Notes
Trust Resilience in Pedagogical Agents: Will Anthropomorphism Help Against Trust Decline?
146Automated agents are becoming more important in all areas of life, including learning and teaching. Thus, for optimal technology-mediated interaction with such automated agents, trust becomes important, especially appropriate calibration of trust. However, it is still unclear, if human-automation trust is similar to human-human trust (media-equation hypothesis) or if certain biases may prompt unique trust dynamics between humans and automated agents (unique agent hypothesis).
147This study contributes to this discussion by tracking trust calibration regarding human and automated agents in a learning task. We hypothesize that trust declines as agent reliability declines (trust calibration) and that anthropomorphism (human-like qualities) should buffer against this decline in line with the unique agent hypothesis. This has already been investigated in other contexts (e.g., de Visser et al., 2016) but not yet in the context of language learning. Context may play a role because this buffering effect of anthropomorphism has not been conclusively demonstrated in previous studies.
148In a conceptual replication of de Visser et al. (2016), n = 30 participants repeatedly guess the meaning of foreign words in a 4x4 within-subject design, that is, with the help of pedagogical agents varying in anthropomorphism (human, robot, smart speaker, and computer) that give recommendations systematically decreasing in reliability over time (from 100%, 67.5%, 50%, to 0%). In each of the 96 trials, participants (1) guess alone, (2) get an agent’s recommendation, and (3) make their final decision. At each step, participants judge their self-confidence and rate their trust in the agent. The final decision (3) is also a measure of behavioral trust. Behavioral and judgment data are analysed via mixed-effects models and ANOVAs. Trust calibration would be indicated by declining trust with declining reliability. The unique agent hypothesis would be supported by an interaction of anthropomorphism and reliability regarding trust.
149Trust in technology / anthropomorphism / artificial intelligence / language learning / pedagogical agent