Logic Alchemy: Scientific Reasoning & Causal Validation

Problem

Agents often overstate causal links, estimates, and conclusions, especially when evidence is incomplete. bionic-reasoning targets high-risk reasoning tasks where unsupported confidence can lead to flawed decisions. It focuses on making claims testable instead of rhetorically persuasive.

How It Works

The skill activates for causal judgment, numerical prediction, and hypothesis validation. It uses a two-layer model. The first layer applies scientific method controls: build a hypothesis, update confidence with Bayesian reasoning, and require falsifiable statements such as “if A, then B” or “X causes Y, with average effect Z”. It also checks confirmation bias, correlation-versus-causation, and survivorship bias. The second layer uses physical reasoning: first principles, conservation constraints, and dimensional analysis to bound predictions. When reasoning crosses three steps, it adds a peer-review-style check and escalates risk controls if gaps appear.

Boundaries

It suits engineering analysis, data reasoning, and strategy evaluation, but not open-ended creative work, vague preference discussion, or quick conversational replies. It still needs sufficient facts and constraints. It reduces reasoning bias, but it does not replace experiments, statistical tests, or factual verification.