Using synthetic data to evaluate consumer attitudes / Maria Petrescu.

Format:

Book

Author/Creator:

Petrescu, Maria, author.

Language:

English

Subjects (All):

Consumer behavior--Research--Methodology--Case studies.

Consumer behavior.

Artificial intelligence--Marketing applications--Case studies.

Artificial intelligence.

Physical Description:

1 online resource

Place of Publication:

London : SAGE Publications Ltd, 2026.

Summary:

This case explores the methodological potential and challenges of using synthetic (silicon) data in marketing research. Specifically, it examines whether large language models (LLMs), such as ChatGPT 5 and Gemini 2.5, can generate samples that approximate human responses to surveys about consumer attitudes toward online reviews. The study compares a dataset of 505 U.S. consumers recruited via Qualtrics with synthetic datasets produced through iterative prompting of LLMs. The prompts varied in specificity, ranging from simple sample generation to more detailed instructions designed to mirror the demographic and attitudinal distribution of the original survey.The datasets were analyzed using multivariate analysis of variance (MANOVA) to assess differences in both demographic characteristics and consumer constructs, including suspicion of ulterior motives and the use of online reviews in decision-making. Results revealed significant differences across data sources: ChatGPT occasionally produced patterns close to the Qualtrics sample, while Gemini often generated implausible or extreme outputs. Within-model inconsistencies across runs further highlighted the sensitivity of synthetic data to prompt design and model variability.The findings suggest that while synthetic data cannot replace human samples, they hold value as methodological complements. Silicon samples can accelerate pretesting, instrument refinement, and hypothesis development. However, ethical and epistemological considerations, including transparency, bias, and reproducibility, remain central. This case offers practical lessons on how to generate, evaluate, and responsibly apply synthetic data in consumer and marketing research.

Notes:

Description based on XML content.

ISBN:

9781036245207

OCLC:

1594889123

Publisher Number:

T301349