New TIC taxonomy diagnoses 6 ways data narratives go wrong

A team of researchers led by Yu Fu from Georgia Tech has published a paper introducing TIC, a Taxonomy of Issues in Data Communication, designed to systematically diagnose how data narratives fail. While prior work has studied isolated errors—such as statistical fallacies, misleading charts, or flawed fact-checking—the community lacked a unified framework to show how these issues arise, propagate, and compound. To fill this gap, the team synthesized existing literature across statistics, visualization, and fact-checking, then qualitatively annotated 700 real-world data narratives from fact-checking sites, research datasets, and controversial media. The resulting taxonomy organizes recurring breakdowns along six dimensions: data, analysis, visual encoding, text, reasoning, and interpretation. These dimensions are placed within a process framework that covers three stages: analysis (how data is collected and processed), narrative construction (how findings are represented and argued), and audience reception (how readers interpret the narrative).

Alongside the taxonomy and framework, the researchers contribute a qualitatively annotated case corpus with coding justifications and an interactive browsing interface. The corpus allows practitioners to study real examples of each issue type, such as cherry-picked data points, inappropriate statistical tests, misleading axis scales, ambiguous language, faulty causal reasoning, and misinterpretation by audiences. By providing a structured lens, TIC enables journalists, data scientists, educators, and AI system designers to identify and prevent breakdowns in data communication. The work is particularly relevant as AI-generated charts and text increasingly shape public understanding. The paper is available on arXiv under the identifier 2607.10523 and includes 22 pages, 7 figures, and 2 tables.