Computational cognitive science and AI

Hanbo Xie

I study how people think, decide, and learn by combining cognitive modeling, think-aloud data, and large language models.

Ph.D. student, Georgia Tech Psychology Princeton CoCoSci Lab visit, Mar-Jul 2025

Cognitive Modeling Think-Aloud Protocols Human-Centered AI

View CV Publications
Hanbo Xie

About

Computational approaches to cognition, decision-making, and human-centered AI.

I am a Ph.D. student in Psychology at the Georgia Institute of Technology, specializing in computational cognitive science. My work asks how language, behavior, and computational models can reveal the hidden structure of human thought.

My thesis uses large language models to analyze think-aloud data from decision-making and learning tasks. Rather than treating behavior as only button presses or choices, I use participants' verbal reports as a richer window into strategies, beliefs, and latent cognitive processes. I am especially interested in when LLMs can help measure these processes reliably, and where human judgment remains essential.

I also study AI systems through ideas from psychology and neuroscience: how models explore, reason, explain decisions, and interact with people. More broadly, I am interested in AI-assisted scientific discovery that expands, rather than replaces, careful empirical work.

Before Georgia Tech, I earned an M.A. from the University of Arizona and spent three years as a full-time research assistant at Peking University's CBCS. From March to July 2025, I visited Tom Griffiths' CoCoSci Lab at Princeton University as a Visiting Student Research Collaborator.

Research Highlights

How language can inform cognitive science, what AI-assisted inference can justify, and how people can understand what AI models learn.

Language as Evidence of Thought

Choices reveal what people do, but often leave open how they arrived there. I study think-aloud reports and strategic language as complementary evidence about beliefs, strategies, and decision processes. Language can help constrain cognitive models and expose differences that behavior alone may miss; it is not a direct readout of someone's thoughts.

Across risky choice, sorting, and learning tasks, our work develops computational ways to connect what people say with what they do, while testing where those connections hold and where they break down.

Selected work:

  • Xie, H., Xiong, H., & Wilson, R. C. (2023). Text2Decision: Decoding Latent Variables in Risky Decision Making from Think Aloud Text. NeurIPS 2023 AI for Science Workshop.
  • Xie, H., Xiong, H., & Wilson, R. C. (2024). From Strategic Narratives to Code-Like Cognitive Models: An LLM-Based Approach in A Sorting Task. First Conference on Language Modeling (COLM).
  • Xie, H.†, Xiong, H. D., & Wilson, R. C. (2025). Rethinking Think-Aloud in the Age of Language Models. PsyArXiv preprint. Preprint.
  • Xie, H.†, Jagadish, A. K., Pan, L., & Wilson, R. C. (2026). Think-aloud reshapes automated cognitive model discovery beyond behavior. 9th Annual Conference on Cognitive Computational Neuroscience (CCN). Preprint.
  • Zhang, Z.*, Xie, H.*, Baker, T., Peters, M., & Wilson, R. C. (2025). Linking strategies to think aloud in a stochastic learning task. In Proceedings of the Annual Meeting of the Cognitive Science Society.

AI for Cognitive Science: What We Can and Cannot Say

AI can generate candidate models, predict behavior, and produce plausible explanations of decisions. But these successes do not by themselves establish what computations generated the behavior. My work asks what AI-assisted methods let us infer about cognition, which alternatives remain indistinguishable, and what additional evidence would separate a predictive surrogate from a mechanistic account.

This includes building models of human decisions, testing how process evidence changes model discovery, and examining cases where apparently successful models rely on the wrong explanatory structure.

Selected work:

  • Zhu, J.-Q.*, Xie, H.*, Arumugam, D., Wilson, R. C., & Griffiths, T. L. (2025). Using reinforcement learning to train large language models to explain human decisions. arXiv preprint arXiv:2505.11614.
  • Xie, H., & Wilson, R. C. (2026). Successful Automatic Model Discovery Can Produce False Mechanisms. OSF.
  • Xie, H.†, Jagadish, A. K., Pan, L., & Wilson, R. C. (2026). Think-aloud reshapes automated cognitive model discovery beyond behavior. CCN. Preprint.
  • Xie, H.*, & Zhu, J*. (2025). Centaur May Have Learned a Shortcut that Explains Away Psychological Tasks. Preprint.

Improving Human Understanding of AI Models and Artifacts

What can we learn about an AI system from its behavior, internal representations, and the artifacts it produces? I study model exploration and social reasoning with tools from cognitive science, and ask how interpretability findings can support defensible mechanistic claims rather than attractive stories. A related goal is to make knowledge found in AI systems understandable and useful to people, while testing whether that knowledge is actually grounded in the model and transferable to humans.

Current projects examine dynamic mentalizing and, through ShareMind, how people understand and use AI-generated artifacts. Claims about underlying mechanisms or human learning remain questions to test, not established outcomes.

Selected work:

  • Pan, L.*, Xie, H.*†, & Wilson, R. C. (2025). Large Language Models Think Too Fast To Explore Effectively. arXiv preprint arXiv:2501.18009. NeurIPS 2025 Poster.
  • Hu, Z., Jiang, C., Xie, H., Murty, N. A. R., & Varma, S. (2026). Mechanistic Interpretability Needs a Logic of Inference. NeurIPS 2026 Workshop on Interpretability for Discovery (accepted). Preprint.

Publications

* Denotes equal contribution, † Denotes Correspondence, Underscore denotes mentee. Use topic filters to navigate.

2026

  • Workshop AI
    Hu, Z., Jiang, C., Xie, H., Murty, N. A. R., & Varma, S. (2026). Mechanistic Interpretability Needs a Logic of Inference. NeurIPS 2026 Workshop on Interpretability for Discovery (accepted). Preprint.
  • Conference Think-Aloud Decision
    Ferguson, C., Xiong, H.-D., Xie, H., & Wilson, R. C. (2026). Characterizing Decision Strategies from Think-Aloud in Risky Choice. 9th Annual Conference on Cognitive Computational Neuroscience (CCN). OpenReview.
  • Preprint AI
    Xie, H., & Wilson, R. C. (2026). Successful Automatic Model Discovery Can Produce False Mechanisms. OSF. Google Scholar.
  • Conference Think-Aloud LLM
    Xie, H.†, Jagadish, A. K., Pan, L., & Wilson, R. C. (2026). Think-aloud reshapes automated cognitive model discovery beyond behavior. 9th Annual Conference on Cognitive Computational Neuroscience (CCN). Preprint.

2025

  • Journal Decision Social
    Qiu, S., Tang, Y., Yu, H., Xie, H., Dreher, J. C., Hu, Y., & Zhou, X. (2025). Toward a computational understanding of bribe-taking behavior. Annals of the New York Academy of Sciences.
  • Conference LLM Decision
    Zhu, J.-Q.*, Xie, H.*, Arumugam, D., Wilson, R. C., & Griffiths, T. L. (2025). Using reinforcement learning to train large language models to explain human decisions. arXiv preprint arXiv:2505.11614. ICLR 2026.
  • Conference LLM Decision
    Pan, L.*, Xie, H.*†, & Wilson, R. C. (2025). Large Language Models Think Too Fast To Explore Effectively. arXiv preprint arXiv:2501.18009. NeurIPS 2025 Poster.
  • Conference LLM AI
    Xie, H.†, Zhu, J. Q., Xiong, H. D., Wilson, R., & Griffiths, T. (2025). Reasoning Across Minds and Machines. In Proceedings of the Annual Meeting of the Cognitive Science Society (Vol. 47).
  • Conference Think-Aloud Learning
    Zhang, Z.*, Xie, H.*, Baker, T., Peters, M., & Wilson, R. C. (2025). Linking strategies to think aloud in a stochastic learning task. In Proceedings of the Annual Meeting of the Cognitive Science Society.
  • Preprint Think-Aloud LLM
    Xie, H.*, & Zhu, J*. (2025, July 12). Centaur May Have Learned a Shortcut that Explains Away Psychological Tasks. https://doi.org/10.31234/osf.io/u7z4t_v1 (submitted).
  • Preprint Think-Aloud LLM
    Xie, H.†, Xiong, H. D., & Wilson, R. C. (2025). Rethinking Think-Aloud in the Age of Language Models. PsyArXiv. https://osf.io/preprints/psyarxiv/6ta3z_v1 (submitted).

2024

  • Journal Decision Clinical
    Fang, Z., Zhao, M., Xu, T., Li, Y., Xie, H., Quan, P., ... & Zhang, R. Y. (2024). Individuals with anxiety and depression use atypical decision strategies in an uncertain world. eLife, 13.
  • Conference Think-Aloud LLM
    Xie, H., Xiong, H., & Wilson, R. C. (2024). From Strategic Narratives to Code-Like Cognitive Models: An LLM-Based Approach in A Sorting Task. First Conference on Language Modeling (COLM).
  • Conference Think-Aloud Decision
    Xie, H., Xiong, H., & Wilson, R. C. (2024). Evaluating Predictive Performance and Learning Efficiency of Large Language Models with Think Aloud in Risky Decision Making. Computational Cognitive Neuroscience (CCN), MIT.

2023

  • Journal AI
    Xie, H. (2023). The promising future of cognitive science and artificial intelligence. Nat Rev Psychology.
  • Conference Think-Aloud Decision
    Xie, H., Xiong, H., & Wilson, R. C. (2023). Text2Decision: Decoding Latent Variables in Risky Decision Making from Think Aloud Text. NeurIPS 2023 AI for Science Workshop.
  • Conference Think-Aloud LLM
    Xie, H., Xiong, H., & Wilson, R. C. (2023). Computational introspection: Can large language models reveal cognitive algorithms from human language? Poster session presented at the 5th Chinese Computational and Cognitive Neuroscience Conference, Beijing, China.

2022

  • Conference Decision Learning
    Guo, Y., Song, S., Xie, H., Gao, X., & Zhang, J. (2022, February). ARIMA and RNN for Selection Sequences Prediction in Iowa Gambling Task. In 2022 2nd International Conference on Artificial Intelligence and Signal Processing (AISP) (pp. 1-6). IEEE.

2020

  • Conference Social Learning
    Song, S*., Xie, H.*., Speekenbrink, M., Zhang, J., Gao, X., & Zhou, X. (2020, October). The computational basis of individuals' learning under uncertainty in groups with collective goals. Oral presentation at the Society for Neuroeconomics, Vancouver, Canada.

Blog

Essays and notes at the intersection of cognitive science and AI.

The Future of AI for Science: From Problem Solving to Problem Formation

Scientific discovery, understanding, and better questions

What happens when AI makes answers easier to find, and why turning discoveries into understanding matters for the questions science can ask next.

Read post

What Does It Mean to Understand?

Questions, representations, and human-centered understanding

Why understanding may be less about finding one true representation and more about preserving the right structure for the questions a person or system needs to answer.

Read post

The Understanding Bottleneck

AI, collaboration, and the cost of shared context

AI can compress the path from idea to output without equally compressing the path to shared understanding—creating a new bottleneck in review, coordination, and responsibility.

Read post

Research Taste After Implementation Becomes Cheap

Problem selection, evaluation, and AI-era research

Why AI makes the visible parts of research cheaper, and why that shifts more weight onto choosing important problems, defining honest evaluations, and distinguishing polished artifacts from understanding.

Read post

Performance Scales More Easily Than Insight

Scaling and its limits in computational cognitive science

How large behavioral datasets and powerful AI models can rapidly improve predictive performance while scientific understanding lags behind — and why data and knowledge bottlenecks matter for the future of cognitive science.

Read post

Collaborators

Mentor and Committee

Social Cognition

Think Aloud

Large Language Models and Neural Networks

Mentees

  • Zhenlong Zhang, UCLA Department of Psychology PhD Student
  • Lan Pan
  • Yangtong Feng, Wash U St. Louis

Contact

Email: hanboxie1997@gatech.edu

Address: 750 Ferst Drive, Atlanta, GA 30332

X GitHub Google Scholar