Emotion in an Active Inference Model of Human Driving
A recent submission on arXiv (2608.07480) introduces a broadened framework for valence and arousal within active inference models related to human driving. Active inference serves as a method for simulating adaptive behavior, merging goal-oriented actions with the minimization of uncertainty, and has been utilized in driving scenarios. Previous models overlooked the impact of emotional states, which play a crucial role in decision-making on the road. Earlier studies in non-traffic contexts depicted emotions using the circumplex model along valence and arousal dimensions but were confined to basic discrete state spaces. This new research enhances that approach by deriving affective estimates from a sophisticated active inference model featuring continuous states, taking into account the current state and additional variables. The authors, whose names are not mentioned in the abstract, aim to fill a void in the representation of human driving behavior by incorporating emotional aspects, potentially enhancing the authenticity of driver models in autonomous systems and traffic simulations.
Key facts
- Paper on arXiv with ID 2608.07480
- Proposes expanded formulation of valence and arousal for active inference driving model
- Active inference balances goal-directed action with uncertainty reduction
- Previous active inference driving models did not address affective state
- Prior emotion modeling in active inference used circumplex model with valence and arousal
- Previous work limited to discrete state spaces
- New model uses continuous states
- Affective estimates conditioned on current state
Entities
Institutions
- arXiv