My research asks how people learn to give feedback that engages with others’ work rather than only judging it. I started with experts in academic peer review, then turned to novices in graduate courses. I also study how people learn from each other in online communities. Across these projects, I combine qualitative methods, such as interviews and content analysis, with computational methods, such as epistemic network analysis, social network analysis, and topic modeling.
Peer review is one of the main ways scholars build knowledge together, but little is known about how this happens in the exchanges between reviewers and authors. Analyzing published review records from a Nature journal, my colleagues and I identified three forms of knowledge co-construction: quick consensus, negotiated integration, and tested integration. Across these forms, reviewers’ questions led to more productive co-construction than their suggested solutions. I have since extended this work to conference review at ICLR. There, reviewers with different levels of confidence showed different patterns of epistemic cognition, and ICLR’s open, computer-supported review led to more discussion among reviewers than review at a Nature journal.
Understanding expert practice is not enough to guide novices. Research on feedback literacy has focused mainly on how students make sense of and use the feedback they receive; I study what it takes to give it. My dissertation examines a doctoral writing course in which students took part in scaffolded peer review workshops, using interviews to study how they develop the ability to give feedback. Preliminary findings show that novices struggle most with making sense of peer drafts and with seeing themselves as reviewers. A related study examines the same question in a doctoral reading club, where students learned to give feedback without a set structure. To understand how new reviewers are prepared, I also reviewed peer review training across publishers, universities, and conferences, finding that training varies widely, and proposed directions based on the cognitive apprenticeship model.
Students often treat peer feedback as a task to complete rather than a chance to understand a peer’s work. This is usually blamed on limited domain knowledge, but research on framing suggests another explanation. Students may frame the same activity as inquiry, first trying to make sense of a peer’s work and then evaluating it, or as task completion, doing what the assignment requires. Framing is cued by how a task is structured and by the materials and tools students use. A detailed rubric, for example, may lead students to check items off a list rather than follow the author’s reasoning. Using design-based research, I will design, test, and refine peer feedback activities, such as open-ended reading prompts and collaborative questioning, in undergraduate and graduate courses, working toward design principles and tools that instructors can adopt.
AI is part of this work. As AI makes it easy for students to hand off their judgment, peer feedback becomes a setting where they can practice being accountable for it. An AI tool that writes comments for students may cue task completion even more strongly than a detailed rubric. I will examine whether AI can be designed to do the opposite, by asking reviewers questions while leaving the judgment to them, and where such support falls short. My team built a prototype of this kind at the 2026 NYEDTECH Hackathon, where it won second place.
Much learning happens outside school, in online communities where people teach and learn from each other. On Kaggle, a data science competition platform, I used social network analysis to study how people form teams, and found that experts and novices differ in who joins team-seeking forums, who benefits from them, and who goes on to win. In an online AI painting community, I used topic modeling of more than 180,000 posts to trace how members discussed and learned about AI tools as the technology changed.