A Dyadic Active Inference Model of Conflict Conversations: An Exploratory Study on 16 Couples

We describe an active inference formulation of romantic couples during conflict conversations. We model partners' hidden states and the resulting observable sound and motion behaviors and their evolution in time. Agent models are equipped with an internal model of their own state and their partner's hidden state, i.e., a basic version of emotional theory of mind. This internal model enables the agent models to predict and subsequently act towards future behavioral observations, giving rise to patterns such as turn-taking or demand–withdraw-like interactions. We show that several, but not all model parameters can be recovered from simulated experimental data. Fitted parameter profiles from mixed-gender 16 couples suggest four couple-level clusters. These clusters differ in joint behavioral preferences and hidden state transition dynamics, exhibiting patterns consistent with female-demand & male-withdraw, male-demand & female-withdraw, and balanced behavioral engagement. We conclude that active inference is a promising modeling framework for interpersonal processes, even though parameter identifiability remains challenging, which limits interpretability.

Authors

Publication Details

Journal
TAM DataHub
Published
2026-11-19
DOI
https://doi.org/10.60834/tam-datahub-48
Primary Topic
Embodied and Extended Cognition
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

A Dyadic Active Inference Model of Conflict Conversations: An Exploratory Study on 16 Couples

Dominik Endres, Koosha, A. koosha, A. koosha, Tahmineh, Tahmineh
TAM DataHub
Embodied and Extended Cognition
preprint

A Dyadic Active Inference Model of Conflict Conversations: An Exploratory Study on 16 Couples

Dominik Endres, Koosha, A. koosha, A. koosha, Tahmineh, Tahmineh
preprint en

Abstract

We describe an active inference formulation of romantic couples during conflict conversations. We model partners' hidden states and the resulting observable sound and motion behaviors and their evolution in time. Agent models are equipped with an internal model of their own state and their partner's hidden state, i.e., a basic version of emotional theory of mind. This internal model enables the agent models to predict and subsequently act towards future behavioral observations, giving rise to patterns such as turn-taking or demand–withdraw-like interactions. We show that several, but not all model parameters can be recovered from simulated experimental data. Fitted parameter profiles from mixed-gender 16 couples suggest four couple-level clusters. These clusters differ in joint behavioral preferences and hidden state transition dynamics, exhibiting patterns consistent with female-demand & male-withdraw, male-demand & female-withdraw, and balanced behavioral engagement. We conclude that active inference is a promising modeling framework for interpersonal processes, even though parameter identifiability remains challenging, which limits interpretability.

TAM DataHub
Embodied and Extended Cognition
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon