GroupAffect-4: A Multimodal Dataset of Four-Person Collaborative Interaction

2026-05-19 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, quick

Summary

GroupAffect-4 is a new multimodal dataset designed to analyze affect in co-located four-person groups, addressing the fragmentation in existing affective computing and social signal processing corpora. It comprises data from 40 participants organized into 10 four-person groups, each completing four distinct collaborative tasks: information pooling, negotiation, idea generation, and a public-goods game. Participants were instrumented with wrist-worn physiology sensors, eye-tracking glasses, and close-talk microphones. The dataset integrates continuous affect self-reports, post-task questionnaires, task outcomes, and Big-Five personality scores, all time-aligned. It boasts high data quality, covering over 91% of expected physiology windows and 98% of eye-tracking windows, with a confirmed affective manipulation check. GroupAffect-4 defines fifteen benchmarkable targets across within-person state, between-person traits, and group dynamics, and includes leave-one-group-out feasibility baselines. The dataset is publicly available with a BIDS-inspired structure, Croissant metadata, and open processing scripts.

Key takeaway

For research scientists investigating human interaction or group affect, GroupAffect-4 offers an unprecedented multimodal resource to overcome data fragmentation. You can utilize its time-aligned physiological, eye-tracking, audio, and self-report data from four-person groups to develop and benchmark models for within-person states, between-person traits, and group dynamics. This dataset enables more holistic and ecologically valid studies of collaborative behavior.

Key insights

The GroupAffect-4 dataset integrates diverse multimodal data to enable comprehensive analysis of affect in co-located four-person group interactions.

Principles

• Multimodal data captures complex group dynamics.
• Time-alignment is crucial for integrated analysis.
• Ecologically varied tasks enhance dataset utility.

Method

Participants in 4-person groups complete varied tasks while instrumented with physiology sensors, eye-trackers, and microphones, with data time-aligned to self-reports and personality scores.

In practice

• Analyze within-person states and group dynamics.
• Benchmark models on 15 defined targets.
• Utilize open processing scripts for data.

Topics

• GroupAffect-4
• Multimodal Datasets
• Affective Computing
• Group Dynamics
• Eye Tracking
• Physiological Sensing

Code references

• meisamjam/GroupAffect-4

Best for: AI Scientist, Research Scientist

Related on AIssential

• MIND: multimodal integration with neighbourhood-aware distributions — MIND integrates incomplete multimodal data using a VAE with neighbourhood-aware priors, improving robustness and performance in tasks like cancer stratification.
• VideoFDB: Evaluating Full-Duplex Vision-Speech Capabilities in Conversational Agents — VideoFDB is the first benchmark to evaluate full-duplex audio-visual conversational agents, revealing current systems' limitations in nonverbal grounding.
• Data-Driven Decoding of Russell's Circumplex Model of Affect — Transformer embeddings, particularly multimodal fusion, intrinsically encode Russell's circumplex model of affect, bridging psychology and representation learning.
• Frozen Multimodal Embeddings for Personality and Cognitive Ability Assessment in Asynchronous Video Interviews — Frozen multimodal embeddings effectively assess personality and cognitive ability from asynchronous video interviews, outperforming baselines.
• BEHAVE: A Hybrid AI Framework for Real-Time Modeling of Collective Human Dynamics — Collective human dynamics can be modeled as complex dynamical systems using behavioral fields derived from physical micro-signals.
• Cohesion-6K: An Arabic Dataset for Analyzing Social Cohesion and Conflict in Online Discourse — Divisive online discourse consistently attracts disproportionately higher user engagement compared to unifying narratives.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.