How can we ensure the validity of participatory research methods in HCI for mental health in the current AI age? HCI research often overlooks the gap between the actual tech- and AI- literacy of participants and that assumed by experimental designs. This assumption risks the validity of research findings, as results predicated on participants’ faulty conceptions of what AI is may not reflect the reality of the problem being investigated. My thesis investigates the effects of AI (il)literacy on participatory research methods through a mixed-methods study in the form of a series of semistructured interviews with mental health practitioners that evaluate if and how their perceptions towards mental health change pre and post educational intervention. I aim to quantify this gap with established survey tools and support the findings with qualitative interview data.

This is an ongoing project!

• View Proposal

Given that LLMs have demonstrated significant biases (eg. racism, sexism, heteronormativity, etc.), what harm might arise when they meet with the historically biased field of mental health? This project aimed to evaluate diagnostic prejudices in LLMs based on intersecting identities presented in patient profiles. I conducted a pilot study and presented initial results as a poster at KDD 2025. I also wrote an essay exploring the risks and potentials of LLMs in mental health care, which was published recently.

Transformative justice is an alternative approach to creating community safety that does not depend on the harmful carceral norms of policing. TJ processes, which address harm when it occurs, are community-based and rooted in accountability, with the goal of collective sensemaking in order to move towards shared understanding, repair, and healing. Importantly, it also seeks to transform the conditions that facilitated the harm to ensure that it is less likely, or impossible, for it to occur in the future. To learn more, Partisse Cullors wrote this great piece for the Harvard Law Review. If you'd like more book recommendations I'm happy to share! You can find my email at the bottom of this website.

REPaiR is a platform we are developing to support the work of transformative justice practitioners. Based on a participatory design process, we are exploring practitioners' current attitudes towards and uses of technology, identifying ways that technology can support their practises, working to establish safety guardrails, and designing a trauma-informed platform based on the community's positionality.

This is an ongoing project!

• View Poster

Ferrari, T., Epps, A., Ames, R., Wang, C. (2025). Participatory AI Research and Development to Support Transformative Approaches to Justice and Community Safety. ACM Collective Intelligence Conference, La Jolla, California, USA, August 4-6, 2025. Poster presentation.

This project aimed to understand the varying degrees of alignment between foundational LLMs and the values held by transformative justice practitioners, in order to make informed decisions about whether and how LLMs can be implemented in this context. I led the development of an evaluation protocol rooted in community values (as identified by part of our research process for REPaiR) in order to assess this alignment.

• View Paper

Wang, C., Epps, A., Ferrari, T., Ames, R. (2025). AI for Abolition? A Participatory Design Approach. In Proceedings of the Workshops at the Fourth International Conference on Hybrid Human-Artificial Intelligence. (Selected for oral presentation)

Speculative fiction for tech-driven menstrual equity completed as a final project for a graduate seminar course on HCI along with two other students. Through a series of design exercises analyzing a speculative technology through various value-oriented lenses, we created a set of fictional artifacts exploring how this technology could support (or hinder) menstrual justice.

• View Project

This was a final project from a graduate seminar course on generative AI and LLMs that I completed with two other students. We developed a novel-diffusion-based model architecture, inspired by discrete-diffusion models for text data, to perform symbolic regression and benchmarked it against an existing deep learning approach.

• View Project

I worked on a project to improve upon prior sex-specific kidney function models which enable researchers to understand risk factors for kidney injury, accounting for the historically overlooked biological differences between sexes. My work contributed to a publication in Mathematical Biosciences.

• View Paper

Leete, J., Wang, C., Lopez-Hernandez, F. J., & Layton, A. T. (2022). Determining risk factors for triple whammy acute kidney injury. Mathematical Biosciences, 347, 108809.