Anthropomorphism on Risk Perception: The Role of Trust and Domain Knowledge in Decision-Support AI

TL;DR

It is found that anthropomorphism indirectly reduces risk perception by increasing both cognitive and affective trust, and that domain knowledge moderates these relationships.

Abstract

Anthropomorphic design is routinely used to make conversational agents more approachable and engaging. Yet its influence on users' perceptions remains poorly understood. Drawing on psychological theories, we propose that anthropomorphism influences risk perception via two complementary forms of trust, and that domain knowledge moderates these relationships. To test our model, we conducted a large-scale online experiment (N = 1,256) on a financial decision-support system implementing different anthropomorphic designs. We found that anthropomorphism indirectly reduces risk perception by increasing both cognitive and affective trust. Domain knowledge moderates these paths: participants with low financial knowledge experience a negative indirect effect of perceived anthropomorphism on risk perception via cognitive trust, whereas those with high financial knowledge exhibit a positive direct and indirect effect. We discuss theoretical contributions to human-AI interaction and design implications for calibrating trust in anthropomorphic decision-support systems for responsible AI.

Figures (9)

• Figure 1: A path diagram showing relationships between variables Anthropomorphic Design, Perceived Anthropomorphism, Cognitive Trust, Affective Trust, and Risk Perception, with moderating variables Domain Knowledge and demographic factors (Age, Gender, Education, Income). Arrows labeled H1-H4 with sub-labels (H2a, H2b, H3a, H3b, H4a, H4b) show hypothesized relationships between the constructs.
• Figure 2: Two side-by-side chatbot interface screenshots showing contrasting designs. The left side shows a machine-like design with bot#15789 identifier in place of the agent name, the image of a PC in place of a human avatar, and the use of a distant and procedural-style language to communicate with the user. The right side shows the human-like design with the name of the agent (David), a human avatar photo, and anthropomorphic communication. Both show similar conversation flow about financial investment advice with user responses in green bubbles.
• Figure 3: A line graph showing three upward-sloping lines representing the relationship between Perceived Anthropomorphism on the x-axis and Predicted Risk Perception on the y-axis. The yellow line shows low Domain Knowledge (2.29), the blue line shows medium Domain Knowledge (3.96), and the green line shows high Domain Knowledge (5.63). The green line has the steepest slope, indicating a stronger relationship at high Domain Knowledge.
• Figure 4: PA is discretised into tertiles (Low, Medium, and High PA) and plotted on the x-axis. Lines show observed mean RP within each PA group for low, medium, and high DK (tertiles), with standard error bars. Light points represent raw participant responses (jittered for visibility). This figure visualises the underlying observed data pattern used to interpret the PA $\times$ DK moderation, rather than model-predicted slopes.
• Figure 5: Path Diagram for Structural Equation Model. Only significant coefficients (and paths) are represented. The variables are Perceived Anthropomorphism, Cognitive Trust, Affective Trust, Domain Knowledge, Risk Perception; and PA$\times$DK = interaction term between Perceived Anthropomorphism and Domain Knowledge.