Research

Abstract 

As artificial intelligence (AI) becomes more autonomous and socially present, it is critical to understand how people accept AI not just as a technological tool, but also as an agent capable of (semi)autonomous decision-making and interaction. With a meta-analysis of 287 effect sizes representing over 119,000 individuals, this research examines the factors driving human acceptance of AI. Through a dual-perspective framework, AI as a tool versus AI as an agent, the authors identify key AI characteristics, including capability, role, expertise scope, and anthropomorphism, that significantly influence acceptance. These engineerable AI characteristics, along with contextual and individual factors, form an AI–task–user framework that explains AI acceptance across different use scenarios and user groups. These findings contribute to the discourse on AI acceptance and human–AI interactions, revealing a small, decreasing reluctance to accept AI and, more importantly, directing future research to empirical testing and theory building of AI acceptance from an agentic perspective. This research also provides an actionable user-centered design roadmap for practitioners to develop and communicate AI features that align with human expectations and enhance positive responses, especially at a time when agentic AI is rapidly becoming a technological and societal reality.

Abstract

To balance the need for privacy and the benefits of big data analytics, regulators around the world are giving consumers control over their data, allowing them to choose whether or not to voluntarily share their purchase history data with firms. Intuition suggests that voluntary data sharing benefits consumers who can now choose to share their data only when it is profitable to do so. To investigate this argument, we built a model in which a monopolistic firm sells a repeatedly purchased product to consumers over two periods, and consumers decide whether or not to share their purchase history data with the firm, who can use it in the future to price-discriminate against them. We found that, compared to when data collection is completely outlawed, voluntary data sharing can benefit the firm but at the consumer’s expense. Moreover, regulations that mandate firms to better protect consumer data against data breaches can backfire on consumers. Finally, we show that under voluntary data sharing, a firm’s ability to offer consumers a monetary incentive to share their data can improve profits without hurting consumers. Taken together, these findings underscore the surprising effects of voluntary data sharing and caution public policymakers of how certain data policies that, on the surface, seem purely beneficial can lead to unintended consequences.

Abstract 

Peer-to-peer marketplaces and the sharing economy are reshaping many markets. Airbnb is attractive to many customers because of its non-standardized, diverse selections and superior value over hotels, but Airbnb properties also come with more uncertainty than hotels. Images of the property can help resolve uncertainty. This study focuses on the background image (the image displayed in search results) and examines how the background image’s content features (living room, bedroom, and interior design) and aesthetic features (clarity, brightness, and contrast) affect the booking rate during a 16-night end-of-year holiday period. The authors develop a model that includes the amount of visual information, property characteristics, and host characteristics, and correct for endogeneity by using a pair of markets (New York City and San Francisco) to calculate propensity scores and construct instrumental variables. The results show that having a background image that features the living room and shows more interior design elements increases the booking rate, while featuring a bedroom decreases it. The effects of the content features are larger than the effects of the aesthetic features, and the effects are economically significant: for example, the effect of featuring the living room in the background image translates into a 35% increase in the booking rate, which amounts to $728 more revenue during the holiday period.

Abstract

Digital twins of consumers have emerged as a promising approach to simulate consumer thinking, feeling, and decision-making. Grounded in the psychological theory, which conceptualizes behavior as a function of both personal traits and contextual factors, this research proposes and validates a dual-component framework for constructing Large Language Model (LLM)-based consumer digital twins. Fine-tuning on consumer-specific data including usergenerated content allows the model to internalize individual traits, preferences, cognitive and behavioral patterns, while retrieval-augmented generation (RAG) dynamically incorporates information specific to consumer context at inference. We demonstrate and validate the framework with Amazon e-commerce data, constructing 304 personified digital twins and evaluating their performance in predicting purchase decisions and review contents. The digital twins can predict future purchases with an average accuracy of 86% and to generate product reviews with strong semantic alignment to actual consumer-generated content (cosine similarity above 0.94). By aligning LLM adaptation techniques with the foundational psychological theory about behavior, our method enables psychologically grounded simulations of individual-level consumer behavior at a scale. This research contributes to the literature on generative AI, synthetic agents, and digital twins in consumer research, and at the same time, offers a new methodology for theory-driven modeling and privacy-compliant personalization in practice.