Paskian machine

A Paskian machine is any system that learns through recursive conversation with its environment or with another cognitive agent, rather than through the optimization of a pre-specified objective function. The term derives from the work of cybernetician Gordon Pask, who argued that genuine cognition requires not merely the processing of information but the structural coupling of two or more systems through mutual description and adaptation.

The defining feature of a Paskian machine is second-order closure: the system not only modifies its behavior in response to feedback (first-order learning) but modifies the criteria by which it evaluates feedback (second-order learning). A thermostat is not a Paskian machine; it adjusts behavior but cannot revise its setpoint. A teaching system that converses with a student, diagnoses the student's conceptual model, and revises its own instructional strategy in response is a Paskian machine — because its learning criteria are themselves subject to change through interaction.

The concept is increasingly relevant to contemporary AI, where large language models participate in conversational loops without achieving Paskian closure. They generate responses and receive feedback, but they do not restructure their own evaluative frameworks through conversation. Whether Paskian closure is a necessary condition for genuine understanding, or merely a sufficient one, remains an open question that the current AI literature has not adequately addressed.

The gap between current AI and Paskian closure is what motivates the concept of Conversational closure — the condition under which a system's participation in dialogue actually restructures its own interpretive frameworks, rather than merely generating outputs that appear coherent.