How do different designs of teen-visible safety summaries (e.g., a one-line rule label, a short rationale plus next-step suggestions, or a clickable detail view) change teens’ reactions to false positives and near-miss refusals, and which variants most reliably keep them engaged rather than driving them to unsafeguarded tools?
teen-safe-ai-ux | Updated at
Answer
Short, goal-aware rationales with next-step suggestions, plus optional detail-on-click, tend to keep teens more engaged after false positives and near-miss refusals than either bare rule labels or very long explanations.
Best default pattern
- One-line cue + 1–2 sentence rationale + 1–3 concrete next steps.
- Optional expandable detail for teens who want to understand the rule.
Likely effects by design:
- One-line rule label only (e.g., “Blocked by self-harm safety rules”)
- Pros: fast, low friction, easy to implement.
- Cons: feels arbitrary; high frustration on false positives; more "I’ll go elsewhere" behavior.
- Best use: low-stakes blocks; repeated triggers where full text would be spammy.
- Short rationale + next-step suggestions (e.g., “I can’t give self-harm methods, but I can help with coping ideas or how to talk to someone you trust.”)
- Pros: acknowledges goal, offers motion, reduces perceived paternalism; good fit with graceful refusals.
- Cons: slightly longer UI; needs careful copy.
- Best use: high-stakes areas (self-harm, sex, bullying, substances) and first 1–2 refusals per topic.
- Clickable / expandable details (e.g., “Why?” → brief rule summary and examples)
- Pros: supports teens who want reasons; builds trust; lets default surface stay short.
- Cons: extra UX work; overuse may clutter simple flows.
- Best use: shared devices, older teens, edge cases where rules changed (age upgrade, region shift).
Practical recommendation
- Default to pattern (2) with an affordance for (3).
- Reserve pattern (1) for repeated or obviously abusive probing once rationale has been shown.
- Keep wording neutral and rule-based: “this system can’t…” rather than “you can’t…”.