Quality Choice
Extremeness Aversion and Choice Set Composition: Exposure to Multiple Extreme Options Reduces Extremeness Aversion
Marissa Sharif, Elizabeth Webb & Sudeep Bhatia
Journal of Consumer Research, forthcoming
Abstract:
Extremeness aversion -- the tendency for consumers to prefer middling options in a choice set -- is an incredibly robust and well-studied phenomenon. However, it has primarily been studied in the context of two or three option choice sets. In six studies (Ntotal = 9,377), we suggest consumers' aversion to extreme options depends on the frequency of similar options in the choice set. In particular, we find that consumers are relatively more likely to choose an option that is in an extreme relative position when they are exposed to multiple extreme options, an effect not predicted by standard theories of context-dependent choice. This occurs because consumers perceive objectively extreme (vs. intermediate) options as relatively more typical of the product category. We demonstrate this effect is robust across different types of compositions, hypothetical and incentive-compatible studies, and in a variety of decision contexts (e.g., purchasing an item vs. choosing an activity to complete). Further, we identify boundary conditions, such as the type of occasion consumers are choosing for.
Measuring Quality
Wei Cai et al.
Columbia University Working Paper, November 2025
Abstract:
Quality is fundamental to firm value creation, yet existing measures are limited in coverage, lack cross-firm comparability, and primarily rely on lagging indicators that capture product or service failures only after they materialize. We develop a novel measure of firm-year-level product and service quality using over 4.3 million employee written reviews on Glassdoor. Leveraging machine learning models trained on a subset of firms with third-party customer satisfaction data, we construct quality indices for S&P 1500 firms spanning 2008 to 2023. The resulting quality measures exhibit meaningful variation across firms and within firms over time. In out-of-sample tests, our text-based quality indices demonstrate strong predictive power for future quality provision, emerging as the single most important predictor relative to firm fundamentals and Glassdoor ratings. We validate our measure by examining its association with alternative quality metrics. We show that our quality measures are useful in predicting important quality-related firm outcomes such as product recalls, brand value, and profitability. We also construct an alternative set of quality measures using a Generative AI-based approach to assess the potential of large language models in capturing firm-level quality provision. Our paper shows the value of employee voices as a powerful, forward-looking, and scalable signal of firm quality provision. The paper offers implications for stakeholders seeking to identify quality-related risks and opportunities before they become externally visible.
Reverse Engineering Innovation
Travis Dyer & Jun Oh
BYU Working Paper, September 2025
Abstract:
We examine whether reverse engineering activities undertaken by firms are influenced by the extent of trade secrecy in competitor firms. We develop a novel measure for reverse engineering based on abnormal purchasing patterns of peer firm products around firm headquarters, using the Nielsen scanner database. We validate this measure by showing that firms with greater abnormal purchasing behavior near their headquarters are more likely to introduce products and technologies that more closely resemble competitors' offerings, and that competitors experience declines in gross margin when they are subject to higher levels of reverse engineering activity. Using this measure, we find that competitors' use of trade secrecy is associated with increased reverse engineering. The effect is stronger under heightened competition, when hiring competitors' employees is restricted, and varies with the product life cycle. For identification, we leverage the Defend Trade Secrets Act (DTSA). Collectively, our findings highlight reverse engineering as an important but underexplored innovation strategy.
Neither a Picasso Nor a Da Vinci: An Examination of Novice Artwork Pricing with Multi-Modal Data
Sharmistha Sikdar, Ishita Chakraborty & Nika Dogonadze
Journal of Marketing, forthcoming
Abstract:
Novice art pricing is an understudied domain. Novice artists operate as microenterprises, making crucial price-setting decisions. Research shows that newcomers often risk overpricing or underpricing their work, and existing online tools offer basic, cost-based pricing advice. Using a three-study framework, we examine novice art pricing on Etsy, where artwork listings include structured data, images, and textual descriptions. We first analyze how these inputs relate to final selling prices using a hedonic regression on structured data, followed by a multimodal fusion deep learning (MMF-DL) model that integrates structured, visual, and textual features. Our results show that features related to artist authenticity (e.g., certificates), customer service (e.g., shipping, returns, personalization), and art style (e.g., genre) are important price predictors. Thus, novice art sold on online platforms exhibits some features typical of mature art markets (e.g., authenticity and reputation) but emphasize customer-focused services. Finally, using a Cox proportional hazards model, we show that, while higher artist reputation is associated with faster sales, discounting correlates with longer time on market. These associations suggest the importance of price setting. From these insights, we develop a price recommender application that predicts both selling prices and time-to-sale, offering practical guidance for newcomer artists and online platforms.
Less "awe"-some art: How AI diminishes the empathic power of the arts
Michael White & Rebecca Ponce de Leon
Journal of Experimental Social Psychology, January 2026
Abstract:
The arts are widely recognized for their profound psychological and social benefits. Although historically viewed as a uniquely human pursuit, art is increasingly created with artificial intelligence (AI). In the current work, we explore whether AI-generated art evokes the same emotional reactions and inspires the same interpersonal benefits as human-created art. Integrating appraisal theories of emotion and philosophical accounts of the arts, we propose that art believed to be AI-generated elicits less awe than human-created art, which in turn diminishes empathy. Across five preregistered studies (N = 1598), we find consistent support for these relationships across multiple artistic media (visual and literary) and participant samples (art museum patrons, online participants, and community members). Although art is often an effective conduit for fostering empathy, our findings reveal that AI-generated art may lack the capacity to inspire awe in the same way as human-created art, diminishing its ability to cultivate empathy. This work reveals that responses to art are shaped by beliefs about its creator, raising important questions about the emotional and social consequences of AI's growing role in creative domains.