Proceedings of the Annual Meeting of the Cognitive Science Society

Existing personality models describe stable individual differences well, but often leave underspecified how such differences generate moment-to-moment expressive behaviour. Acting theory faces a complementary problem: performers must reliably externalise a character's internal state so that intentions, affect, and personality can be inferred from movement, speech, and posture alone. We address this gap by introducing a computational formalisation of the Laban–Malmgren System (LMS), a framework from movement analysis and acting pedagogy that links internal disposition to observable action. Building on Laban's analysis of movement qualities and Malmgren's extension for character work, we formalise four core internal and external dimensions of behaviour and develop a generative cognitive architecture in which: 1) stable trait parameters define priors over a latent expressive state, 2) state transitions are shaped by context and actions, and 3) a stochastic policy maps the latent state to observable behaviour. The resulting model provides a principled, action-centred link between internal dispositions, state dynamics, and expressive behaviour, making the theory accessible to simulation, quantitative analysis, and embodied artificial agents. We demonstrate its computational feasibility through a public proof-of-concept implementation, Persona: an interactive digital portrait in which an avatar conveys its internal state through detailed bodily expression and responds dynamically to viewer interaction (Saatchi Gallery, London, finalist for the 2024 Lumen Prize). By formalising a theory rooted in expert artistic practice, this work offers a complementary, action-centred perspective on personality modelling and lays the groundwork for future empirical validation.