Empathic Nuances of Machine-Mediated Asylum-Seeker Narratives

Abstract

Abstract The use of artificial human-like agents within the legal context of immigration is still a largely understudied topic. We employed a computer- generated asylum seeker with a narrative interpolated by ChatGPT4 to examine whether “uncanniness” in machine-mediation affected empathic decision-making. In an online study, 466 participants were instructed on the United Nations’ legal standard for granting asylum. They were then asked to simulate the role of an immigration officer by deciding whether an asylum narrative, conveyed either by written text or spoken by a computer- generated asylum seeker, met the standard. Results show that regardless of the narrative's mode of delivery, the same facts overwhelmingly led to the granting of asylum. Our findings encourage new applications of machine-mediated approaches to the promotion of social justice initiatives.