Kiran Tomlinson

Kiran Tomlinson

Senior Researcher

Microsoft Research

About

I’m a Senior Researcher at Microsoft Research in the AI Interaction and Learning group, where I’m working on understanding and improving the interactions between people and their LLM-based assistants. More broadly, my research studies preferences and decision-making through algorithmic and machine learning methods.

I recently completed my PhD in Computer Science at Cornell University, where I was fortunate to be advised by Jon Kleinberg. My dissertation was on human decision-making at the individual and collective level. During my PhD, I interned at Microsoft Research with Jennifer Neville on recommendations in networks and at Microsoft’s Office of Applied Research with Longqi Yang and Mengting Wan on multi-organization recommendation. In Winter and Spring 2023, I was a visiting instructor at Carleton College teaching Data Structures and Mathematics of Computer Science.

When away from my desk, I spend my time learning to fly, playing guitar, building 8-bit computers, playing video games, biking, listening to music, flying quadcopters, bouldering, and playing pool. I have additional interests in spaceflight, Premier League football, and Formula 1.

[High-res headshot] [CV of Failures]

Recent news

Sep ‘25 Our paper on LLM problem-solving failures was accepted to NeurIPS ‘25 as a spotlight!

🗣 Sep ‘25 Gave a talk on AI applicability to work at the Wharton Business & Generative AI Conference.

🔍 Aug ‘25 Given the misunderstandings of our research on AI and jobs, we wrote a blog post clarifying what our research says.

🗞️ Jul ‘25 Our preprint on AI and jobs has gotten quite a bit of media coverage, including in Newsweek, Business Insider, CNBC, and Fortune. (Unfortunately, a lot of it misrepresents the paper.)

🛠 Jul ‘25 Our workshop What Can(’t) Transformers Do? was accepted to NeurIPS ‘25!

All news

Collaboration Network

People are red, papers are blue.

Recent Papers

(2025). Working with AI: Measuring the Occupational Implications of Generative AI. arXiv.

PDF Cite Dataset Slides Video

(2025). Lost in Transmission: When and Why LLMs Fail to Reason Globally. Accepted to NeurIPS ‘25 (spotlight).

PDF Cite

(2025). When the Universe is Too Big: Bounding Consideration Probabilities for Plackett-Luce Rankings. AISTATS.

PDF Cite Code Poster

(2025). Replicating Electoral Success. AAAI.

PDF Cite Code Poster Slides Video DOI