Uploaded February 2025 | Updated September 2026, 3 days ago
Abstract: Peer review is a cornerstone of scientific progress. However, the ecosystem faces growing challenges including increasing submission volumes, reliance on first-time reviewers, and AI-generated content. My research explores the potential and perils of leveraging advances in AI to facilitate and strengthen this ecosystem through interactive scaffolding for reviewers, meta-reviewers, and authors. While traditional guidelines offer only static, high-level advice, my work envisions how to create AI-driven interactive guidance that is contextual, personalized, and self-reflexive. My ReviewFlow system offers contextual cues, in-situ knowledge support, and notes-to-outline synthesis to help novice reviewers write better reviews with greater confidence. Likewise, my MetaWriter system investigates the value of extractive summarization and generative summaries to assist meta-reviewing. Despite benefits, both studies revealed friction for key social considerations—such as loss of agency, laziness, and skepticism in how others appropriate these tools—that pose challenges for broader deployment. My work envisions a future where generative AI research agents act like training wheels for new researchers—providing structured guidance until they gain confidence and expertise, ultimately enhancing the peer review ecosystem, facilitating scientific discovery, and fostering collaborative knowledge production.
Bio: Lu Sun is a Ph.D. candidate in the Cognitive Science Department and the Design Lab at UC San Diego, advised by Prof. Steven Dow and Prof. Kristen Vaccaro. Her research sits at the intersection of Human-Computer Interaction (HCI), Human-Centered AI, Natural Language Processing (NLP), and Learning Science. Her work takes inspiration from cognitive science and education theory to design and develop AI-driven scaffolding systems to support complex knowledge work. She investigates how interactive AI guidance can enhance the academic peer review ecosystem, providing scaffolding for novice reviewers, meta-reviewers, and authors. She employs a mixed-methods approach to examine both the opportunities and societal
challenges of human-AI collaboration. During her doctoral research, she has collaborated with Microsoft Research and Amazon Science. For more information: lusunhci.github.io/site.
Abstract: Peer review is a cornerstone of scientific progress. However, the ecosystem faces growing challenges including increasing submission volumes, reliance on first-time reviewers, and AI-generated content. My research explores the potential and perils of leveraging advances in AI to facilitate and strengthen this ecosystem through interactive scaffolding for reviewers, meta-reviewers, and authors. While traditional guidelines offer only static, high-level advice, my work envisions how to create AI-driven interactive guidance that is contextual, personalized, and self-reflexive. My ReviewFlow system offers contextual cues, in-situ knowledge support, and notes-to-outline synthesis to help novice reviewers write better reviews with greater confidence. Likewise, my MetaWriter system investigates the value of extractive summarization and generative summaries to assist meta-reviewing. Despite benefits, both studies revealed friction for key social considerations—such as loss of agency, laziness, and skepticism in how others appropriate these tools—that pose challenges for broader deployment. My work envisions a future where generative AI research agents act like training wheels for new researchers—providing structured guidance until they gain confidence and expertise, ultimately enhancing the peer review ecosystem, facilitating scientific discovery, and fostering collaborative knowledge production.
Bio: Lu Sun is a Ph.D. candidate in the Cognitive Science Department and the Design Lab at UC San Diego, advised by Prof. Steven Dow and Prof. Kristen Vaccaro. Her research sits at the intersection of Human-Computer Interaction (HCI), Human-Centered AI, Natural Language Processing (NLP), and Learning Science. Her work takes inspiration from cognitive science and education theory to design and develop AI-driven scaffolding systems to support complex knowledge work. She investigates how interactive AI guidance can enhance the academic peer review ecosystem, providing scaffolding for novice reviewers, meta-reviewers, and authors. She employs a mixed-methods approach to examine both the opportunities and societal
challenges of human-AI collaboration. During her doctoral research, she has collaborated with Microsoft Research and Amazon Science. For more information: lusunhci.github.io/site.
