Persuaded, Not Informed: Incentive-Misaligned Witnesses Defeat In-Context Grounding

Abstract:Language-model agents increasingly answer questions over customer-relationship management (CRM) records, such as whether to qualify a sales lead. We identify a failure mode not addressed by a stronger model: when the context contains an assertion by a party with an incentive toward optimism - here the sales representative, a witness recorded in the CRM - the model treats the assertion as evidence and clears deals the company's own records deem unacceptable. Across 100 lead-qualification tasks from CRMArena-Pro, the representative asserts an acceptable timeline in every call and an acceptable budget in 76; on the 31 tasks where such an assertion contradicts the price list and installation policy, a model reading only the transcript clears the deal in 29 of 31 cases. The signature is consistent across seven models from four providers (misled on 87-97%); scale and explicit reasoning confer no resistance. Only 3 of 35 genuine failures involve no assertion: the failure is persuasion, not missing information. We contribute a diagnostic method rather than an architecture: (i) a bucket analysis that separates persuasion from information gaps, (ii) a same-information control showing that supplying the records to the model lowers strict accuracy from 41 to 18 while raising recall - precision collapses - and (iii) a compute-step control that holds extraction fixed and varies only who computes Budget and Timeline. The margin ranges from 42 points on an inexpensive model to 2-5 points on models that already compute correctly; on the strongest models the arms are within confidence intervals, so the pattern is a consistent direction and a soundness property, not a proved performance floor. We pre-specify a generalization test that returns a negative result, characterize the precondition (a policy exactly specified in the inputs), and release all evaluation artifacts.

9 pages, 4 figures, IEEE conference format. Ancillary files contain the evaluation harness, pre-specifications, and per-run result files

Computation and Language (cs.CL); Artificial Intelligence (cs.AI)

arXiv:2609.28854 [cs.CL]

(or arXiv:2609.28854v1 [cs.CL] for this version)

https://doi.org/10.48550/arXiv.2609.28854 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

accuracy_sweep.json

anthropic_acc.json

compute_control_claude-fable-5-1.json

compute_control_claude-sonnet-5.json

compute_control_claude-sonnet-5_run1.json

compute_control_claude-sonnet-5_run2.json

compute_control_gpt-4o-mini.json

compute_control_gpt-5.6-sol.json

compute_control_gpt-5_seed1.json

compute_control_gpt-5_seed2.json

compute_control_kimi-k2.6.json

compute_control_o3-mini_seed1.json

compute_control_o3-mini_seed2.json

compute_control_qwen3.8-max.json

err_budget_rows.json

gullibility_rows.json

gullibility_sweep.json

gullibility_sweep_providers.json

gullibility_sweep_v5.json

invalid_config_OUTCOME.md

invalid_config_preregistration.json

knowledge_qa_OUTCOME.md

knowledge_qa_preregistration.json

memory_answerable_classification.json

ov_accuracy_sweep.py

ov_anthropic_acc.py

ov_compute_control.py

ov_config_eval.py

ov_gullibility.py

ov_gullibility_sweep.py

ov_provider_sweep.py

ov_provider_sweep_v5.py

ov_rag_extract_code.py

rag_extract_code.json

rag_extract_code_tonly.json

(41 additional files not shown) You must enabled JavaScript to view entire file list.

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.