Research
Active and past lines of research, plus tools and datasets built along the way. Tagged by cluster — see the graph for how they connect.
Content for this page is in progress. In the meantime, the research graph and CV are live, and there's a short colophon if you're curious how this site was built.
How can generative models help people break out of creative ruts? Work at TRI on surprise, taste, and physiological signals as steering cues for human–AI co-creation, with a series of CHI / HCI workshop papers from 2022–2026.
Ubiquitous, in-the-moment prompts that surface alternative framings at the exact moment someone gets stuck. Builds on emotion-regulation research and ambulatory-intervention methods.
Probabilistic models of what people find pleasing, surprising, or wow-worthy in generated artifacts. Extends preference and valuation work from neuroeconomics into the design and AI-generation setting.
fMRI investigations of intertemporal choice, dietary self-control, and how the brain values delayed rewards. Postdoctoral work at Duke with Scott Huettel.
How the brain represents trust, social value, and reward in social contexts. Oxytocin, theory-of-mind, fMRI. Predoctoral research at NIMH with Andreas Meyer-Lindenberg.
Cognitive reappraisal as a treatment mechanism in social anxiety disorder; neural correlates of CBT response. Stanford with Philippe Goldin and James Gross.
Theoretical and empirical work on how the brain models its own predictions — with applications to expectation, surprise, and attention. [Placeholder: confirm scope/years.]
Force-directed visualization of 50+ publications grouped into seven conceptual clusters, with a marching-squares contour overlay and spreading-activation interaction. The conceptual center of this site.
View the graph →A Claude-based CLI tool that ingests a DOI or full-text PDF and emits cluster-tagged concept nodes into the graph. Lightweight, idempotent, and dependency-free beyond the API client.
Node script that reads LaTeX .bib files from a private CV repo and emits the JSON powering the CV and graph pages — so the website never falls out of sync with the canonical publication list.