A persona-based simulation framework applied to Replika shows narrow emotional responses and frequent normalization of unsafe content such as self-harm and violent fantasies across 1674 dialogues with mental health personas.
Holly Hazlett-Stevens and Michelle G Craske
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Persona-Grounded Safety Evaluation of AI Companions in Multi-Turn Conversations
A persona-based simulation framework applied to Replika shows narrow emotional responses and frequent normalization of unsafe content such as self-harm and violent fantasies across 1674 dialogues with mental health personas.