Context Stability
Deterministic Slices: Designing Questions AI Can Ask Reliably
Deterministic slices constrain questions so answers stay consistent and explainable.
Rendering diagram...
Question templates apply stable filters to produce deterministic answers.
TL;DR
- • Design questions around stable dimensions.
- • Avoid ambiguous or overlapping filters.
The problem (layman)
- • AI can ask questions with ambiguous or unstable filters.
- • Results become hard to reproduce.
Why it matters
- • Deterministic questions build user trust.
- • They enable repeatable evaluation.
Symptoms
- • Same question yields different results across sessions.
- • AI produces inconsistent explanations.
Root causes
- • Unclear filter paths.
- • Dimensions with overlapping meanings.
What good looks like
- • Defined set of reliable question patterns.
- • Stable dimensions with explicit roles.
How to fix (steps)
- • Define a list of approved question templates.
- • Test them against the model.
- • Publish these templates to users.
Pitfalls
- • Allowing free‑form questions without guardrails.
- • Ignoring role‑playing dimensions.
Checklist
- • Question templates defined.
- • Templates tested for stability.
- • User guidance published.