Jeff Huang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Jeff Huang has a total of 72 co-authors. Here are the top co-authors by number of shared publications:
- Tongyu Zhou
Brown University, Providence, Rhode Island, USA
8 shared publications - Joshua Kong Yang
Brown University, Providence, Rhode Island, USA
6 shared publications - Ji Won Chung
Department of Computer Science, Brown University, Providence, Rhode Island, USA
4 shared publications - Nicole R. Nugent
Alpert Medical School of Brown University
4 shared publications - Talie Massachi
University of Pennsylvania
4 shared publications - Amy Wei Xiao
Department of Human-Centered Design and Engineering
3 shared publications
Latest Publications
- Open Access
- 23 September 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Human-Computer Interaction
- Vol. 10 (6) , 1-41
Attract & Engage Visitors for Tabular Data Maintenance: a Longitudinal Naturalistic Field Study
Maintaining an evolving tabular dataset over time requires attracting and engaging enough accurate, unpaid contributors to visit and maintain it. We developed Drafty, a large publicly accessible tabular dataset of Computer Science faculty profiles. We observed Drafty’s visitors’ edits to the data, the accuracy of those edits, and engagement with the system. In March 2022, we released new add-ons to create four “sources” to attract more visitors: organic, static, asking, and dynamic. Our dynamic sources create automatic insights from our evolving tabular dataset by automatically creating aggregate statistics and sharing them on CS Open Rankings and Twitter. Visits from automatic insights made 3 times more edits than Drafty’s everyday organic visitors. Also, visits from our low-effort automatic dynamic sources and high-effort asking sources (i.e., friendsourcing) had similar accuracy. Our longitudinal study demonstrates that new features requiring minimal human effort (i.e., automatic insights) can help attract and sustain an anonymous community of visitors to maintain our large, evolving tabular dataset.
- 23 September 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Human-Computer Interaction
- Vol. 10 (6) , 1-41
Open AccessAttract & Engage Visitors for Tabular Data Maintenance: a Longitudinal Naturalistic Field Study
Maintaining an evolving tabular dataset over time requires attracting and engaging enough accurate, unpaid contributors to visit and maintain it. We developed Drafty, a large publicly accessible tabular dataset of Computer Science faculty profiles. We observed Drafty’s visitors’ edits to the data, the accuracy of those edits, and engagement with the system. In March 2022, we released new add-ons to create four “sources” to attract more visitors: organic, static, asking, and dynamic. Our dynamic sources create automatic insights from our evolving tabular dataset by automatically creating aggregate statistics and sharing them on CS Open Rankings and Twitter. Visits from automatic insights made 3 times more edits than Drafty’s everyday organic visitors. Also, visits from our low-effort automatic dynamic sources and high-effort asking sources (i.e., friendsourcing) had similar accuracy. Our longitudinal study demonstrates that new features requiring minimal human effort (i.e., automatic insights) can help attract and sustain an anonymous community of visitors to maintain our large, evolving tabular dataset.
- Open Access
- 23 September 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Human-Computer Interaction
- Vol. 10 (6) , 1-34
Therapy as an NLP Task: Comparing LLMs and Human Peers Behaviors in CBT Sessions
Large language models (LLMs) are increasingly being used as ad hoc therapists. While prior research has found that LLMs outperform human counselors in generating single-turn empathetic responses, fewer studies have compared their behaviors across multi-turn sessions. In this study, we compare the session-level behaviors of human peer counselors with those of an LLM, both trained on the same manual to deliver multi-turn, single-session Cognitive Behavioral Therapy (CBT). Our three-phase, mixed-methods study involved: (a) an 18-month ethnography of a peer support platform, where seven counselors iteratively refined CBT prompts through 110 self-counseling sessions and 60 weekly focus groups; (b) a novel session generation method that allows direct, controlled comparison of human and LLM counselors under matched conditions—client responses were drawn from publicly available human-led CBT sessions while counselor responses were generated by a CBT-prompted LLM; and (c) expert evaluations conducted by three licensed clinical psychologists. Through data triangulation, our results show a trade-off. Human peer counselors use relational techniques to interpret subtle cues, adapt CBT to users’ values and cultural contexts, and use strategies such as small talk and contextually relevant self-disclosure to build rapport and guide the session, but often at the expense of session structure and therapeutic focus. LLM counselors, on the other hand, demonstrate greater methodological adherence to CBT techniques, but struggle to sustain turn-taking, frequently fail to distinguish between clinically important and trivial content, and are more prone to lecturing and imposing solutions. LLM counselors also tend to produce “deceptive empathy”, excessively anthropomorphic responses that can inflate user expectations of genuine human care. Taken together, our findings imply that while LLMs may outperform human counselors when generating a single-turn interaction, their ability to lead multi-turn sessions is more limited, highlighting that therapy cannot be reduced to a standalone natural language processing (NLP) task. We conclude by mapping concrete design opportunities and ethical guardrails for hybrid human-AI systems, emphasizing the risks of over-attributing human relational subjectivity to current LLMs.
- 23 September 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Human-Computer Interaction
- Vol. 10 (6) , 1-34
Open AccessTherapy as an NLP Task: Comparing LLMs and Human Peers Behaviors in CBT Sessions
Large language models (LLMs) are increasingly being used as ad hoc therapists. While prior research has found that LLMs outperform human counselors in generating single-turn empathetic responses, fewer studies have compared their behaviors across multi-turn sessions. In this study, we compare the session-level behaviors of human peer counselors with those of an LLM, both trained on the same manual to deliver multi-turn, single-session Cognitive Behavioral Therapy (CBT). Our three-phase, mixed-methods study involved: (a) an 18-month ethnography of a peer support platform, where seven counselors iteratively refined CBT prompts through 110 self-counseling sessions and 60 weekly focus groups; (b) a novel session generation method that allows direct, controlled comparison of human and LLM counselors under matched conditions—client responses were drawn from publicly available human-led CBT sessions while counselor responses were generated by a CBT-prompted LLM; and (c) expert evaluations conducted by three licensed clinical psychologists. Through data triangulation, our results show a trade-off. Human peer counselors use relational techniques to interpret subtle cues, adapt CBT to users’ values and cultural contexts, and use strategies such as small talk and contextually relevant self-disclosure to build rapport and guide the session, but often at the expense of session structure and therapeutic focus. LLM counselors, on the other hand, demonstrate greater methodological adherence to CBT techniques, but struggle to sustain turn-taking, frequently fail to distinguish between clinically important and trivial content, and are more prone to lecturing and imposing solutions. LLM counselors also tend to produce “deceptive empathy”, excessively anthropomorphic responses that can inflate user expectations of genuine human care. Taken together, our findings imply that while LLMs may outperform human counselors when generating a single-turn interaction, their ability to lead multi-turn sessions is more limited, highlighting that therapy cannot be reduced to a standalone natural language processing (NLP) task. We conclude by mapping concrete design opportunities and ethical guardrails for hybrid human-AI systems, emphasizing the risks of over-attributing human relational subjectivity to current LLMs.
- Open Access
- 21 September 2026
- Published by Taylor & Francis in European Journal of Psychotraumatology
- Vol. 17 (1)
Understanding the interplay of social context and physiology on psychological outcomes in trauma-exposed adolescents: a protocol and baseline data report
1. Trauma-exposed adolescents are at risk for a host of deleterious outcomes, particularly stress-sensitive disorders such as acute stress disorder (ASD), post-traumatic stress disorder (PTSD), dep...
- 21 September 2026
- Published by Taylor & Francis in European Journal of Psychotraumatology
- Vol. 17 (1)
Open AccessUnderstanding the interplay of social context and physiology on psychological outcomes in trauma-exposed adolescents: a protocol and baseline data report
1. Trauma-exposed adolescents are at risk for a host of deleterious outcomes, particularly stress-sensitive disorders such as acute stress disorder (ASD), post-traumatic stress disorder (PTSD), dep...
- Open Access
- 12 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 4800-4814
SVGraffiti: Remixing the Web with Vector Illustrations
- 12 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 4800-4814
Open AccessSVGraffiti: Remixing the Web with Vector Illustrations
- 5 May 2026
- Published by Wiley in Neurogastroenterology & Motility
- Vol. 38 (5) , e70326
A Mixed‐Methods Analysis of Trauma‐Exposed Adolescents' Gastrointestinal Symptoms as Captured in Online Social Communication
Background: Traumatic stress can disrupt the adolescent gut‐brain axis, and unintentional social reinforcement can lead to highly impairing disorders of gut‐brain interaction (DGBI). However, social support is critical during the peritraumatic period, highlighting a unique challenge for caregivers and practitioners. The current study examined the ways in which post‐trauma gastrointestinal concerns were discussed via online social messaging (e.g., texting, social media) among a sample of trauma‐exposed adolescent patients. Methods: One hundred ninety‐one trauma‐exposed adolescents were recruited from a US children's hospital emergency department. Topic modeling and key‐word identification were used to examine naturalistic online communication (e.g., text messages), yielding a subsample of youth who used online social messaging to communicate about their gastrointestinal symptoms. Inductive content analysis was used to characterize communication about GI symptoms during the nine‐month period immediately following presentation for a potentially traumatic event. Key Results: Youth were most likely to share about symptoms with peers (i.e., friends, romantic partners) rather than adults. Responses were usually caring/affirming or offered practical support; dismissive responses were rare. GI symptoms and life stressors were often mentioned in the same conversation. Themes of avoidance and impairment were also identified. Conclusions and Inferences: Findings demonstrate the role of digital social attention in illness reinforcement, which complicates social support during the peritraumatic period. This avenue for peer reinforcement should be considered in treatment planning and discussions of family guidelines for managing symptoms.
- 5 May 2026
- Published by Wiley in Neurogastroenterology & Motility
- Vol. 38 (5) , e70326
A Mixed‐Methods Analysis of Trauma‐Exposed Adolescents' Gastrointestinal Symptoms as Captured in Online Social Communication
Background: Traumatic stress can disrupt the adolescent gut‐brain axis, and unintentional social reinforcement can lead to highly impairing disorders of gut‐brain interaction (DGBI). However, social support is critical during the peritraumatic period, highlighting a unique challenge for caregivers and practitioners. The current study examined the ways in which post‐trauma gastrointestinal concerns were discussed via online social messaging (e.g., texting, social media) among a sample of trauma‐exposed adolescent patients. Methods: One hundred ninety‐one trauma‐exposed adolescents were recruited from a US children's hospital emergency department. Topic modeling and key‐word identification were used to examine naturalistic online communication (e.g., text messages), yielding a subsample of youth who used online social messaging to communicate about their gastrointestinal symptoms. Inductive content analysis was used to characterize communication about GI symptoms during the nine‐month period immediately following presentation for a potentially traumatic event. Key Results: Youth were most likely to share about symptoms with peers (i.e., friends, romantic partners) rather than adults. Responses were usually caring/affirming or offered practical support; dismissive responses were rare. GI symptoms and life stressors were often mentioned in the same conversation. Themes of avoidance and impairment were also identified. Conclusions and Inferences: Findings demonstrate the role of digital social attention in illness reinforcement, which complicates social support during the peritraumatic period. This avenue for peer reinforcement should be considered in treatment planning and discussions of family guidelines for managing symptoms.
- Open Access
- 16 October 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Human-Computer Interaction
- Vol. 9 (7) , 1-20
Tone Indicators: Designing Accessible CMC Cues for Neurodiverse Users
As messaging becomes increasingly salient to social connection and well-being, there exists a gap in understanding and support for neurodivergent users who may struggle with interpreting tone in computer-mediated communication (CMC). We look to existing accessibility practices among neurodivergent communities to thus propose a system of alternative cues that explicitly flag tone and intent, in hopes of better scaffolding neurodiverse communication styles. Through a mixed-method study of conversational pairs (N=18), we assessed the impact of these tone indicators on participants' experiences of synchronous messaging and social connection, interpreting our findings in the context of disability-focused design for messaging interfaces. We found that these alternative cues lowered the overall cognitive load of live tone interpretation for neurodivergent users, enabling a greater sense of confidence and control over self-presentation. In turn, participants reported more positive and in-depth conversations with their partners. We further observed that some participants typed more slowly, wrote longer messages, and used additional communication cues, like emojis and text stylization, concurrently. Participants' use and feedback on this tool also illuminated social context and collaboration's role in cue ambiguity. Our work subsequently suggests future research and design directions on supporting communicative accessibility via the intentional design of alternative cue systems and co-customization.
- 16 October 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Human-Computer Interaction
- Vol. 9 (7) , 1-20
Open AccessTone Indicators: Designing Accessible CMC Cues for Neurodiverse Users
As messaging becomes increasingly salient to social connection and well-being, there exists a gap in understanding and support for neurodivergent users who may struggle with interpreting tone in computer-mediated communication (CMC). We look to existing accessibility practices among neurodivergent communities to thus propose a system of alternative cues that explicitly flag tone and intent, in hopes of better scaffolding neurodiverse communication styles. Through a mixed-method study of conversational pairs (N=18), we assessed the impact of these tone indicators on participants' experiences of synchronous messaging and social connection, interpreting our findings in the context of disability-focused design for messaging interfaces. We found that these alternative cues lowered the overall cognitive load of live tone interpretation for neurodivergent users, enabling a greater sense of confidence and control over self-presentation. In turn, participants reported more positive and in-depth conversations with their partners. We further observed that some participants typed more slowly, wrote longer messages, and used additional communication cues, like emojis and text stylization, concurrently. Participants' use and feedback on this tool also illuminated social context and collaboration's role in cue ambiguity. Our work subsequently suggests future research and design directions on supporting communicative accessibility via the intentional design of alternative cue systems and co-customization.
- Open Access
- 12 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 455-459
Designing Systems for Long-term Sensemaking
- 12 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 455-459
Open AccessDesigning Systems for Long-term Sensemaking
- Open Access
- 27 September 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-15
L.ink: Procedural Ink Growth for Controllable Surprise
- 27 September 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-15
Open AccessL.ink: Procedural Ink Growth for Controllable Surprise
- Open Access
- 4 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 1156-1164
Copyrighting Generative AI Co-Creations
- 4 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 1156-1164
Open AccessCopyrighting Generative AI Co-Creations
- Open Access
- 25 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-17
VidSTR: Automatic Spatiotemporal Retargeting of Speech-Driven Video Compositions
- 25 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-17
Open AccessVidSTR: Automatic Spatiotemporal Retargeting of Speech-Driven Video Compositions