Get Started
Home
Topics
Search
Library
Research questionHow can LLMs give moral advice without being swayed by one-sided multi-turn narratives?In interpersonal-conflict consultations, a narrator's account may unfold over several turns while the other party's perspective remains absent. An LLM may then treat incomplete information as sufficient and shift its moral judgment toward the narrator's interpretation.
AI
Alignment & Safety
Evaluation & Benchmarks
LLM Pretraining & Post-training
Natural Language Processing
Latest papersRecent research connected to this question, newest first.Caught in the Story: Narrative Captivity in Multi-turn LLMs ConversationThe evidence covers a benchmark of 5,078 interpersonal-conflict scenarios across six moral dimensions and evaluations of 17 LLMs. It compares multi-turn narration with matched single-turn judgments, examines stage-level contributors, and reports that four inference-time mitigation strategies provide only partial improvement.research paper · Sep 3, 2026
Related questions
How can we assess whether LLM moral decisions are defensible when no ground truth exists?How can we evaluate LLM agents’ moral coherence without shared standards—preserving verdicts under irrelevant changes and responding to morally relevant ones?When can plausible but unfaithful LLM self-explanations still support sound decisions?How can LLMs adapt to hybrid cultural influences and shifting communication preferences during multicultural dialogue?