---
name: yogsoth-ai/competing-hypothesis-construction
source: https://app.decimal.ai/s/yogsoth-ai-competing-hypothesis-construction@1/SKILL.md
source_sha256: d6b7ef17dd90
---

# Competing Hypothesis Construction

Construct multiple competing hypotheses for the same phenomenon: actively counter confirmation bias, maintain epistemic openness by building genuinely different explanations in parallel, and design decisive predictions that can distinguish them.

## When to Use

- The researcher already has a "preferred hypothesis" and needs to actively challenge it
- The phenomenon has multiple plausible explanations, and premature convergence would lead in the wrong direction
- You need to show reviewers or funders that alternative explanations have been considered
- When designing experiments, you need to determine which variable best discriminates between competing explanations

Not applicable: the phenomenon has overwhelming evidence supporting a single explanation → directly use deductive-hypothesis-generation to refine that explanation.

## Thinking Framework

**Avoid confirmation bias by generating genuinely different explanations, then find discriminating predictions**

The core logic of competing hypothesis construction:

1. **Force diversity**: competing hypotheses must be genuinely different at the mechanism level, not variants of the same mechanism
2. **Symmetric treatment**: each hypothesis is treated with equal seriousness; the preferred hypothesis is not allowed special treatment
3. **Discriminating predictions**: find decisive predictions that distinguish the hypotheses — what result supports H1 but contradicts H2, and vice versa
4. **Matrix comparison**: reveal structural differences between hypotheses through a systematic matrix

**Quality criteria for competing hypotheses**:
- **Genuinely competing**: two hypotheses give different causal explanations for the same phenomenon (not weaker/stronger versions of the same explanation)
- **Mutual exclusivity**: there exists at least one observable result that can support one while contradicting the other
- **Comparability**: both hypotheses have clear testable predictions

## Budget Gate

| Tier | Competing hypotheses | Discriminating predictions | Comparison matrix | Falsifiability |
|------|---------|---------|---------|---------|
| S | ≥2 genuinely competing hypotheses | ≥1 discriminating prediction | simplified version (2×2) | 1 falsification scenario per hypothesis |
| M | ≥3 competing hypotheses | ≥2 discriminating predictions | full matrix (hypotheses × predictions) | full falsification per hypothesis |
| L | ≥4 competing hypotheses | ≥3 discriminating predictions | full matrix + experiment design suggestions | full falsifiability audit |

## Default Reference Flow

1. Call the `competing-hypothesis-generation` SOP (via the `competing-hypothesis-matrix` tactic): force generation of hypotheses that are genuinely different at the mechanism level
2. Call the `discriminating-prediction-design` SOP: design discriminating predictions for each pair of competing hypotheses
3. Call the `hypothesis-comparison-matrix` SOP: build a hypotheses × predictions comparison matrix to reveal structural differences
4. Call the `falsifiability-check` SOP: generate a falsification scenario for each hypothesis

## context-checkpoint

Record after each round:
- Competing hypothesis list (core mechanism statement for each hypothesis)
- Mechanism difference analysis (at what level the hypotheses are genuinely different)
- Discriminating prediction list (which pair of hypotheses each prediction can distinguish)
- Comparison matrix (hypotheses × predictions, annotated support/contradict/neutral)
- Recommended decisive experiment directions

<!-- BEGIN available-tables (generated) -->

## Available Tactics

Optional, no fixed order; the final leaf is always a sop.

| Tactic | When to use |
| --- | --- |
| competing-hypothesis-matrix | Tactic: Multi-hypothesis management — generate competing hypotheses, design discriminating predictions, build a structured comparison matrix |

## Available SOPs

Optional, no fixed order; the final leaf is always a sop.

| SOP | When to use |
| --- | --- |
| falsifiability-check | SOP: check whether a hypothesis meets the falsifiability criterion |

<!-- END available-tables (generated) -->