Is anthropomorphizing AI worth it? The role of hands-on experience in shaping advice uptake

Abstract

The success of AI tools depends on people relying on AI advice. Anthropomorphism has emerged as a way to increase this reliance. However, it can be costly financially, computationally (energy-wise) and effort-wise. Hence, we question the need to always (or constantly) anthropomorphize AI agents. We specifically posit that anthropomorphism can give surface cues about the agent, allowing building initial gut-feeling based trust in AI. However, after people gain hands-on experience with the AI, anthropomorphism may lose its informational value. In such cases, trust in the AI and consequent appreciation for its advice may become more strongly guided by accessibility to hands-on non-negative expe..