AMTP Proceedings 2026
Document Type
Conference Proceeding
Publication Date
April 2026
Abstract
Large-scale qualitative coding is a bottleneck in marketing research, especially for complex, multi-level phenomena like corporate cancel culture. This paper uses over 100 well-known cancellation events as a testbed to evaluate whether GPT can act as a reliable, low-cost “independent rater” for rich categorical schemes. We develop a multi-dimensional coding framework (trigger locus, controversy source, social issue, moral foundation, and target level) and implement a staged GPT pipeline: zero-shot prompts benchmarked against human “gold standard” coders, scaled to larger samples, then extended with few-shot prompting and fine-tuning for harder constructs such as moral foundations. We compare GPT variants and roles, treating models as additional coders rather than as research subjects. The contribution is methodological: we show how to decompose cancel culture into theoretically meaningful dimensions and how to harness GPT—carefully and transparently—to label them at scale, offering best practices for marketing scholars facing ever-growing textual datasets.
DOI
10.20429/amtp.2026.86
Recommended Citation
Abraham, Sairah and Poehlman, T. Andrew, "AI-Based Multi-Dimensional Coding of Cancel Culture: Opportunities and Limits of LLM-Based Categorization" (2026). AMTP Proceedings 2026. 86.
https://digitalcommons.georgiasouthern.edu/amtp-proceedings_2026/86
