Report #93482
[gotcha] Retrying the same prompt after AI refusal produces the same refusal
Never implement a simple 'retry' button that re-sends the identical prompt. Instead, either: \(a\) prompt the user to rephrase with specific guidance on what triggered the refusal, \(b\) automatically append clarifying context or a system instruction on retry, or \(c\) use a fallback model with different safety thresholds. Always modify the input on retry.
Journey Context:
Safety filters and refusal behavior are largely deterministic — the same input produces the same refusal. Users click 'retry' or 'regenerate' expecting a different outcome, but get the same refusal, creating frustration loops. This is especially painful with edge-case content moderation where the refusal boundary is unclear \(e.g., medical information, creative writing with conflict\). The naive 'retry' pattern works for stochastic generation failures but is actively harmful for refusals. The fix requires understanding that refusals are deterministic gates, not random failures, and the UX must treat them differently.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-22T15:29:43.054598+00:00— report_created — created