About Me

Hi, I’m Trenton! I am a final-year EECS PhD Candidate at the University of Michigan AI lab, advised by Jenna Wiens. During Summer 2024, I interned at Microsoft Research Redmond with the Augmented Reasoning & Learning Group mentored by Adith Swaminathan and Tobias Schnabel. I’m on the industry job market this year, looking for Research Scientist roles.

I work on ML alignment with domain-specific values throughout the entire ML lifecycle, from training to post-deployment. To date, I’ve worked on:

  • Aligning ML models with fairness criteria when learning from biased labels (MLHC ‘22, ICML ‘24)
  • Measuring the steerability of LLMs, with applications to personalization (NeurIPS SafeGenAI ‘24, pre-print)
  • Ensuring ML models are resistant to abuse post-deployment, with applications to health insurance risk adjustment (NeurIPS ‘24)

My work often takes inspiration from healthcare and policy. To that end, I’ve worked with interdisciplinary teams featuring clinicians, economists, bioethicists, data scientists, and more. Technically, my work has touched areas including ML fairness, strategic classification, causal inference, and LLM evaluation.

I was previously an CS MS student @ Stanford (2021) and earned a BA in American Studies along the way (2020). Earlier I worked in video machine learning robustness with HazyResearch and open-domain conversational AI for the Alexa Grand Prize Socialbot Challenge with the Stanford NLP Group. I was also a contributor to Google BIG-bench.

Contact. If you’re another researcher in the machine learning, causal inference, or healthcare space–please reach out if you want to chat about research!

For current Michigan students interested in working with me, please email me to set up a one-off coffee chat if you’d like to discuss some research ideas.

Email: ctrenton (at) umich (dot) edu

Recent News

  • [8/6/2025] I started performing as a jazz pianist in Ann Arbor! Anyway, that inspired me to write some thoughts about jazz + GenAI.

  • [7/30/2025] I was a finalist for the College of Engineering’s Beyster Computational Fellows Program, representing Michigan CSE.

  • [7/30/2025] Passed my thesis proposal exam (title: Predictions as behaviors: Steering machine learning models towards high-level principles).

  • [5/28/2025] My pre-print on the steerability of LLMs with Microsoft Research & Netflix is now on arXiv! We’ve open-sourced our evaluation framework on [Github]. Learn more at our website and visualize some of our results here.

Selected Publications

Chang, Trenton, Schnabel, Tobias, Swaminathan, Adith, and Wiens, Jenna. A Course Correction in Steerability Evaluation: Revealing Miscalibration and Side Effects in LLMs. arXiv preprint. [pre-print] [website] [demo]

Jabbour, Sarah*, Chang, Trenton*, Antar, Anindya Das*, et. al. Evaluation Framework for AI Systems in “the Wild.” arXiv preprint. [paper]

Chang, Trenton, Warrenburg, Lindsay, Park, Sae-Hwan, Parikh, Ravi B., Makar, Maggie, and Wiens, Jenna. Who’s Gaming the System? A Causally-Motivated Approach for Detecting Strategic Adaptation. NeurIPS 2024. [paper]

Chang, Trenton, Nuppnau, Mark, He, Ying, Kocher, Keith E., Valley, Thomas S., Sjoding, Michael W., Wiens, Jenna. Racial differences in laboratory testing as a potential mechanism for bias in AI: A matched cohort analysis in emergency department visits. PLOS Global Public Health. [paper]

Chang, Trenton, Wiens, Jenna. From Biased Selective Labels to Pseudo-Labels: An Expectation-Maximization Framework for Learning from Biased Decisions. ICML 2024. [paper] [code]

Chi, Ethan A., Paranjape, Ashwin, See, Abigail, Chiam, Caleb, Chang, Trenton et. al. Neural Generation Meets Real People: Building a Social, Informative Open-Domain Dialogue Agent. SIGDIAL 2022. [paper] [video]

Chang, Trenton, Sjoding, Michael W., and Wiens, Jenna. Disparate Censorship & Undertesting: A Source of Label Bias in Clinical Machine Learning. MLHC 2022. [paper] [video]

Srivastava, Aarohi, …, Chang, Trenton, …, et. al. Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models. TMLR 2022. [paper]

Chi, Ethan A., Chiam, Caleb, Chang, Trenton et. al. Neural, Neural Everywhere: Controlled Generation Meets Scaffolded, Structured Dialogue. Alexa Prize Proceedings 2021. [paper]

Service

  • Reviewed: AAAI, AISTATS, CHIL, ICLR, ICML, MLHC, ML4H, KDD (workshop), NeurIPS, TMLR. Best Reviewer Award at Research2Clinics, 2021 NeurIPS workshop.
  • Outreach Chair, ML4H Symposium Program Committee (2025)
  • Michigan AI Blog Co-Coordinator (2024-2025)
  • Workflow Chair, ML4H Symposium Program Committee (2024)
  • University Relations Chair, Computer Science & Engineering Graduate Student Organization, University of Michigan (2023-2024)
  • Panelist, Summer Research Opportunity Program, University of Michigan (2023)
  • AI Lab Graduate Admissions Committee Volunteer, Division of Computer Science & Engineering, University of Michigan (2022 & 2024)

Teaching & Mentoring

Undergraduate/graduate level

  • [08/2023 - 12/2023] Graduate student instructor, Causality and Machine Learning, EECS 598-009, University of Michigan
  • [01/2021 - 03/2021] Research Mentor, Stanford ACM

K-12 level

  • [07/2023 - 08/2023] Workshop organizer, Xplore Engineering & Discover Engineering, Division of Computer Science & Engineering, University of Michigan
  • [07/2022] Instructor, AI4ALL
  • [06/2020 - 08/2020] Instructor, Inspirit AI
  • [06/2019 - 08/2019] Residential Counselor/Teaching Assistant for Artificial Intelligence, Stanford Pre-Collegiate Studies