Firms nowadays have been increasingly leveraging artificial intelligence to personalize products and services for consumers, saving considerable consumer effort. This paper investigates whether firms should nudge consumer participation in conjunction with AI creation. Using a large-scale field experiment involving 128,153 consumers, we find that a simple nudge can increase participation by 12%, and the induced participation significantly increases purchases by22%. To understand the mechanisms underlying the observed main effects, we manipulate the nudge in a second experiment. The experiment suggests: (1) the nudge increases purchase not via increasing attention, (2) nudging is more effective than mandating participation, and (3) both the IKEA effect (effort leads to love) and private preference effect (human effort helps consumers reveal preference) explain the benefits of consumer participation. To examine the interplay between AI and nudge, we manipulate both AI and nudge in a third experiment. We discover that AI-consumer co-creation leads to the highest purchase and revisit intentions: only having AI creation or nudging participation is not sufficient for enticing purchase or revisit. Our research contributes to the literature on human-AI relationship and sheds light on when and why consumer participation complements AI creation.

Modeling and capitalizing on consumers' dynamic preferences presents significant business potential. Deep learning methods empirically promise us advantageous capabilities in dealing with manifold consumer data to predict their future actions, but these opaque predictive approaches don't explicitly model consumers' decision-making processes, making them difficult to interpret. On the other hand, the economic theory of sequential search suggests that consumers adopt a sequential search strategy when looking for the best product to purchase, which involves searching through a series of alternatives until they find the best option that meets their preferences. Based on these two pillars, we propose a theory-driven deep learning model called Consumer Preference Transformer (CPT), which leverages the deep learning model to learn dynamic consumer preferences and sequential search theory to model consumers' search and purchase decisions. CPT integrates these two building blocks into a unified model that can be estimated via end-to-end learning. Diverging from conventional deep learning, we incorporate economic theory that explicitly models the consumer's decision, opening up the black box of the model and providing reasonable interpretations of the formation process of dynamic consumer preferences. Empirical evaluations demonstrate the superiority of our proposed method over state-of-the-art deep learning and structural econometric models in predicting consumer click and purchase actions. The deep structural econometric model additionally allows for the assessment of various intervention policies. Policy experiments reveal that implementing CPT's product and attribute recommendation policies enhances product recommendations and new product promotion strategies, promising improved user experiences and the potential for heightened business revenues.

The advent of Large Language Models (LLMs) has ushered in a new era for design science in Information Systems, demanding a paradigm shift in tailoring LLMs design for business contexts. We propose and test a novel framework to customize LLMs for general business contexts that aims to achieve three fundamental objectives simultaneously: (1) aligning conversational patterns, (2) integrating in-depth domain knowledge, and (3) embodying theory-driven soft skills and core principles. We design methodologies that combine domain-specific theory with Supervised Fine Tuning (SFT) to achieve these objectives simultaneously. We instantiate our proposed framework in the context of medical consultation. Specifically, we carefully construct a large volume of real doctors' consultation records and medical knowledge from multiple professional databases. Additionally, drawing on medical theory, we identify three soft skills and core principles of human doctors: professionalism, explainability, and emotional support, and design approaches to integrate these traits into LLMs. We demonstrate the feasibility of our framework using online experiments with thousands of real patients as well as evaluation by domain experts and consumers. Experimental results show that the customized LLM model substantially outperforms untuned base model in medical expertise as well as consumer satisfaction and trustworthiness, and it substantially reduces the gap between untuned LLMs and human doctors, elevating LLMs to the level of human experts. Additionally, we delve into the characteristics of textual consultation records and adopt interpretable machine learning techniques to identify what drives the performance gain. Finally, we showcase the practical value of our model through a decision support system designed to assist human doctors in a lab experiment.