mindits

Mindits are discrete representations that encode aspects of an agent’s mental state, such as beliefs, intentions, or preferences. They function as modular units within cognitive architectures, enabling reasoning, planning, and action selection. The term mindit blends mind and unit to reflect their role as compact, portable representations of cognitive content.

A mindit typically consists of content (the propositional or evaluative information), provenance (where it originated), and a status (active or inactive, with an associated confidence). Types commonly discussed include belief mindits, goal mindits, and intention mindits. Mindits can be queried by the system to inspect current beliefs, desires, and commitments, and they may be updated through perception, reflection, or interaction with other agents.

In artificial agents, mindits support explainable AI by exposing internal state in human-readable form. They also aid modular planning: a planner can select actions that align with the active mindits or revise them as new information arrives. In cognitive science, mindits serve as an abstraction for modeling how humans hold and revise beliefs, desires, and intentions within a dynamic environment.

The concept is not standardized and appears across a range of theoretical discussions rather than a single unified framework. Critics warn that overly coarse or underdefined mindits can fail to capture the nuance of real cognition and risk conflating beliefs, desires, and intentions.

Related concepts include mental representation, belief states, intentions, goals, cognitive architectures, and explainable AI.