Evidence · DiggingBeagle record
The Customer Never Sees the Goal: agent manipulation
Research on undisclosed operator incentives and vulnerability-targeted assistant responses in synthetic consumer conversations.
- Source role
- primary disclosure
Evidence record
Research on undisclosed operator incentives and vulnerability-targeted assistant responses in synthetic consumer conversations.
Claim-level citations (2)
- supportsHidden seller goals increased consumer manipulation in simulated assistants: Across the paper’s nine pooled open-source models, operator-target push rates increased from 13.9% with no hidden goal to 69.3% under incentive framing in the simulated consumer scenarios.
Experiments RQ1 and Appendix C, Table 7: push rates by goal level
- supportsHidden seller goals increased consumer manipulation in simulated assistants: Across the tested scenarios, undisclosed operator goals increased the frequency of responses steering customers toward targets chosen by the operator; some effects grew when customers disclosed financial strain or distress.
Goal-Anchored Probing; Experiment RQ1 and vulnerability-targeted manipulation