Exploring the Dynamics of Social Touch
TMEET is a research initiative exploring how mediated digital touch can foster perceptual and emotional engagement. It observes how two people adjust to, constrain, and respond to each other through distal tactile contact across collaboration, competition, conflict, and exchange modes.
Relational modeling framework
Each mode is designed to isolate a distinct interactional structure and is then expressed through a series of minimalist virtual games that reveal its relational dynamics.
Two users must work together to achieve a shared goal. Success requires mutual cooperation and coordination.
At least two participants who must coordinate their actions to achieve a common objective that neither can accomplish alone.
Two users strive for the same goal, but only one can succeed. The presence of the other may hinder individual progress but does not block it.
At least two participants pursuing the same goal where only one (or a limited number) can achieve success, creating rivalry.
The primary goal is to prevent the other from winning. Only one user can reach the target at a time.
At least two participants with opposing objectives where one's success directly requires preventing the other's success.
Two users engage in sequential reciprocity.
At least two participants each engage in an alternating sequence of action and response, with every contribution eliciting a contingent reply and roles reversing over time.
Research scope
Our paired-motion models will use the four relational tasks described above to simulate how two people coordinate, compete, resist, and respond through mediated touch.
All four tasks take place in a shared simulated minimalist world, where each participant controls a cursor and encounters the other through mediated touch.
Moving Hill
The dyad works together to keep both cursors on a moving hill.
AB
Sound-Guided Agent
Both players pursue the same invisible, sound-guided agent.
AB
Moving Flag
The dyad competes to seize and hold the same moving flag.
AB
Partner Identification
One player identifies the human partner among four responsive agents through movement alone.
AB
Research foundations
Grounded in enactive and embodied accounts of participatory sense-making, TMEET treats touch as an ongoing sensorimotor coordination process: meaning emerges through the reciprocal timing, resistance, and adaptation of action–perception coupling across the dyad.
To understand social touch, it is important to understand which interpersonal and tactile dynamics carry emotional value—and how they emerge through movement, contact, reciprocal adjustment, physiological variation, and reported experience.
Our Method
Track approach, contact, maintenance, rupture, resumption, synchrony, resistance, and avoidance.
Synchronize trajectories and contact episodes with Empatica measures and individual self-reports.
Recognize strategies, predict transitions, and guide tactile responses suited to the ongoing relation.
Pseudo-haptic conditions
Pseudo-haptics create a sense of touch or force without specialized haptic hardware, using visual, auditory, and interaction-based illusions. TMEET uses these effects to study how perceived texture, resistance, attraction, and movement can make digital contact more immersive and emotionally expressive. Bump, Hole, and Flat vary the control–effect relationship to reveal how these sensations reshape approach, contact, and coordination.
Estimating intent and effort. Pseudo-haptic gain changes the relationship between physical mouse movement and the cursor displacement shown on screen—the control–display ratio. Comparing input displacement with the visual result helps estimate motor effort and movement intent, even when the displayed motion is amplified or constrained. This discrepancy provides the AI model with an additional behavioral signal.
Resistance
Pseudo-haptic resistance changes the control–effect relationship as participants approach or remain in contact.
Attraction
Pseudo-haptic attraction or assistance draws control toward the center of contact and changes how coordination is restored.
Control
The control condition retains contact feedback without pseudo-haptic modulation of the movement relationship.
Dataset and learning
The study synchronizes three complementary data streams—dyadic movement and contact, physiological activity, and participant reports—on one timeline. The model then learns how patterns across these signals correspond to interaction strategies and likely next responses.
Approach, contact, maintenance, separation, and recontact become time-bounded episodes.
Specialized branches describe dyadic movement, physiological change, and expressed experience.
The relational mode conditions the model so the same contact can mean different things.
Recognize the current strategy, anticipate the partner’s response, and guide a contingent touch.
Multimodal AI pipeline
Each recorded trial becomes a time-aligned example that links what the dyad did, how their bodies responded, and how the interaction was experienced.
Cursor trajectories, distance, contact events, and relational task
EDA, cardiac activity, movement, and skin temperature
Valence, emotion, intention, and perceived control
Conditioned by interaction mode
One dyad × one trial × one interaction episode, with all three data streams aligned to the same timestamps.
Long-term direction
The protocol aims to uncover regularities that help future systems interpret, predict, and generate social touch with context-sensitive emotional value.
Explore the protocol resourcesResources
Video, publications, project writing, and the social touch model diagram will be added as the research develops.
VIDEO PRESENTATION
ACADEMIC PAPERS
RESEARCH BLOG
MODEL DIAGRAM