The prevalent paradigm for understanding what constitutes a ‘good’ data visualization, and what we are likely teaching business students, relates to a conventional wisdom of efficiency, clarity, transparency, and faithful representation of truth. Teaching about the ethics of visualizations seems to be largely absent from business school curricula. This paper suggests that business students who make and use visualizations should be taught to use a code of ethics for visualizations based not only on conventional wisdom, but also on ideas from the Machine Learning/AI and ‘ethical visualizations’ communities. Elaborating on Sheppard’s code, the code suggested here incorporates ideas about bias, transparency, user agency, and privacy. Ethical awareness and practice become even more important as more data collection, processing, and visualization is shifted to Machine Learning and Generative AI. A threshold model is adapted to understand when visualization decisions require more scrutiny. Ideas from the code and the threshold model are used to analyze choices made by the CDC’s COVID-19 Rapid Response Team for an early pandemic visualization that circulated widely beyond the initial report in which it appeared. This and other examples underscore the need for business professionals to consider their visualizations from the perspective of the code.
Read full abstract