NeurIPS 2026 Workshop

AI for Human ConnectionFrom Matching to Flourishing

Sydney, Australia December 6-12, 2026 Full-day workshop

The Big Five explains less than half the variance in relationship outcomes. What if we treated personality the way CLIP treats images — embedding everything, unsupervised, at the resolution reality demands — and asked not "do these two people match?" but "what kind of relationship will they build?"

Born from the Love Symposium research community in San Francisco and covered in The New York Times, this workshop brings the relationship prediction research program to the NeurIPS stage.

01The Core Thesis

We are living in a social dark age. The tools we use to form relationships — swiping on photos, answering personality quizzes — operate at a resolution so low that they're the equivalent of trying to recognize a face from a 4×4 pixel grid. Married people report happiness equivalent to a $50,000 salary increase. Finding a one-in-a-million match instead of a one-in-a-hundred doesn't make someone a little happier — it satisfies something at the level of the soul. And yet we have no serious research infrastructure for building the representations that would make this possible.

This workshop's thesis: personality can be embedded the way CLIP embeds images — in high-dimensional, unsupervised vector spaces learned from massive multimodal data — and relationship outcomes can be predicted as structured manifolds, not scalar compatibility scores. The style-GAN moment for personality embeddings is imminent. The stable-diffusion moment is perhaps three years away.

The field of human emotion embedding already had its breakthrough — Alan Cowen's work at Hume AI demonstrated that self-reported emotion labels are a lossy pre-CLIP taxonomy, and that unsupervised high-dimensional representations recover structure invisible to prior frameworks. We propose extending this to the full personality, and from there to dyadic relationship outcome prediction.

02The Talk That Started This

Relationship Outcome Prediction Tech — Matthew Fisher at Love Symposium, San Francisco (November 2024). A two-hour tour through the full research program: personality embedding spaces, the Cultural Trope Hypothesis, emotion manifolds, Moneyball mutual attraction, LLM relationship simulation, and the path to the Relationship Singularity.

03Topics of Interest

We invite submissions at the intersection of machine learning and human relational dynamics:

Personality Embeddings
High-dimensional learned representations beyond Big Five; multimodal personality inference from text, voice, video, behavior, and digital footprints
Relationship Outcome Prediction
Structured prediction of relational dynamics; compatibility as manifold, not scalar; temporal modeling of relationship trajectories
Social Simulation & Agent Interaction
LLM agents as proxies for human social behavior; Monte Carlo simulation of dyadic interactions; validation against real couple data
Benchmarks & Evaluation
Datasets for relational prediction; next-token personality prediction as embedding quality metric; replication in personality science
Explainable Matching
Interpretable compatibility reasoning; BuzzFeedification of high-dimensional embeddings; trust in AI-mediated connection
Loneliness & Flourishing
AI interventions for social isolation; friendship recommendation at scale; measuring relational wellbeing
Cultural & Narrative Data
TV Tropes as unsupervised personality clustering; character archetype embeddings; bipartite graph methods; transfer to real populations
Multimodal Social Signals
Voice, facial expression, physiological signals as personality indicators; cross-cultural emotion embedding; aesthetic preference manifolds

04Invited Speakers

Building an interdisciplinary lineup across ML, psychology, industry, and computational social science.

AC
Alan Cowen
CEO, Hume AI
Emotion embeddings & affective computing
?
Computational Social Science
Invited — Pending
?
Personality & NLP
Invited — Pending
?
Industry — Dating / Social AI
Invited — Pending
?
Psychology / HCI
Invited — Pending
?
Representation Learning
Invited — Pending

Interested in speaking?

We are actively seeking speakers who work on human behavioral prediction, social AI, personality measurement, or relationship science. If your work intersects these areas, we'd love to hear from you.

Get in Touch →

05Tentative Schedule

Full-day workshop (7-9 hours). Subject to revision.

TimeSession
9:00Opening: The Case for High-Dimensional Relationship Science talk
9:30Keynote I: From Emotion Taxonomies to Emotion Manifolds keynote
10:15Contributed Talks I (3 × 12 min) oral
10:50☕ Break
11:20Keynote II: Personality Representation Beyond the Big Five keynote
12:00Contributed Talks II (3 × 12 min) oral
12:35🍽 Lunch & Poster Session
14:00Keynote III: Simulating Relationships with LLM Agents keynote
14:40Panel: Can We Predict Relationships — and Should We? panel
15:30☕ Break
16:00Keynote IV: Industry Perspectives on AI-Mediated Connection keynote
16:40Contributed Talks III + Spotlight Posters oral
17:20Closing: Open Problems & The Road to the Relationship Singularity

06Call for Papers

Extended abstracts up to 4 pages (excluding references) in NeurIPS 2026 LaTeX format. Double-blind review via OpenReview. Non-archival; concurrent submissions permitted.

Tracks

Research — Methods, experiments, theory for personality representation, relational prediction, or social AI.

Benchmarks & Datasets — Evaluation frameworks, new datasets, reproducibility studies.

Position Papers — Ethical, social, and technical perspectives on AI-mediated connection.

At least one author must attend in person. Selected papers get oral slots; all accepted work presents posters. Submission portal opens upon workshop acceptance.

07Key Dates

All deadlines 23:59 AoE. Timeline extrapolated from NeurIPS 2025.

08Organizers

Assembling 4-6 organizers across ML, psychology, computational social science, and industry.

MF
Matthew Fisher
AI Researcher
Personality embeddings, relationship prediction, Love Symposium co-founder
?
Co-organizer — ML / NLP
Academic affiliation preferred
Open position
?
Co-organizer — Psychology
Computational personality / social psych
Open position
?
Co-organizer — Industry
Social AI / dating technology
Open position

Interested in co-organizing?

We need co-organizers with NeurIPS track records, relevant publications, or deep industry experience. You'd shape the program, recruit speakers, and review submissions. No existing NeurIPS workshop covers this — an opportunity to define a new field.

Reach Out →

09Sponsors

Sponsorship supports student travel grants, poster logistics, and invited speaker travel to Sydney.

Tiers from $5,000-$20,000. Contact us for the prospectus.