Chamada de Artigos para Dossiê "Generative Artificial Intelligence, Synthetic Data, and Synthetic Respondents in Marketing and Market Research"

Decision-oriented empirical research in marketing and management is undergoing a structural transformation. Advances
in generative artificial intelligence, particularly large language models (LLMs), have opened the possibility that empirical
evidence itself can be partially constructed, expanded, or stress-tested by generating valid and meaningful synthetic
data and synthetic respondents. These developments promise substantial gains in speed, scale, and cost efficiency in
obtaining useful data, while enabling empirical inquiry into populations that are rare, sensitive, geographically dispersed,
or otherwise difficult to access.
Importantly, the diffusion of synthetic data practices is not confined to academic experimentation. Commercial research
providers and analytics platforms are increasingly incorporating synthetic personas (respondents), hybrid samples, and
automated iteration routines into their workflows, often blending human and AI-generated data in ways that are not
yet well understood or consistently governed (Argyle et al., 2023; Bisbee et al., 2024). While computational simulations
have long been discussed in the methodological literature as tools for theory testing and robustness analysis (Davis et al.,
2007; Leavitt et al., 2021; Olenick & Outland, 2025), the rise of generative models fundamentally alters the nature of this
discussion. Unlike parameter-driven simulations grounded in explicitly defined population models, LLM-based systems
can generate data through probabilistic architectures trained on vast, opaque datasets, introducing new intellectual and
organizational challenges, such as assessing the validity and reliability of the data (Bisbee et al., 2024).
The central question facing scholars and scientists, therefore, is no longer whether synthetic data and respondents will
be used, but under what conditions they are appropriate, for which research purposes, with what risks, and under which
standards of validation, reliability, transparency, and accountability. These questions are not merely technical. They are
deeply organizational and institutional, shaping how evidence is obtained and evaluated, how decisions are justified, and
how responsibility for knowledge claims is distributed and confirmed within organizations and research ecosystems.
This Special Issue of the Brazilian Administration Review invites contributions that treat synthetic data and synthetic
respondents as a joint challenge and problem of theory, methods, governance, and, ultimately, acceptance. Beyond
consumer-level applications, we particularly encourage research that examines how executives, managers, analysts, and
academic scholars assess the credibility of AI-generated or AI-augmented evidence, how such assessments interact with
organizational risk, compliance, and accountability structures, and how they ultimately influence strategic and tactical
decision-making.
Recent work in marketing and consumer research emphasizes that LLM-generated ‘silicon samples’ may be most
appropriate for exploratory and preparatory research stages but pose substantial risks to confirmatory inference and subgroup
analysis if used uncritically (Sarstedt et al., 2024). Furthermore, recent research demonstrates that while LLM-generated
data may approximate central tendencies observed in human samples, it likely exhibits reduced variance, sensitivity to
AI prompt design, instability over time, and systematic distortions across subgroups, and therefore raises concerns for
inference, generalization, and reproducibility (Bisbee et al., 2024; Dominguez-Olmedo et al., 2024). At the same time, work
on technological agency and perceived controls to ensure data quality suggests that acceptance of algorithmic outputs
depends not only on statistical performance but also on users’ perceptions of transparency, participation, and controllability
in the data generation process (Sundar & Marathe, 2010). Parallel debates are also emerging around disclosure norms and
responsible use of generative AI in academic and competitive research, as universities, organizations, and journals struggle
to define consistent expectations for transparency and accountability (Hair & Sabol, 2024). An improved understanding of
when and how these technical and perceptual dimensions jointly shape organizational adoption is, therefore, a central
objective of this Special Issue.