Mahault Albarracin is a researcher in cognitive computing and computational cognitive science working at the intersection of artificial intelligence, social cognition, and philosophy of mind. Her research develops formal models of agency, social coordination, and collective intelligence grounded in active inference and Bayesian approaches to belief dynamics.

She is an Associate Researcher with the Laboratoire d’Analyse Cognitive de l’Information (LANCI) at Université du Québec à Montréal and a Visiting Researcher in Social Robotics at Sheffield Hallam University. Her work combines formal modelling, simulation, and embodied robotics to investigate theory of mind, empathy, epistemic alignment, and the dynamics of shared intentionality in artificial and biological systems.

Dr. Albarracin’s broader interests include the epistemology of machine intelligence, the structure of social norms and scripts, and the role of predictive processing frameworks in explaining affective and interpersonal phenomena. She regularly lectures on artificial intelligence, cognition, and society, and her publications span computational modelling, philosophy of cognitive science, and social systems theory.

Mahault Albarracín’s program offers unusually direct conceptual and computational bridges between (i) scripts as socially shared normative patterns, (ii) affordance landscapes as action possibilities structured by cultural practices, and (iii) active inference as a modelling framework in which action and perception are forms of inference over a generative model of the world. Her work formalises scripts as both internal schemata and external social orders (dual-aspect scripts), enabling a principled account of “material minds” in which environments carry actionable norms rather than merely “context.”  Norms become externalised into material environments via social activity, and are “brought about” in ongoing engagement rather than located strictly inside agents or strictly “out there.” 

The key design opportunity for HCI/robotics is to treat artefacts and spaces as normative interfaces: they do not only present physical constraints, but also cue deontic expectations (“what one should do here”). This suggests a concrete robotics agenda: build robots that can (1) perceive artefacts and spatial configurations as deontic cues, (2) infer which social script is currently in play, and (3) choose actions that are simultaneously task-effective and norm-compliant, while remaining sensitive to pluralism, ambiguity, and contestability of norms. This aligns naturally with active-inference notions of deontic value (policy priors elicited by cues), which already formalise how observable cues can steer agent policy selection toward culturally expected behaviour. 

A necessary second half of the story, especially salient for “material minds”, is the inverse direction: robots are themselves artefacts that can reshape human routines, attention, and social meaning. Longitudinal and ethnographic studies of domestic robots (notably robot vacuums) show households actively reorganising cleaning practices, spatial arrangements, and affective relationships around the device.  This motivates a “reflexive externalism”: robots should be designed not only to read scripts but also to publish scripts responsibly, via transparent cues, user-contestable norms, and careful limits on persuasive or manipulative influence.