Good Robots, Bad Robots: Morally Valenced Behavior Effects on Perceived Mind, Morality, and Trust

Abstract

Both robots and humans can behave in ways that engender positive and negative evaluations of their behaviors and associated responsibility. However, extant scholarship on the link between agent evaluations and valenced behavior has generally treated moral behavior as a monolithic phenomenon and largely focused on moral deviations. In contrast, contemporary moral psychology increasingly considers moral judgments to unfold in relation to a number of moral foundations (care, fairness, authority, loyalty, purity, liberty) subject to both upholding and deviation. The present investigation seeks to discover whether social judgments of humans and robots emerge differently as a function of moral foundation-specific behaviors. This work is conducted in two studies: (1) an online survey in which agents deliver observed/mediated responses to moral dilemmas and (2) a smaller laboratory-based replication with agents delivering interactive/live responses. In each study, participants evaluate the goodness of and blame for six foundation-specific behaviors, and evaluate the agent for perceived mind, morality, and trust. Across these studies, results suggest that (a) moral judgments of behavior may be agent-agnostic, (b) all moral foundations may contribute to social evaluations of agents, and (c) physical presence and agent class contribute to the assignment of responsibility for behaviors. Findings are interpreted to suggest that bad behaviors denote bad actors, broadly, but machines bear a greater burden to behave morally, regardless of their credit- or blame-worthiness in a situation.

From Warranty Voids to Uprising Advocacy: Human Action and the Perceived Moral Patiency of Social Robots

“…Agents' moral status may be understood as having two primary dimensions: moral agency and moral patiency. Moral agency is the capacity to be good and do good (Banks, 2020a) and its relevance to robots has received ample attention in extant literature. Less attention has been paid to moral patiency-the ways in which robots may be victims or beneficiaries of (im)moral action (Gunkel, 2018).…”

“…Understanding the robot-as-PMP can be challenging in that mental models are proverbial black boxes (Rouse and Morris, 1985) and people often will not overtly ascribe moral status to robots as they would to humans despite judging their behaviors as similarly good or bad (Banks, 2020a). It is useful, then, to draw on moral typecasting theory (MTT; Gray and Wegner, 2009) in tandem with eliciting hypothetical stories to infer mental-model content (cf.…”

Framing Effects on Judgments of Social Robots’ (Im)Moral Behaviors

“…Evidence is mixed on whether social robots and humans may be judged similarly (e.g., Banks, 2020a ) or differently (e.g., Malle et al, 2015 ) for (im)moral actions. This divergence may be a function of framing differences.…”

“…In contrast to the morally valenced scenarios in Study 1, these videos presented morally ambiguous behaviors—responses that both upheld and violated each moral foundation ( Krakowiak and Oliver, 2012 ). This was necessary because behavior evaluations are known to align with moral upholding/violation ( Banks, 2020a ), while ambiguous scenarios permit behavioral evaluations to vary according to the frame. Participants were told that Ray was asked to talk about a time when she encountered particular situations.…”

“…Although survey procedures leveraging video stimuli have the benefit of efficiently recruiting large samples, extant evidence indicates that mediated presentations of robots garner different social and moral evaluations compared to in-person exposures ( Schreiner et al, 2017 ; Banks, 2020a ). Because robots are physical embodiments of AI, Study 2 was replicated in a face-to-face setting to determine whether copresence may differently foster framing effects.…”