Behavioral Health Expert for Safe AI Conversations

Mercor; United States; Remote; USD 55,000 - 83,000; Part time

Job summary

Cincinnatus LLC is seeking behavioral health experts to evaluate how AI models respond in sensitive conversations, including relationships, wellbeing, and personal beliefs. This is a part‑time W‑2 role with 20–40 hours per week, placed at a leading AI lab as part of the extended workforce.

Ideal candidates have 3+ years in mental health or related fields and can clearly justify professional judgments while maintaining neutrality and precise written feedback.

Qualifications

• A degree in psychology, counseling, social work, behavioral health, or related field, or equivalent experience.
• Minimum 3 years of professional mental health or related field experience.
• Ability to stay neutral and nonjudgmental across diverse perspectives and explain reasoning.
• Strong written communication skills with precise, well-structured feedback.
• Willingness to commit 20 hours/week.

Responsibilities

• Evaluate conversations between users and AI models for neutrality and safety.
• Identify patterns of concern and document clear rationales.
• Develop rubrics, guidelines and reference responses for balanced replies.
• Design test scenarios to probe non-crisis topics.
• Collaborate with researchers to maintain consistent evaluation standards.

Skills

Neutrality

Written comm

Motivational interviewing

AI safety

Boundaries

Cognitive distortions

Sycophancy

Education

Degree in psychology, counseling, social work, behavioral health, or related field

Job description

Cincinnatus LLC is seeking behavioral health experts to evaluate how AI models respond in sensitive conversations, including relationships, wellbeing, and personal beliefs. This is a part‑time W‑2 role with 20–40 hours per week, placed at a leading AI lab as part of the extended workforce.

Ideal candidates have 3+ years in mental health or related fields and can clearly justify professional judgments while maintaining neutrality and precise written feedback.