practitioners in the Age of Generative AI – Microsoft Research

People

The Tools for Thought team is interdisciplinary, mixing experts in social science, computer science, engineering, and design. Our workstreams reflect this mix, combining depth in user research, cutting-edge technology, and new user experiences.