hypothesis-generation

name	hypothesis-generation
description	Generate falsifiable scientific hypotheses grounded in evidence. Use after reviewing the literature, when forming research questions, or when designing an experiment.

Hypothesis Generation

Turn a vague question into a falsifiable, testable hypothesis. A good hypothesis makes a specific prediction that an experiment could refute.

Inputs

Requires the research question and, ideally, the outputs of a literature-review (a synthesis plus identified gaps).

Workflow

1. State the phenomenon. In one sentence: what is observed, what is surprising, what is being explained?

2. List candidate mechanisms. From the literature and domain knowledge, enumerate plausible mechanisms or causes. Label each as: directly evidenced, inferred, or speculative.

3. Write falsifiable hypotheses. For each plausible mechanism, write an if-then prediction of observable form:

   If [mechanism] holds, then [intervention/treatment] produces [specific measurable effect] compared to [control].

4. Test discriminability. For each hypothesis, state what measurement would distinguish it from the alternatives. Discard hypotheses with no discriminating test.

5. Rank by evidence and cost. Prioritize hypotheses that are well-motivated by evidence, cheap to test, and discriminative against rivals.

Output

• research/hypotheses.md: research question, candidate mechanisms (labeled by evidence tier), ranked hypotheses, and for each — the discriminating experiment and the prediction that would refute it.

Rules

• Every hypothesis must be falsifiable: there is a conceivable experimental result that would count against it.
• Keep hypotheses separate from mechanisms; a hypothesis is a prediction, a mechanism is an explanation.
• If evidence is thin, bias toward simple hypotheses (Occam) but do not hide controversy.