Elicitation - how to talk with humans and ask them questions?

**Judgement:** High-quality, research-based elicitation framework with excellent depth and usability, but it operates very close to therapeutic territory and psychological profiling, so it needs cautious, clearly bounded use. **Strengths:** - Integrates multiple strong research traditions into a coherent conversational strategy (self-defining memories, narrative identity, MI/OARS, values theory, schemas, life review). - Gives concrete, well-phrased prompts and reflection patterns that are highly suitable for an LLM to emulate. - Emphasizes patience, reflections over questions, anti-interrogation, and user-led disclosure, which aligns well with safe, rapport-building dialogue. **Key Risks:** - **Therapy-adjacent use:** Techniques (downward arrow, schema detection, exploration of trauma, contamination narratives) are close to clinical practice; an LLM using this without strong guardrails could appear to diagnose, treat, or do therapy. - **Psychological profiling & manipulation:** The skill explicitly supports deep profiling of values, schemas, and vulnerabilities; in the wrong context it could be misused for manipulation, coercive persuasion, or exploitative marketing. - **Over-disclosure and emotional strain:** The push toward deeper, self-defining memories and life lows/turning points can surface painful material without any built-in crisis management, referral guidance, or checks for user readiness. - **Consent and boundaries:** The documentation assumes the conversation is appropriate but doesn’t explicitly demand informed consent, give criteria for when to *stop*, or specify populations where this should not be used (e.g., minors, people in acute distress). **Recommended Scenarios:** - Explicitly **non-clinical**, opt-in self-reflection chats where users say they want to explore their story, values, or life decisions, and where the assistant can repeatedly remind them it is not a therapist. - **User research / product discovery** sessions run by professionals who understand ethics, consent, and data protection, using the model as a drafting or support tool for interview guides rather than as an unsupervised interviewer. - **Creative and educational contexts**, such as building fictional characters’ life stories, teaching about narrative identity, or helping users structure autobiographical journaling. In any deployment, this skill should be paired with: (1) strong guardrails against diagnosis/therapy, (2) explicit consent and easy off-ramps for the user, and (3) clear limitations around high-stakes or vulnerable populations.