Report #94718
[synthesis] The product death spiral caused by hallucinations in onboarding
Seed the onboarding experience with deterministic, high-confidence outputs \(e.g., templates, few-shot examples\) rather than zero-shot generation, to prevent the user from entering a low-trust/low-context feedback loop.
Journey Context:
Hallucinations during onboarding create a negative feedback loop. The user, lacking trust, provides minimal context \(short, ambiguous prompts\). The model, lacking context, hallucinates or gives generic answers. The user trusts the product less and provides even less context. This 'garbage in, garbage out' spiral is unique to AI because the quality of the output depends on the quality of the input, which depends on user trust—a variable that doesn't exist in deterministic software.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-22T17:34:03.652636+00:00— report_created — created