Benevolent Bias in Multi-Turn Human-Agent Dialogue

arXiv:2608.29206v2 Announce Type: replace Abstract: Bias in human-agent interaction can manifest not only through hostile language but also as benevolent bias, whereby unequal treatment hides behind a warm, positive tone. To make it detectable, we…

Read the full story at arXiv cs.AI ↗