A human example
A player notices a window while an optional waiting remark is queued.
What the caller supplies
The game supplies equally prioritized eligible recordings, the real trigger for each and recently completed speech.
What happens next
Jev recommends at most one cue or silence. The game rejects changed contexts and owns playback.
Illustrative integration example. The measured first-run choices are documented separately.
Potential value: medium
Could improve the timing and variety of authored reactions without generating new dialogue.
Evidence confidence: low
The optional game integration exists. Its first eight-case diagnostic chose silence throughout; a randomized listening comparison and positive-selection evidence are still missing.
The rating describes support for this claim. It is separate from Jev's returned probability. How we assign ratings.
Evidence, including disagreement
- Define ambiguity before evaluating a workflow. reference. Specify what uncertain and mixed cases should do.
- Parable dialogue advice chose silence in every first-run case. first-party. Conservative silence is not evidence of better dialogue.
The next test
20 scripted playthrough paths plus human review, covering return visits, waiting, interruptions and scene changes.
Compare against
- Current deterministic director
- Optional Jev advice
- Always silence for optional remarks
Measure
- Repeated lines
- Interruptions
- Stale recommendations rejected
- Player preference
- Added delays
Decision after the test
Keep narration optional. Require no repeat or stale-playback regression and a clear preference in reviewed playthroughs.
The report will retain inputs, question versions, every attempt and failure examples. We will update the confidence rating after reviewing the result.
Use a related Engine recipe
Recipes are implementation starting points. Their presence does not mean the protocol above has passed.