Researcher

Research

My work includes studies of how machine learning literacy shapes people's perceptions of algorithmic fairness, evaluations of language model judgments, research on models' internal representations, and AI-supported learning. These projects use different methods and are at different stages.

More recently, I've been exploring what can be learned from models' internal representations and whether AI tutors can better identify students' misconceptions, track their understanding, and support their learning.

Affiliations

Selected Research

Algorithmic fairness and human perceptions

RISE Lab.

Research on how learning about machine learning affects people's preferred definitions of algorithmic fairness. Related work on interdisciplinary approaches to measuring perceptions of fairness in machine learning was published at EAAMO 2021 (listed under Peer-reviewed publications).

Language model evaluation

Studies of how language model judgments respond to framing and context, including race and gender bias in hiring decisions, persuasion, harm, and refusal. Sycophancy and contrarianism are also topics under investigation. These investigations use different methods. The hiring-bias replication was presented in October 2024.

Model representations

Research using linear probes to investigate what information can be recovered from language models' internal representations, including whether clickability can be recovered from headlines with observed click-through rates. This work examines what information is present in a model's representations. It does not claim that probing reveals the causal mechanisms behind the model's judgments. Presented in November 2025.

Evaluating AI Tutors and Learner Models

Ongoing research.

We're investigating how AI tutoring systems identify students' misconceptions, estimate what they've learned, and decide what support to provide next. Our experiments compare large language models, Bayesian Knowledge Tracing, and hybrid approaches that combine conversational diagnosis with structured learner models.

Interpreting emoji-based product sentiment

SeqHub Analytics, 2018–2021.

I served as the primary AI research contributor for a company exploring how emoji responses could help interpret sentiment toward products. The approach compared emotion-bearing words with words describing emoji meanings, clustered emojis, and analyzed bursts of responses to infer the dominant expressed sentiment.

Talks

  • Audio Intelligence: A Transparent Framework for Analyzing Long-Form Speech. With Dr. Andre Martin, University of Notre Dame. TRAP Lab, September 2026.
  • Exploring the Representation of Clickability in Large Language Models. TRAP Lab, November 2025.
  • Towards Evaluating Look-Think-Do: AI Coteacher and Learning Management Platform. TRAP Lab Unconference, June 2025.
  • Look-Think-Do: An AI Co-teacher. With J Phillips. Yale Innovation Summit, Yale Ventures, May 2025.
  • The Look-Think-Do Approach: AI-Assisted Education for Agency and Creativity. TRAP Lab, April 2025.
  • GPT's Race and Gender Biases in Hiring: Replication and Insights. TRAP Lab, October 2024.

Research reports

Apart Research, 2026.

Probing LLM Preferences: Demographic Framing, Elicitation Context, and Incentive-Driven Trade-offs.

Research sprint report.

Read report

Peer-reviewed publications

Boykin, C.M., Dasch, S.T., Rice, V., Lakshminarayanan, V.R., Togun, T.A., Brown, S.M. (2021).

Opportunities for a More Interdisciplinary Approach to Measuring Perceptions of Fairness in Machine Learning.

EAAMO ’21, ACM.

View DOI

Hui, P., Gysler, S., Uduman, M., Togun, T.A., et al. (2018).

MicroRNA Signatures Discriminate Between Uterine and Ovarian Serous Carcinomas.

Human Pathology, 73, 130–138.

View DOI

Vassilakopoulou, M., Togun, T., Dafni, U., Cheng, H., Bordeaux, J., et al. (2014).

In Situ Quantitative Measurement of HER2 mRNA Predicts Benefit from Trastuzumab-Containing Chemotherapy in a Cohort of Metastatic Breast Cancer Patients.

PLoS ONE, 9(6), e99131.

View DOI

Liu, Z., Hu, Z., Pan, X., Li, M., Togun, T.A., et al. (2011).

Germline Competency of Parthenogenetic Embryonic Stem Cells from Immature Oocytes of Adult Mouse Ovary.

Human Molecular Genetics, 20(7), 1339–1352.

View DOI

Other writing

Togun, T.A. (2022).

Purpose and Problem Solving: Developing Leaders in the Classroom.

Next Gen Learning.

Read article