CV
Education
- Ph.D. in Information Science (Human-Computer Interaction), Cornell University, Aug. 2021 - Dec. 2026
- B.A. in Linguistics (Dean’s List), University of California, Berkeley, Aug. 2016 - Dec. 2019
Selected Research
- Conversational Dynamics & Multimodal Prediction in Computer-Mediated Communication, Jan. 2022 - Aug. 2024
- Multimodal cues as externalizations of predictive state: Established that prosody, gaze, and facial expression function as externalizations of speakers’ and listeners’ real-time predictive models — when these channels are progressively suppressed (video → audio → transcript), word prediction degrades systematically in both accuracy and sensitivity to affect (COLING 2025; Proc. of the 31st Int’l Conference on Computational Linguistics, pp. 7949–7962)
- Novel computational measure of prediction: Developed an LLM-based method to approximate human next-word prediction in real-time speech, replacing traditional neuroscience paradigms with a scalable corpus-and-LM approach for quantifying prediction failure in conversational turn-taking
- Differential cost of lean channels for non-native speakers: Showed that L2 listeners are disproportionately affected by the loss of multimodal cues — limited L2 exposure and negative affect compound the cost of lean channels, leaving NNS with higher processing load and less reliable prediction than native speakers under the same conditions (CogSci 2025; eScholarship)
- Statistical modeling pipeline: Built reproducible analysis pipelines for lab-collected behavioral data combining GLMMs (lme4, lmerTest), LASSO-regularized regression, and multimodal behavioral coding (ELAN, Praat) across in-lab studies (cumulative N = 300+)
- Theoretical contribution: Identified which cognitive mechanisms break down in which CMC modalities — providing the empirical foundation for XPLAIN’s intervention design
- AI-Mediated Communication as Cognitive Scaffolds — XPLAIN, Aug. 2024 - Present
- Theory-driven system vision: Designed a real-time AI scaffold for non-native speakers in online meetings, grounded in the prediction-breakdown findings above; targets cognitive bottlenecks (semantic prediction failure, temporal pressure, comprehension/production trade-offs) at specific stages of speech processing (CHI EA 2026, CSCW 2025, CUI 2026 under review)
- Novel cognitive-fluency metric: Developed a psycholinguistic measure that synchronizes four multimodal behavioral channels — speech disfluencies, gaze shifts, mouse clicks, and facial expressions — with specific speech-processing stages, enabling real-time inference of emerging cognitive trouble
- Phase 1 — Concept validation via three Wizard-of-Oz studies: Designed and ran three lab-based WoZ studies in which a confederate simulated the AI agent’s anticipatory cues, rephrasing, and timing adjustments to validate intervention type, timing, and individual fit:
- Baseline scaffolding study (N = 38): Evaluated three core scaffolds — lexical clarifications, idea/content suggestions, and topic summaries — in dyadic virtual meetings; demonstrated improvements in 30–80% perceived communicative efficiency, participation, and inclusivity, moderated by individual differences in language proficiency, personality, and prior AI experience (short WIP at CSCW 2025; full manuscript in preparation)
- Personalization study (N = 40): Tested context-adaptive interventions calibrated to user-level cognitive dimensions; companion latent class analysis on qualitative interview data (n = 29) identified eight two-class groupings across thematic domains (e.g., AI usage and trust, communication strategies, feature-specific evaluations), revealing that user profiles reflect domain-specific adaptation patterns rather than a global typology (CogSci 2026, accepted)
- Translation study (N = 27): Examined how L1- vs. L2-cued translation scaffolds support real-time turn-taking under domain-specific lexical demand, integrating prediction theory, cognitive load theory, and language-dependent memory theory (manuscript in preparation)
- Phase 2 — Real-time evaluation pipeline for naturalistic conversation: Engineered an evaluation pipeline that ingests synchronized multimodal streams (speech, gaze, mouse-click, facial expression) during live online-meeting conversations and outputs real-time estimates of users’ performance and cognitive states (emerging prediction trouble, processing load, repair initiation); modeled intervention effects with event-centered GLMMs and interaction-state shifts via latent transition models; cumulative 100+ customer-discovery interviews
- Listener perception (pre-registered, in-field): Designed a pre-registered two-phase blind/informed listener-perception study (OSF) measuring listeners’ implicit detection of AI-assisted conversational speech, acoustic cue attribution, and the moderating effect of prior AI-knowledge — testing whether the same multimodal traces that enable AI scaffolding (disfluency, timing, prosody, visible ease/strain) trigger negative attributions of competence and authenticity from native evaluators, surfacing a “double bind” for non-native speakers using AI support
- Translation to product & impact: Selected for the Cornell eLab Spring 2025 cohort with $5,000 prototype-build funding; identified educational and workplace use cases for bridging knowledge gaps and reducing communication-driven social friction
- Sustaining Public Goods via Prosocial Behavior, Citizens & Technology Lab (PI: J. Nathan Matias), Cornell University, Jan. 2025 - Present
- Lead statistician on parallel pre-registered field experiments across four Wikipedia language communities (Arabic, German, Persian, Polish; N = 15,558 editors), quantifying the causal effect of peer-to-peer gratitude on upstream reciprocity and sustained volunteer participation
- Built end-to-end causal inference pipelines in R: power analyses, intent-to-treat (ITT) and complier-average causal effect (CACE) estimation, multilevel models with cluster-robust standard errors, sensitivity analyses, and cross-community meta-analytic pooling
- Designed reproducible analysis & visualization workflow handling longitudinal behavioral data (edits, contributions, retention) over multi-month observation windows
- Translated statistical findings into actionable strategies for online community health and social sustainability; manuscript under review at PNAS
- Multilingual FrameNet Project, International Computer Science Institute, Aug. 2019 - Dec. 2019
- Studied structures of English lexical databases, ran model trainings on semantic relations, and refined labelings for multilingual alignment
Professional Experience
- Founder & Product Lead, XPLAIN (Proactive Meeting AI), Aug. 2024 - Present
- Architected a real-time AI-MC system utilizing LLM-based cognitive scaffolding to mitigate miscommunication in online meetings
- Recruited & directed a 19-person cross-functional research team (UX, Engineering, Data)
- Secured traction via venture showcases & demos via Cornell eLab incubator
- Content Strategist (Trust and Safety), ByteDance Inc., Dec. 2020 - Jun. 2021
- Architected a new content moderation framework for sensitive word detection AI models across TikTok, Helo, Fictum (>500M users)
- Established HITL data-labeling standards and QA evaluation metrics for AI safety models
- Linguist Intern - Ontology in NLU, Facebook, July - Aug. 2019
- Built the first Mandarin NLU model for Facebook AI Assistant chatbot with culture-specific syntactic and semantic adaptations
Technical Skills
- Interactive AI Systems & Prototyping: Wizard-of-Oz concept validation, real-time multimodal evaluation pipelines (synchronized speech, gaze, mouse, facial expression streams) for naturalistic conversation, web-based experiment platforms, full-stack MVP development
- AI/LLM Methods: Prompt & context engineering, in-context learning, LLM-based behavioral simulation, surprisal & next-word-prediction extraction from LMs, speech transcription (Whisper), LSTM/sequence-model fine-tuning (PyTorch, CANDOR multimodal corpus), AI agent task design & benchmarking, human-in-the-loop (HITL) evaluation protocols
- Quantitative & Statistical Modeling: GLMMs / multilevel models, causal inference (field experiments, parallel design), latent class & latent transition analysis, regression with regularization (LASSO), time-series prediction, event-centered analyses
- Qualitative & Mixed-Methods Research: Semi-structured interviews, inductive thematic coding, contextual inquiry, focus groups, multimodal behavioral coding/annotation, customer-discovery interviews (100+), worker co-design
- Research Methods: Experimental design, user studies (80+ in-lab), surveys, usability testing, prototyping
- Programming & Tools: Python (pandas, NumPy, scikit-learn, statsmodels, OpenAI/Anthropic APIs), R (tidyverse, lme4, lmerTest), SQL, JsPsych, PsychoPy, Praat, ELAN, Qualtrics, Atlas.ti, Figma, Miro, Git/GitHub, LaTeX, JupyterLab
Languages
- Mandarin Chinese — Native
- English — Fluent
- German — B2 (Goethe-Institut PASCH scholar, 2015)
Publications
He, W. P. & Fussell, S. R. (2026). "User Heterogeneity in AI-Mediated Communication: Extending Cognitive Theories via Latent Class Analysis." Proceedings of the 48th Annual Meeting of the Cognitive Science Society (CogSci 2026). Rio de Janeiro, Brazil. (Accepted)
He, W. P. & Fussell, S. R. (2026). "XPLAIN: A Proactive Scaffold Across Speech Processing Stages—Supporting Non-Native Speakers in Real-Time AI-Mediated Turn-Taking." Proceedings of the ACM Conference on Conversational User Interfaces (CUI 2026). (Under review)
He, W. P. (2026). "Disfluency as a Window into Cognitive Mediation: Psycholinguistic Metrics for Evaluating AI-Integrated Spoken Communication." Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems. Barcelona, Spain.
He, W. P. & Fussell, S. R. (2025). "Proactivity in Scaffolding Comprehension and Production in Real-Time Turn-Taking: A Case Study of Bridging Communication Gaps for Non-Native Speakers." Companion Publication of the 28th ACM Conference on Computer-Supported Cooperative Work and Social Computing (CSCW Companion '25). Bergen, Norway. ACM.
He, W. P. & Fussell, S. R. (2025). "Difference in the Cognitive Mechanism of Predictive Processing in Computer-Mediated Communication: A Comparison Study of L2 Speakers." Proceedings of the 47th Annual Meeting of the Cognitive Science Society (CogSci 2025). San Francisco, CA.
He, W., MacDonald, C. C., Yoo, Y., Eizayaga, M., Shim, R., Katreczko, L. D., & Fussell, S. R. (2025). "Exploring Content Predictability in Turn-taking Through Different Computer-Mediated Communications." Proceedings of the 31st International Conference on Computational Linguistics (COLING 2025), pages 7949–7962. Abu Dhabi, UAE. Association for Computational Linguistics.
Matias, J. N., Kamin, J., Al-Kashif, R., He, W. P., Klein, M., & Pennington, E. "How Thanking Peers Sustains Volunteer Participation in Public Goods: Parallel Field Experiments in Four Wikipedia Language Communities." Proc Natl Acad Sci USA (PNAS). (Under review)
Manuscripts In Preparation
- He, W. P. & Fussell, S. R. Cognitive Trade-offs Between Speech Disfluencies and Active Engagement from Proactive AI-Mediated Scaffolds in Turn-taking.
- He, W. P. & Fussell, S. R. Context-Adaptive Translations for Optimized Processing under Time-Pressured AI-Mediated Turn-Taking for Non-Native Speakers.
- He, W. P. & Fussell, S. R. Deploying Cognitive Measures as Personalization Metrics for Proactive AI-Mediated Scaffolds in Turn-taking: A Mixed-Methods Study.
Talks & Presentations
April 13, 2026
Conference Presentation at CHI Conference on Human Factors in Computing Systems (CHI EA '26), Barcelona, Spain
October 18, 2025
Conference Presentation at ACM Conference on Computer-Supported Cooperative Work (CSCW Companion '25), Bergen, Norway
July 30, 2025
Conference Presentation at 47th Annual Meeting of the Cognitive Science Society (CogSci), San Francisco, CA, USA
January 19, 2025
Conference Presentation at 31st International Conference on Computational Linguistics (COLING), Abu Dhabi, UAE
January 01, 2024
Poster at The International Conference on Interdisciplinary Advances in Statistical Learning, San Sebastian, Spain
January 01, 2022
Poster at The International Conference on Interdisciplinary Advances in Statistical Learning, San Sebastian, Spain
January 01, 2021
Talk at The 28th Manchester Phonology Meeting, Online
January 02, 2020
Talk at The 56th Linguistics Colloquium, Online
January 01, 2020
Poster at 22nd Biennial International Conference of Infant Studies, Online
Teaching
Teaching Assistant, Cornell University (2022–2025):
Guest Lectures:
- INFO 4450 — Content Prediction in the Fluency of Computer-Mediated Communication (Spr 2024)
- INFO 6310 — Mechanisms of Content Prediction in Computer-Mediated Communication (Fall 2024)
Student Mentorship (INFO 4900 Independent Study): Coached 25 undergrads and 14 MEng students in research execution and professional development.
Grants & Awards
- Entrepreneurship Lab (eLab) Start-Up Fund, Cornell University, 2025 — $5,000
- Research & Travel Grant, Cognitive Science Program & Cornell University, 2022, 2024, 2025 — $5,000
- SAGE Fellowship, Cornell University, 2021-2023 — ~$65,000/year
- Linguistics Society of America Summer Grant, UC Berkeley, 2019 — $1,500
- International PASCH Junior Scholarship, Goethe-Institut, Germany, 2015 — $15,000
Service & Leadership
- Journal Club Co-Lead, Cognitive Science of Language Lab (PI: Prof. Morten H. Christiansen), Cornell University, 2021 - 2023 — Led weekly discussions on cutting-edge research in psycholinguistics, computational linguistics, and language cognition for two academic years
- W.E. Cornell, 2023-24 Cohort — Guided entrepreneurship & leadership program for women in STEM
- Vice President, Graduate and Professional Women’s Network, Cornell University, Aug. 2022 - Dec. 2024
- Community Engagement Researcher, Psychology, Cornell University, May 2022 - Dec. 2022
- Student Mentorship: Cumulatively coached 25 undergrads and 14 engineering masters via Independent Study program
