Selected works
Zhong, Z., “From Privacy Washing to Sustainable Data Strategies: An AI-Enabled Large-Scale Investigation”, AMA 2026, ISMS 2025, EMAC 2025, preparing for submission to Information Systems Research (UTD24, AJG/ABS: 4*) [Full Paper]
Abstract: We first developed analytical models showing that firms can enhance profits and consumer surplus by adopting sustainable data strategies. To validate the model and enable scalable policy measurement, we introduced PrivaAI, an AI- based agent with a novel framework derived from regulation, literature, expert input, and human ratings. PrivaAI, leveraging structured chain-of-thought reasoning and enhanced retrieval-augmented generation, was fine-tuned and tested on 28,360 human-labeled evaluations (1,418 policies) to align with consumer perceptions by 92.5%. Applying it to policies from 10 million websites classified by Open PageRank and GICS sectors, we find: (i) lower-ranked providers systematically underperform, consistent with the model’s assumption; (ii) marked industry heterogeneity; and (iii) regulatory asymmetries—under weaker regulation (CCPA), higher-ranked firms often show uneven improvements suggesting privacy washing, whereas stronger regulation (GDPR) yields broader, more balanced improvements. A pre-registered lab experiment with 610 US-representative participants further shows that PrivaAI evaluations explain intentions regarding usage, data sharing, and trust, while validating that the features it flags as suggesting privacy washing align with consumer perceptions. Together, these findings provide a scalable foundation and actionable insights for benchmarking practices, detecting privacy washing, and guiding firms and regulators toward sustainable data strategies that reinforce long-term trust and social responsibility.
Zhong, Z., X. Li, and S. Yiu, “One Consumer, Multiple Locations: Geographic Targeting with Registration Data under Privacy Constraints”, preparing for submission to Marketing Science (UTD24, AJG/ABS: 4*) [Full Paper]
Abstract: Geographic targeting often treats location as a single field. We show why that is no longer enough when platforms observe several plausible locations but privacy regulation makes some costly to process. Using 110,326 transactions from a leading Chinese live-event platform, including 63,559 verified orders linking national-ID origin, mobile-number locality, and event city, we estimate a location-prior hierarchy. Mobile locality is closer to consumption than national-ID origin in 58.5% of transactions, yet 45.0% of mobile-event distances exceed 500 kilometers. Origin-mobile divergence predicts purchases outside both registered locations, not broader distance demand, and adds little genre information after destination-market structure. In ranking audits, behavioral trajectories dominate registration signals (.898 vs. .634 AUC). A catalog-based conditional logit over feasible listed events validates the hierarchy without using the decoupled-market outcome: mobile distance raises holdout top-decile hit from 54.0% to 69.8%, while adding origin reaches 70.2%. The operational rule is clear: start with mobile, separate spatial from taste ranking, and add origin only when segment-level lift clears its privacy cost.
Yang, Z and Z. Zhong*, “Public Data Infrastructure and the Distribution of IT Business Value: Evidence from China's Big Data Pilot Zones ”, under review at Information Systems Research (UTD24, AJG/ABS: 4*) [Full Paper]
Abstract: Can government-led data factor market policies enhance firm productivity, and if so, do they benefit technology leaders or digital laggards? We address this question by exploiting China's National Big Data Comprehensive Pilot Zones (NBDCPZ) as a quasi-natural experiment. Applying difference-in-differences estimation to 3,666 listed firms across 251 cities (2010--2021), we find that NBDCPZ designation increases firm-level total factor productivity by 5.92%, with instrumental variable estimates confirming robustness. Mechanism analysis identifies firm-level digital transformation---measured through textual analysis of annual reports---as the critical channel. Crucially, heterogeneity analyses reveal a “leveling-up”pattern: productivity gains are broad-based across technology intensities and firm sizes, with non-state-owned and service-sector firms benefiting most. These findings suggest that data factor market policies function as inclusive digital infrastructure that democratizes access to data capabilities, thereby narrowing rather than exacerbating the digital divide. Our study contributes to the emerging literature on the business value of data as a production factor and offers policy implications for designing inclusive digital transformation initiatives.
Zhong, Z., “Data Decay in AI-Driven Marketing: Strategic Mimicry, Aggregation Failure, and Platform Countermeasures”, under review at Journal of Academy of Marketing Science (FT50, ABDC: A*) [Full Paper]
Abstract: Marketing strategy treats behavioral data as a strategic asset that appreciates with scale. We show that this premise can invert when consumer-deployed AI agents adapt strategically to platform classifiers. In a dynamic model of strategic mimicry, accumulated behavioral traces depreciate, depreciation accelerates in AI-agent share, and data can reach a full-mimicry region in which behavioral targeting loses informational value. The platform's optimal response shifts across admission control, data provenance, and accelerated retraining as AI-agent share rises. Calibrating the model on 110,326 transactions from a major Chinese ticketing platform, a pre-rollout machine-speed trace proxy implies 59% data-value erosion (USD 1.6-4.9 million per year), while mandatory identity provenance reduces proxy-calibrated erosion to 16% (USD 0.1-0.4 million). The framework extends marketing's strategic-asset tradition to AI-mediated exchange and gives managers, regulators, and scholars a decision rule for protecting data value.
Zhong, Z., “Where Synthetic Consumers Fail: Evidence from Four Frontier LLMs and a Large-Scale Ticketing Platform”, under review at International Journal of Research in Marketing (AJG/ABS: 4*, ABDC: A*) [Full Paper]
Abstract: Can synthetic consumers predict what customers will buy and what they do after buying when a firm knows only basic profile information? We test this question with four frontier LLMs, 86,400 real API predictions, and 110,326 transactions from a rare live-event ticketing platform dataset. The answer is limited. The models recover some purchase preferences, especially after seeing a short purchase history. But they fail in three ways. First, aggregate accuracy hides large segmentlevel misses: the models perform well in the platform's native scene but poorly for a large segment outside that scene. Second, they struggle to predict whether buyers actually attend after purchase, even after examples of buyers who did and did not attend. Third, they often disagree about the same customer: in 70.5% of cases, the four models choose different top events. Local-language prompts reduce, but do not remove, these segment gaps. For marketing research, the implication is simple: synthetic consumers can help screen which preferences to test, but they should not replace observed customers when decisions depend on realized behavior or cell-level validity.
Zhong, Z., M. Wang, and S. Yiu, “The Pause That Wasn't Hesitation: Validating What Platform Traces Really Mean”, under review at Journal of the Association for Information Systems (AJG/ABS: 4*, ABDC: A*) [Full Paper]
Abstract: Platform logs increasingly convert small traces into large behavioral claims: a pause becomes hesitation, dwell time becomes interest, a slow reply becomes uncertainty. Yet logs record system events, not mental states. This paper develops an evidence-to-permission framework for deciding when platform traces can support behavioral inference, theory, and automated action. The framework requires analysts to establish three things before a trace is named or deployed: the population the log actually observes, the behavioral domain the trace can support, and the use the evidence warrants. We test the framework on a unique proprietary dataset from a leading Chinese live-experience ticketing platform: 108,924 completed transactions, plus public checks covering 17,780 Stack Exchange questions, 1,590 GitHub issues, and 12,330 e-commerce sessions, including 10,422 non-purchase sessions. In the focal case, pre-payment hesitation time (PPHT) appears to measure checkout hesitation but collapses under validation: PPHT-only prediction is near chance, residual PPHT does not predict no-show, account structure explains 72.1% of variance, and repeated accounts are stable (ICC = 0.464). PPHT is not a useless trace; it is a misassigned trace. Conditional on payment completion, it warrants account-keyed latency, not transaction-level hesitation. Contrast cases show the framework does not merely reject traces: advance-purchase timing supports event-planning inference, check-in timing supports arrival-process inference, and full-funnel e-commerce data show how conclusions change when non-purchase sessions are observed. The broader claim is simple and consequential: platform traces become useful only when evidence assigns what they mean, whom they describe, and what they may legitimately do.
Zhong, Z. and X. Li, “Re-Visiting the Green Puzzle: The Effect of Eco-Positioning on Service Adoption”, ISMS 2024, EMAC 2024, invited resumission at International Journal of Research in Marketing (AJG/ABS: 4*, ABDC: A*) [Full Paper]
Abstract: Service providers are increasingly employing eco-positioning strategies to promote sustainable products. However, these strategies’ effectiveness in driving service adoption, particularly among consumers with varying levels of inertia, remains unexplored. Drawing on the elaboration likelihood model, this research investigates the differential effects of eco-positioning on inertial and new consumers in energy service adoption. In five studies, including a large-scale study of U.S. households’ energy consumption (N = 15,568) and four experimental studies (N = 1,078), we document a consistent inertia utility in green service adoption ($7.69 in the field, $7.44 in the lab), demonstrating that eco-positioning is more effective in increasing service adoption intentions among inertial consumers than new consumers. Through experiments and topic modelling (BERTopic, LDA), we find that eco-positioning more effectively elicits a warm glow and reduces inertial consumers’ service concerns while the sustainability liability effect has a stronger negative impact on new consumers. Through a novel discrete choice model, we quantify that a 3.12% claimed reduction in emissions is equivalent to a $1 incentive in motivating inertial consumers. These findings provide actionable guidance for service providers and policymakers in tailoring eco-positioning intensity to consumers’ inertia utility levels and leveraging quantitative benchmarks to optimise resource allocation between environmental and monetary incentives.
Zhong, Z.*, Z. Zhou*, N. Sullivan, and V. Mak, “(Green) Attention is All You Need: A Dynamic Salience Model of Sustainability Advertising—Theory and Evidence”, *Co-first author, AMA 2026, targeting Journal of Marketing Research (UTD24, AJG/ABS: 4*)
Abstract: This paper identifies the “Green Regret” phenomenon: sustainability advertising temporarily increases consumer valuation but generates post-purchase dissatisfaction as attention shifts. We develop marketing's first dynamic salience model, demonstrating how advertising-induced attention shifts drive temporal preference instability. Study 1 confirms that sustainability advertising raises both purchase intention and return likelihood. Studies 2a and 2b (preregistered) employ lab experiments with attention measures, revealing that attention significantly mediates sustainability advertising's effect on valuation—an asymmetry absent for design advertising. Study 3 (preregistered) use eye-tracking and incentive-compatible BDM auctions to test whether post-purchase reminders can sustain attention and mitigate regret. Our findings reveal that attention constitutes a fundamental mechanism driving sustainability advertising effectiveness, with critical implications for sustainable marketing practice.
Ye, H.*, S. Chen*, Z. Zhong, C. Xiao, H. Zhang, Y. Wu, and F. Shen (2026), “Seeing through the Conflict: Transparent Knowledge Conflict Handling in Retrieval-Augmented Generation”, *Co-first author, Proceedings of AAAI Conference on Artificial Intelligence (AAAI) (CORE rankings: A*, flagship AI conference) [Full Paper]
Abstract: Large language models (LLMs) equipped with retrieval---the Retrieval-Augmented Generation (RAG) paradigm---should combine their parametric knowledge with external evidence, yet in practice they often hallucinate, over-trust noisy snippets, or ignore vital context. We introduce TCR (Transparent Conflict Resolution), a plug-and-play framework that makes this decision process observable and controllable. TCR (i) disentangles semantic match and factual consistency via dual contrastive encoders, (ii) estimates self-answerability to gauge confidence in internal memory, and (iii) feeds the three scalar signals to the generator through a lightweight soft-prompt with SNR-based weighting. Across seven benchmarks TCR improves conflict detection, raises knowledge-gap recovery by +21.4% and cuts misleading-context overrides by -29.3%, while adding only 0.3% parameters. The signals align with human judgements and expose temporal decision patterns.