Predictive Accuracy, Consumer Search, and Personalized Recommendation, with Mengze Shi and Zachary Zhong
Invited for Resubmission to Marketing Science
Firms use predictive technology to attract and direct consumer search through personalized product recommendations. This paper examines the firm’s recommendation strategy by analyzing a key trade-off: accurate recommendations draw high-search-cost consumers into the search process (the “participation-drawing effect”) but may narrow the search intensity of moderate-search-cost consumers (the “search-narrowing effect”). When pricing is inflexible in response to environmental changes, the search-narrowing effect dominates in markets with intermediate predictive accuracy or limited search costs, leading firms to forgo recommendations despite their value in reducing search frictions. However, with pricing flexibility, the no-recommendation strategy is optimal only when both predictive accuracy and search costs are low. Flexible pricing enables firms to capture the surplus from accurate recommendations, strengthening the participation-drawing effect. It also shifts the firm's strategic focus from managing search intensity to managing search participation, increasing the profitability of personalized recommendations. Our findings underscore the dual impacts of personalized recommendations on consumer search behavior and highlight the importance of pricing flexibility in optimizing recommendation strategies. This research provides actionable insights for firms leveraging predictive technologies in customer management.
Product Relevance, Consumer Search, and Competition, with Bing Jing and Mengze Shi
Invited for Resubmission to Marketing Science
When consumers incur search costs to evaluate match, the product sampled first is more likely to make the sale. Product relevance (i.e., the probability of match) then becomes a source of competitive advantage because all else equal, consumers will first sample the product with greater relevance. We examine relevance and price competition in a duopoly. Interestingly, even when consumers have homogeneous search costs, the ex-ante symmetric firms choose different product relevance. The rationale is as follows. If the firms chose an identical relevance, the firm with even a slightly lower price would attract all consumers to first sample its product, intensifying price competition. Differentiation in relevance relaxes price rivalry, as a firm must undercut the competitor's price by a sufficiently large amount to alter consumers' search sequence. Moreover, relevance differentiation expands at low or high search costs but dwindles at intermediate search costs. When search costs are uniformly distributed, one firm chooses a greater relevance and a higher price than the other. Each product is prominent to a different segment of consumers: The consumers with relatively high (low) search costs first sample the product with greater (smaller) relevance.
Comparing Human-Only, AI-Assisted, and AI-Led Teams on Assessing Research Reproducibility in Quantitative Social Science, Social science crowd-analysis with many coauthors | Resubmitted to Nature (2nd Round Revision)
Large Language Models (LLMs) such as ChatGPT are transforming how scientists conduct and validate research. LLMs are thus seen as promising tools to improve scientific reproducibility. We experimentally test how collaboration between researchers and LLM assistants influences the reproduction of quantitative social science findings. Study I (2024) assigned 288 researchers to 103 teams working in three groups: human-only, AI-assisted, and AI-led. In the AI-led group, the LLM conducted reproducibility checks with minimal human oversight. Study II (2025) replicated the design with 95 researchers in 34 teams. Human-only and AI-assisted teams reproduced published results at comparable rates, and both outperformed AI-led teams. Human-only teams also identified more major errors than AI-assisted and AI-led teams. Finally, both human-only and AI-assisted teams outperformed AI-led approaches in both proposing and implementing robustness checks. In an exploratory analysis, we observe that the gap in most outcomes between AI-led and the other two groups began to narrow by the final event of 2024 and was further reduced in 2025. Despite rapid model advances, expert human judgment currently remains indispensable for reliable empirical verification.
Preference for Diversity, with Ying Zeng, Jiajia Liu, and Jingyi Lu
Under Review at Journal of Consumer Research
How do consumers navigate the ubiquitous competition? Prior research has focused on strategies that improve qualification factors that vertically differentiate consumers and determine competition outcomes (e.g., performance or exam scores). Our work uncovers another prevalent yet understudied strategy: diversification on alignment factors (e.g., constellation or token color), which horizontally differentiates consumers without affecting their relative rank. Eight preregistered experiments and a large-scale analysis on gifting decisions in TikTok Live consistently revealed diversification seeking in competitions: consumers prefer to diversify from their competitors, even when explicitly informed that alignment factors do not influence outcomes, when diversification is priced with a premium, and when diversification further disadvantages those already behind. Process evidence supports a motivated reasoning account of belief in obfuscation: when facing competitive disadvantage, consumers are motivated to believe that diversification reduces direct comparisons and obfuscates their disadvantages, thereby improving chances of winning. Consistently, diversification seeking attenuates when (1) competitive motives are weakened, (2) there are few or no competitive disadvantages to obscure, and (3) decisions are made for others (vs. oneself). This research contributes to the literature on consumer competition, motivated reasoning, and differentiation, while also making practical implications for practitioners and individual consumers in competitive contexts.
