The impact of human-AI collaboration types on consumer evaluation and usage intention: a perspective of responsibility attribution

Abstract

Despite the widespread availability of artificial intelligence (AI) products and services, consumer evaluations and adoption intentions have not met expectations. Existing research mainly focuses on AI’s instrumental attributes from the consumer perspective, along with negative impacts of AI failures on evaluations and willingness to use. However, research is lacking on AI as a collaborative agent, investigating the impact of human-AI collaboration on AI acceptance under different outcome expectations. This study examines the interactive effects of human-AI collaboration types (AI-dominant vs. AI-assisted) and outcome expectations (positive vs. negative) on AI product evaluations and usage willingness, along with the underlying mechanisms, from a human-AI relationship perspective. It also investigates the moderating role of algorithm transparency in these effects. Using three online experiments with analysis of variance and bootstrap methods, the study validates these interactive mechanisms, revealing the mediating role of attribution and moderating role of algorithm transparency. Experiment 1 confirms the interactive effects of human-AI collaboration types and outcome expectations on consumer evaluations and usage willingness. Under positive outcome expectations, consumers evaluate and express willingness to use AI-dominant intelligent vehicles with autonomous driving capabilities higher than those with emergency evasion capabilities (AI-assisted). However, under negative outcome expectations, consumers rate autonomous driving capabilities lower compared to emergency evasion capabilities. Experiment 2 examines the mediating role of attribution through ChatGPT’s dominant or assisting role under different outcome expectations. Experiment 3 uses a clinical decision-making system to study algorithm transparency’s moderating role, showing higher transparency improves evaluations and willingness to use AI products and services under negative outcome expectations. Theoretically, this study advances consumer behavior research by exploring the human-AI relationship within artificial intelligence, enhancing understanding of consumer acceptance variations. Practically, it offers insights for better integrating AI products and services into the market.