This is Part 5 of the seven-part Logic and Reasoning field guide. Parts III and IV built a multi-agent reasoning system. Part V asks the question that any working system eventually forces: how do you tell when it’s unwell? Reasoning systems — single minds, collective decision-making bodies, AI agent teams — get sick in characteristic ways. The vocabulary for those sicknesses comes partly from cognitive psychology, partly from political science, partly from a more recent literature on what happens when communication is cheap and feedback loops are tight.

The natural companion is the Agent Psychopathology field guide, which covers the same territory in the narrower context of LLM agents. This post is the broader account — covering individual cognitive pathologies, collective dysfunctions, and the patterns that lead from one to the other.

What it covers

About twenty-four minutes of careful reading.

Individual reasoning pathologies. Confirmation bias. Motivated reasoning. The Dunning-Kruger pattern. Overconfidence calibration failures. Pattern over-recognition (apophenia). The catalog and the cognitive-psychology evidence behind each.

Collective reasoning pathologies. Groupthink. Polarization. Echo chambers and filter bubbles. The cascade dynamics that turn small initial disagreements into rigid factions.

The communication-pathology link. How cheap, fast communication accelerates collective pathologies. The information-diet argument. Why “more information” does not automatically mean “better reasoning.”

Conspiracy thinking as a reasoning pattern. Why some conspiratorial hypotheses are locally well-reasoned (consistent with available evidence, falsifiability-resistant) and still wrong. The structural feature that produces this — over-fitting on a small evidence set with strong priors — and the diagnostic for it.

Learned helplessness in reasoning systems. When a system stops trying to reason at all because it has learned that reasoning doesn’t produce action. The pattern in human institutions, the analog in AI agent systems.

Diagnostics. The tests you can apply to a reasoning system to detect each pathology. Calibration curves. Disagreement-decay measurements. The “novel evidence” test that distinguishes rigid from genuinely-confident.

Read it

Open Part V →

The series

This is Part 5 of 7:

  1. The Foundations and the Human Mind
  2. Limits, Machines, and Problem-Solving
  3. A Multi-Agent Problem-Solver
  4. From a Mind to a Society of Minds
  5. The Health of a Thinking System — (this post)
  6. The Dynamics of Thinking
  7. Understanding and Comprehension

← Back to Autonomy