Artificial Cognition
Artificial cognition is Steve Shaw’s term in Are humans losing the ability to think for themselves? for AI reasoning that enters a human decision process. Shaw frames it as “system three”: a third layer beside fast intuitive thinking and slower deliberative thinking.
The concept is useful because it separates external model reasoning from the user’s own cognition. ChatGPT can supply an answer, explanation, or recommendation, but once that output appears inside a choice, the human decision is no longer only a fast-versus-slow psychology problem. It also includes whether the user treats AI output as evidence, suggestion, shortcut, or authority.
Artificial cognition is not automatically bad. The Marketplace Tech episode says Shaw uses AI daily and allows students to use it. The risk appears when artificial cognition becomes the first or only reasoning path, producing Cognitive Surrender instead of an inspected aid to Human Judgment Under AI.
Key Claims
- AI adds an external cognitive layer to decision-making rather than only speeding up search or drafting.
- “System three” is source terminology for that AI-mediated layer, not a general claim that AI has humanlike consciousness.
- Artificial cognition can help structure work when the user has already formed a goal, hypothesis, or draft.
- It can undermine judgment when users accept AI conclusions without enough independent reasoning or review.
- In learning contexts, artificial cognition should usually follow First Draft Thinking rather than replace the learner’s first attempt.
- In autonomous workflows, artificial cognition needs review gates because AI may act over tasks before a person has challenged its reasoning.
Connections
- Cognitive Surrender - main risk when artificial cognition becomes authority.
- Steve Shaw and Wharton School - source speaker and research setting.
- ChatGPT - example system in the Marketplace Tech source.
- Human Judgment Under AI - boundary for using artificial cognition without outsourcing responsibility.
- AI As Tutor, First Draft Thinking, and AI Shortcut Risk - education and learning implications.
- Agentic Workflow - workflow context where AI reasoning can be paired with tool action.