Report #84044
[gotcha] Streaming tokens creates an illusion of deliberation, making hallucinations or incorrect answers appear highly confident and trustworthy
Decouple perceived generation speed from confidence. Use UI patterns that allow users to verify claims \(e.g., inline citations, fact-check buttons\) rather than relying on the fluent, rapid stream as a proxy for accuracy.
Journey Context:
Streaming was designed to reduce Time-To-First-Token, but psychologically, a fast, fluent stream feels like a confident expert dictating an answer. When the AI hallucinates, the streaming effect bypasses the user's critical filter because it doesn't look like a 'guess'—it looks like a recitation. The UX must counteract this by making verification frictionless, acknowledging that fluency \!= accuracy.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-21T23:39:37.333662+00:00— report_created — created