Almost human: Anthropomorphism increases trust resilience in cognitive agents

Area of Science:

• Human-Computer Interaction
• Social Psychology
• Automation and Robotics

Background:

• Daily interactions involve computers exhibiting human-like behaviors.
• Conflicting theories exist: some suggest applying social norms to computers, while others propose unique human-machine interaction responses.
• Anthropomorphism is proposed as a key factor in understanding trust dynamics in human-agent interactions.

Purpose of the Study:

• To investigate whether social norms apply to human-computer interactions or if unique human-automation responses occur.
• To examine the role of anthropomorphism in the formation, violation, and repair of trust.
• To resolve the apparent contradiction between social norms and human-automation interaction theories.

Main Methods:

• Three experiments were conducted, varying the appearance and behavior of automated agents (computer, avatar, human).
• Participants received advice from these agents, with reliability gradually decreasing.
• Trust resilience, trust repair behaviors, and the impact of uncertainty were measured.

Main Results:

• Anthropomorphic agents demonstrated greater trust resilience, showing higher resistance to trust breakdowns.
• The positive effect of anthropomorphism on trust resilience was amplified under conditions of greater uncertainty.
• Human-like trust repair behaviors implemented by agents significantly reduced perceived differences between agent types.

Conclusions:

• Anthropomorphism is a critical variable in human-agent trust, influencing trust resilience and repair.
• Findings suggest that designing anthropomorphic agents with human-like trust repair mechanisms can enhance user trust.
• Automation design theories and practical applications should carefully consider and incorporate anthropomorphism.