Generative AI for image synthesis has the potential to transform the digital advertising industry. However, a wide range of uncertainties persists regarding its integration into traditional advertising processes, including finding effective implementations, training methodologies, and achievable performance gains. Specifically, two core challenges limit its practical adoption: a search problem of finding high-performing visuals in a vast creative space, and an alignment problem of ensuring brand and campaign compatibility. This paper proposes a novel end-to-end framework that combines a generative AI with two predictive Bayesian neural networks to identify high-performance and brand-acceptable visuals. We develop a cost-effective Bayesian active learning approach solving simultaneously the dual objectives of performance and alignment. We test the framework in a live advertising campaign for an outdoor activities company. Our system generated a portfolio of visuals achieving higher mean click-through rate and more consistency (lower variance) than creatives from both a professional human designer and a competing AI model optimizing purely for aesthetics. This research provides a validated methodology that bridges the gap between the theoretical potential of generative AI and its practical application, offering a cost-effective solution to the critical search and alignment problems in creative design.

Digital platforms increasingly influence most online markets and ecosystems, creating substantial value for their customers, owners, and other partners. Yet, the challenges associated with platform operations, governance, and regulation continue to evolve. This paper aims to help researchers understand the extensive platform literature to facilitate effective academic contributions. First, we lay out emerging challenges in research topics related to platforms, both from an internal (platform design) and external (platform regulation) perspective. Then, we compare techniques for acquiring and using platform data, both with and without the collaboration of the platforms themselves, to facilitate empirical platform research. Our insights highlight the importance of multidisciplinary and multi-method approaches in studying digital platform policies, value chains and regulation.

The exponential expansion of consumer genetic testing has led to an accumulation of massive genomic data sets owned by governments and firms. The prospect of leveraging genetic data for enhancing consumers' health, well-being, and satisfaction through improved personalization, segmentation, and targeting is promising. Nonetheless, this potential has not been studied empirically to date, and it is unknown whether and when firms should invest resources into incorporating genetic data into strategies and processes. The authors address this gap in a study of taste preferences, important drivers of food and beverage consumption. Using a large U.K.-based sample, they find that with sample sizes currently available, genetic data are expected to significantly improve prediction of taste preferences above traditionally used metrics such as demographics, behavioral variables, and even past consumption, especially for tastes that are uncommon in the local diet (e.g., spicy, sour), as they are less expressed behaviorally. The authors conclude that genetic data show immense promise for prediction-based applications when other data sources are limited or uninformative. These findings could have significant implications for public health initiatives, potentially aiding development of personalized nutrition plans and dietary interventions.

The practical use of nonparametric Bayesian methods requires the availability of efficient algorithms for posterior inference. The inherently serial nature of traditional Markov chain Monte Carlo (MCMC) methods imposes limitations on their efficiency and scalability. In recent years, there has been a surge of research activity devoted to developing alternative implementation methods that target parallel computing environments. Sequential Monte Carlo (SMC), also known as a particle filter, has been gaining popularity due to its desirable properties. SMC uses a genetic mutation-selection sampling approach with a set of particles representing the posterior distribution of a stochastic process. We propose to enhance the performance of SMC by utilizing Hamiltonian transition dynamics in the particle transition phase, in place of random walk used in the previous literature. We call the resulting procedure Hamiltonian Sequential Monte Carlo (HSMC). Hamiltonian transition dynamics have been shown to yield superior mixing and convergence properties relative to random walk transition dynamics in the context of MCMC procedures. The rationale behind HSMC is to translate such gains to the SMC environment. HSMC will facilitate practical estimation of models with complicated latent structures, such as nonparametric individual unobserved heterogeneity, that are otherwise difficult to implement. We demonstrate the behavior of HSMC in a challenging simulation study and contrast its favorable performance with SMC and other alternative approaches. We then apply HSMC to a panel discrete choice model with nonparametric consumer heterogeneity, allowing for multiple modes, asymmetries, and data-driven clustering, providing insights for consumer segmentation, individual level marketing, and price micromanagement.

We report a “Double Decoy” experiment designed to separate two competing accounts of the asymmetric dominance effect. The experiment places an additional decoy alternative within the range of existing alternatives, which should leave choice behaviour unaltered if attributes are weighted by their range. Instead, we observe a decrease in the relative proportion of targets chosen, particularly for subjects who exhibited an initial decoy effect. We also observe considerably more variation in individual behaviour than expected. We therefore consider an alternative theory in which attributes values are compared with diminishing sensitivity (via divisive normalization) and assess its performance in an additional discrete choice experiment previously used in the discrete choice literature. We find that divisive normalization captures behaviour better than range normalization and the linear additive Logit model typically used in applied settings. We therefore propose divisive normalization as both a neuro-computational explanation for context effects and a useful empirical tool for applied researchers.

Heavy alcohol consumption has been associated with brain atrophy, neuronal loss, and poorer white matter fiber integrity. However, there is conflicting evidence on whether light-to-moderate alcohol consumption shows similar negative associations with brain structure. To address this, we examine the associations between alcohol intake and brain structure using multimodal imaging data from 36,678 generally healthy middle-aged and older adults from the UK Biobank, controlling for numerous potential confounds. Consistent with prior literature, we find negative associations between alcohol intake and brain macrostructure and microstructure. Specifically, alcohol intake is negatively associated with global brain volume measures, regional gray matter volumes, and white matter microstructure. Here, we show that the negative associations between alcohol intake and brain macrostructure and microstructure are already apparent in individuals consuming an average of only one to two daily alcohol units, and become stronger as alcohol intake increases.

Advances in molecular genetics have led to the exponential growth of the direct-to-consumer genetic testing industry, resulting in the assembly of massive privately owned genetic databases. This article explores the potential impact of this new data type on the field of marketing. Drawing on findings from behavioral genetic research, the authors propose a framework that incorporates genetic influences into existing consumer behavior theory and use it to survey potential marketing uses of genetic data. Applications include business strategies that rely on genetic variants as bases for segmentation and targeting, creative uses that develop consumers’ sense of community and personalization, use of genetically informed study designs to test causal relations, and refinement of consumer theory by uncovering biological mechanisms underlying behavior. The authors further evaluate ethical challenges related to autonomy, privacy, misinformation, and discrimination that are unique to the use of genetic data and are not sufficiently addressed by current regulations. They conclude by proposing an agenda for future research.

Previous research points to the heritability of risk-taking behaviour. However, evidence on how genetic dispositions are translated into risky behaviour is scarce. Here, we report a genetically-informed neuroimaging study of real-world risky behaviour across the domains of drinking, smoking, driving, and sexual behaviour, in a European sample from the UK Biobank (N = 12,675). We find negative associations between risky behaviour and grey matter volume (GMV) in distinct brain regions, including amygdala, ventral striatum, hypothalamus, and dorsolateral prefrontal cortex (dlPFC). These effects replicate in an independent sample recruited from the same population (N =13,004). Polygenic risk scores for risky behaviour, derived from a genome-wide association study in an independent sample (N =297,025), are inversely associated with GMV in dlPFC, putamen, and hypothalamus. This relation mediates ~2.2% of the association between genes and behaviour. Our results highlight distinct heritable neuroanatomical features as manifestations of the genetic propensity for risk taking. One Sentence Summary: Risky behaviour and its genetic associations are linked to less grey matter volume in distinct brain regions.

Aging has profound effects on how we consume, yet marketers often simplify aging to the mere passage of time (chronological age; ChronAge). This simplification overlooks fundamental biological and psychological aspects of the aging process and neglects meaningful variation within ChronAge segments. This paper highlights the need for multidimensional approaches to studying aging consumers and pioneers a framework that integrates biological aging into consumer behavior theory. We introduce a measure of biological age (BioAge), derived from epigenetic data, to the field of marketing. Using Health and Retirement Study data, we empirically demonstrate that BioAge captures distinct aspects of aging, not reflected in ChronAge, and improves prediction of consumer activities and spending. BioAge is most useful for predicting technology use, social engagement, physical activity, and discretionary spending on entertainment. Further, we show that BioAge mediates the impact of time on consumer behavior and operates through objective physiological pathways, such as chronic disease, cognitive functioning, and physical vitality, rather than subjective perceptions of health. We conclude with implications and best practices, including how BioAge can improve segmentation and customer lifetime value optimization, along with the ethical and legal challenges of its use in research and practice.

Digital advertising success hinges on effective creative design, yet identifying optimal combinations of design elements from millions of possibilities remains a costly challenge. Traditional A/B testing of a few options chosen by designers leaves the vast majority of the design space unexplored, potentially missing top-performing ads. We introduce Automatic Bayesian Combinatorial Design (ABCD), a novel AI-driven framework that unifies advertising creative generation and performance testing. ABCD uses Bayesian neural networks with active learning to iteratively design and test batches of creatives, learning from each round to systematically explore the design space and generate increasingly effective ads. We validate the framework through simulations and a large-scale field experiment with a premium shoe brand. The results show that ABCD is significantly more effective than current best practices. In a final validation test, creatives identified by a model trained with ABCD achieved an average click-through rate 48% higher than those from a model trained with Thompson Sampling (3.7% vs. 2.5%). Furthermore, the ABCD-trained model outperformed the benchmark on 5 out of 7 product lines. Our framework offers managers a practical method to enhance creative effectiveness at scale, while providing a significant methodological advancement for high-dimensional combinatorial optimization in marketing.

Recent advances in marketing have shown that both product images and neural measures can be used for aggregate demand forecasting. However, image-based machine learning models often function as black boxes. Similarly, neuroforecasting demonstrates predictive power without explaining what neural signals reflect. This paper bridges these approaches by introducing a framework that identifies interpretable visual features underlying neural valuation signals that forecast demand. To identify these visual features, we develop MotorVAE, a custom variational autoencoder that learns disentangled, interpretable representations from vehicle images. Our key methodological contribution is anchoring machine-learned visual features to independent neural data: we regress neural activity measured while participants view vehicles onto MotorVAE latent dimensions, identifying a subset of visual features that covary with neural valuation signals known to predict aggregate demand. These neuro-anchored features are translated into human-interpretable design elements using latent space traversals and validated through consumer judgments. Empirically, including neural data substantially improves demand prediction relative to survey-based models alone. Overall, this work moves demand modeling from prediction toward explanation, offering a transparent method for identifying which visual product attributes matter for markets.