What's in a Name? Implications of AI Roles and Mind Perception for Human-AI Teams

Area of Science:

• Human-Computer Interaction
• Artificial Intelligence Ethics
• Cognitive Science

Background:

• The term "AI teammate" is increasingly adopted without full understanding of its impact on human-AI collaboration.
• Reframing AI as a teammate may not automatically confer the benefits of human teamwork to human-AI teams.
• Assessing the risks and benefits of AI role reframing is vital for effective AI integration.

Purpose of the Study:

• To investigate how artificial intelligence (AI) roles (teammate, support, tool) influence user acceptance through the lens of mind perception subdimensions: agency and conscious experience.
• To differentiate the effects of AI role perception versus AI role presentation on acceptance outcomes.
• To explore the differential relationships between agency, conscious experience, and technology- vs. human-centered acceptance perspectives.

Main Methods:

• Three studies were conducted, including a survey of weekly AI users and two experimental vignette studies.
• The research examined the relationship between AI roles and acceptance from technology-centered (Technology Acceptance Model, cognitive trust) and human-centered (affective trust) viewpoints.
• A key methodological distinction was made between AI role perception and AI role presentation.

Main Results:

• Perceiving AI as a "teammate" correlated positively with higher mind perception compared to less collaborative roles.
• While perceiving AI as a teammate consistently enhanced acceptance, presenting AI as a teammate negatively impacted some acceptance outcomes, even after accounting for mind perception.
• Agency and conscious experience showed distinct relationships with technology- and human-centered outcomes, underscoring the need for integrated perspectives.

Conclusions:

• Shifting AI roles towards a "teammate" designation can improve acceptance through enhanced mind perception.
• Strategic implementation of AI roles is necessary, as artificially labeling AI as a "teammate" may lead to unintended negative consequences.
• Intermediate AI role labels, such as "support," may more accurately reflect the current capabilities of many AI systems.