Trust and empathy toward artificial agents under task-language uncertainty

Area of Science:

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

Background:

• Artificial agents are increasingly used to aid human decision-making.
• Users often face situations with incomprehensible task content or system reasoning.
• Understanding how users form trust and empathy towards AI under uncertainty is crucial.

Purpose of the Study:

• To investigate the formation of trust and empathy towards artificial agents under task-language uncertainty.
• To examine the impact of explanations on trust and empathy processes.
• To explore the dissociation between performance-based and social evaluations of AI.

Main Methods:

• An online quiz-based experiment was conducted.
• Task language (comprehensible vs. incomprehensible) and explanation (present vs. absent) were manipulated.
• Pre-post changes in trust and empathy were measured, along with confidence, responsibility attribution, and task accuracy.

Main Results:

• Task comprehension positively impacted performance, confidence, and self-responsibility attribution.
• Trust increased significantly in the incomprehensible task-language condition, amplified by explanations.
• Explanations did not improve task accuracy, but empathy increased across conditions, modulated by language clarity.

Conclusions:

• Trust in artificial agents can increase even when users cannot fully comprehend the task or independently evaluate performance.
• Explanations play a role in enhancing trust, particularly under conditions of uncertainty.
• A dissociation exists between performance-based assessments and social evaluations (trust, empathy) of AI systems.