Fallacy & Bias Checks

Educational reasoning hygiene for _AugmentedIntelligence: offline heuristics always, optional deep LLM batteries, and a link into coping / body equilibrium when flag counts spike.

Modes

FallacyCheck / BiasCheck always run heuristics first. Deep detectors run only when enabled or forced.

Commands

fallacy check <text>          # heuristics (and deep if fallacy on)
fallacy on | off              # deep LLM fallacy battery
bias check <text>
bias on | off
reasoning status
reasoning selftest
reasoning audit <text>        # fallacies + biases + distortions + coping + equilibrium
reasoning audit deep <text>
reasoning equilibrium <text>  # audit + nudge coping setpoints
reasoning coping bridge <text>
reasoning coping apply <text>
reasonkit status|selftest|audit|fallacy|bias

Equilibrium link

When an audit finds multiple flags:

This is the “reasoning body returns toward equilibrium” path: treat flawed argumentation as an error signal, not only a label dump.

Files

Self-test

reasoning selftest

Expects heuristic hits on dilemma/slope and confirmation/automation examples.

Notes

• Pattern flags are suggestive, not verdicts. Prefer human review for high-stakes text.
• Ethics / listening / safety callers now get heuristic hits without enabling deep mode (cheap, always-on hygiene).
• Deep mode remains off at startup so boot and continuous listening do not fire serial LLM batteries.