Paper Type
ERF
Abstract
As anthropomorphic agents become increasingly embedded in organizational workflows, users face a novel and undertheorized security challenge: evaluating both the authenticity of an agent and the legitimacy of its requests simultaneously. This paper introduces two analytically separable constructs. Surface ambiguity refers to perceptual uncertainty about whether an agent is human, artificial, or adversarial. Interpretive ambiguity refers to uncertainty about the intent and legitimacy of a request itself. We argue that resolving one form of ambiguity can impair the other. Extending Media Richness Theory into adversarial contexts and drawing on dual-process theory, we propose that anthropomorphic realism backfires: agents with high perceptual fluency suppress the analytical scrutiny needed to detect manipulative intent, disproportionately exposing users to interpretive ambiguity. We present a pre-registered 2×2 between-subjects experiment in which surface ambiguity is manipulated via response latency and interpretive ambiguity via authority and urgency framing, with credential disclosure as the primary behavioral dependent variable. We predict that the highest rate of credential disclosure will occur under conditions of low surface ambiguity paired with high interpretive ambiguity. Theoretical and practical implications for IS security research and organizational defense design are discussed.
Paper Number
1476
Recommended Citation
Kim, Zisu; D'Arcy, John; and Straub, Detmar, "When Anthropomorphic Realism Backfires: Surface vs. Interpretive Ambiguity in AI-mediated IT Support Requests" (2026). AMCIS 2026 Proceedings. 15.
https://aisel.aisnet.org/amcis2026/sig_sec/sig_sec/15
When Anthropomorphic Realism Backfires: Surface vs. Interpretive Ambiguity in AI-mediated IT Support Requests
As anthropomorphic agents become increasingly embedded in organizational workflows, users face a novel and undertheorized security challenge: evaluating both the authenticity of an agent and the legitimacy of its requests simultaneously. This paper introduces two analytically separable constructs. Surface ambiguity refers to perceptual uncertainty about whether an agent is human, artificial, or adversarial. Interpretive ambiguity refers to uncertainty about the intent and legitimacy of a request itself. We argue that resolving one form of ambiguity can impair the other. Extending Media Richness Theory into adversarial contexts and drawing on dual-process theory, we propose that anthropomorphic realism backfires: agents with high perceptual fluency suppress the analytical scrutiny needed to detect manipulative intent, disproportionately exposing users to interpretive ambiguity. We present a pre-registered 2×2 between-subjects experiment in which surface ambiguity is manipulated via response latency and interpretive ambiguity via authority and urgency framing, with credential disclosure as the primary behavioral dependent variable. We predict that the highest rate of credential disclosure will occur under conditions of low surface ambiguity paired with high interpretive ambiguity. Theoretical and practical implications for IS security research and organizational defense design are discussed.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
SIG SEC