Philosophers have argued that stakes affect knowledge: a given amount of evidence may suffice for knowledge if the stakes are low, but not if the stakes are high. By contrast, empirical work on the influence of stakes on ordinary knowledge ascriptions has been divided along methodological lines: “evidence‐fixed” prompts rarely find stakes effects, while “evidence‐seeking” prompts consistently find them. We present a cross‐cultural study using both evidence‐fixed and evidence‐seeking prompts with a diverse sample of 17 populations in 11 countries, speaking (...) 14 languages. Our study is the first to use an evidence‐seeking prompt cross‐culturally, and includes several previously untested populations (including indigenous populations). Across cultures, we do not find evidence of a stakes effect with our evidence‐fixed prompt, but do with our evidence‐seeking prompt. We argue that the divergent results reveal a tension within folk epistemology: people's beliefs about when it is appropriate to ascribe knowledge differ significantly from their actual practice in ascribing knowledge. (shrink)

Though educators often deal with stressful social conflicts, many face them ad hoc without much training. We studied if and how virtual agents can help University staff manage student-teacher conflicts. We explored educators' verbal, behavioral, and physiological reactions to a virtual agent that brought up a student-teacher conflict and held exit-interviews. Our qualitative analysis revealed that virtual agents for conflict training were positively received, but not for conflict mediation with cross-cultural differences. Those with non-Western backgrounds felt that an agent could (...) help “save face,” whereas Westerners preferred to resolve conflicts in person. In line with this, participants with a Western background rated the virtual agent to be less competent compared to those with non-Western backgrounds. While physiological measures only allow for limited conclusions, we found that participants who believed that the agent was controlled by a human had higher normalized hear rate variability than people who thought that the agent was autonomous. We discuss implications for implementing virtual agents for training purposes, the impact of physiological signals, and the need to consider cultural and individual differences. (shrink)

This report documents the program and the outcomes of Dagstuhl Seminar 21381 "Conversational Agent as Trustworthy Autonomous System ". First, we present the abstracts of the talks delivered by the Seminar’s attendees. Then we report on the origin and process of our six breakout groups. For each group, we describe its contributors, goals and key questions, key insights, and future research. The themes of the groups were derived from a pre-Seminar survey, which also led to a list of suggested readings (...) for the topic of trust in conversational agents. The list is included in this report for references. (shrink)