The Dangers of trusting Stochastic Parrots: Faithfulness and Trust in Open-domain Conversational Question Answering
This addresses the problem of misleading trust in AI systems for users and developers, revealing that certain linguistic cues can mask inaccuracies, which is an incremental finding building on known issues of unfaithfulness in large language models.
The paper investigates how advanced linguistic dialog behaviors, like lexical alignment, increase user trust in open-domain conversational question answering systems, even when the responses are unfaithful, highlighting risks in systems that mimic user input while providing incorrect answers.
Large language models are known to produce output which sounds fluent and convincing, but is also often wrong, e.g. "unfaithful" with respect to a rationale as retrieved from a knowledge base. In this paper, we show that task-based systems which exhibit certain advanced linguistic dialog behaviors, such as lexical alignment (repeating what the user said), are in fact preferred and trusted more, whereas other phenomena, such as pronouns and ellipsis are dis-preferred. We use open-domain question answering systems as our test-bed for task based dialog generation and compare several open- and closed-book models. Our results highlight the danger of systems that appear to be trustworthy by parroting user input while providing an unfaithful response.