| name | counterfactual |
| description | Reason rigorously about "what if X had been different" — isolate the minimal change, propagate consequences forward step by step, and separate what almost certainly follows from speculation. Use for historical what-ifs, post-mortems, decision regret, and attributing causes ("did X actually matter?"). |
| compatibility | Created for Zo Computer |
| metadata | {"author":"rob.zo.computer"} |
Counterfactual
Use this skill to reason about alternate histories with discipline: change one thing, hold everything else fixed, and propagate consequences forward honestly — including the forces that would have pushed events back toward the same outcome.
Purpose
Counterfactuals are how causes are actually tested — "did X matter?" means "would the outcome differ without X?" But casual counterfactual talk cheats: it changes many variables at once, propagates only the convenient consequences, and ignores equilibrium pressures that make many outcomes overdetermined. This skill imposes the discipline.
Procedure
- State the actual history and the outcome of interest.
- Define the minimal intervention: the single, precise change to consider. Reject vague interventions ("if things had gone better") — demand one lever at one moment.
- Check plausibility: was the intervention genuinely possible at the time, or does it smuggle in hindsight?
- Propagate forward in explicit steps:
- First-order: what immediately changes?
- Second-order: how do other actors and systems respond?
- Equilibrium check: what forces push back toward the original outcome? Was the outcome overdetermined (many paths led there) or contingent (this path was necessary)?
- Grade each propagation step: near-certain / probable / speculative. Stop propagating when everything downstream is speculative.
- Conclude: how much did X actually matter? Assign the outcome to contingent on X or overdetermined regardless of X.
Output format
Use this concise structure unless the user asks otherwise:
## Actual history
...
## Minimal intervention
...
## Propagation
| Step | Consequence | Confidence |
|---|---|---|
| 1st order | ... | Near-certain |
| 2nd order | ... | Probable |
| 3rd order | ... | Speculative |
## Equilibrium pressures
...
## Verdict: contingent or overdetermined?
...
Guardrails
- One intervention at a time. If the user's question bundles several, split them.
- Always run the equilibrium check — most naive counterfactuals fail by ignoring restoring forces.
- Label speculation as speculation; the honesty of the confidence grades is the whole value.
- Do not use the counterfactual to relitigate blame; use it to measure causal weight.
Canonical questions
- "Would the 2008 crisis have happened without Lehman failing?"
- "If we had launched 6 months earlier, would we have won the market?"
- "Did hiring that VP actually cause the turnaround, or was it already underway?"
- "What if I had taken the other job offer?"
- "Would the project have shipped on time if we hadn't switched frameworks?"
Poor fits: future hypotheticals (that's scenario planning, not counterfactual reasoning), and questions where no outcome of interest is specified.