Prediction or Causation? Towards a Redefinition of Task Controllability

Area of Science:

• Cognitive Science
• Artificial Intelligence
• Behavioral Psychology

Background:

• Research on controllability perception is increasing.
• Current definitions may confuse controllability with predictability.
• Cognitive tasks require precise definitions for accurate study.

Purpose of the Study:

• To propose an updated algorithmic definition of controllability.
• To differentiate controllability from predictability in agent environments.
• To guide the design of cognitive tasks investigating control perception.

Main Methods:

• Conceptual analysis of existing definitions.
• Reframing controllability through causal influence on state transitions.
• Distinguishing agent agency from environmental predictability.

Main Results:

• Controllability is redefined as an agent's capacity to exert causal influence over environmental state changes.
• This reframing separates the concept of control from mere prediction of outcomes.
• The proposed definition aids in designing experiments that isolate the perception of control.

Conclusions:

• An updated definition of controllability, focusing on causal influence, is crucial for advancing research.
• This distinction prevents confounding factors in cognitive tasks related to agency and decision-making.
• The revised framework supports more rigorous investigation into how agents perceive and interact with controllable environments.