Two summaries are presented here: 1) existing ethical codes and frameworks, and 2) high-level concepts listed in the literature.

The 11 codes listed were prepared by the Association for Computing Machinery (ACM), the United Kingdom (UK) Government Data Science Ethical Framework, Accenture, and similar organizations. Each code is described in a paragraph; for example, the Data Science Association’s Code of Professional Conduct “provides a comprehensive section on quality of data and evidence, with practical guidance for conducting data science projects. [...] However, the code lacks a wider consideration of how data science can cause harm.” From these ethical codes, nine key concepts are identified and grouped into three focus areas: general code of conduct, data-related challenges, and model-related challenges. An example of a concept in the third group is “subjective model design.”

The literature search came up with another set of concepts that generally overlapped the concepts in the codes. The concepts from the literature search fell into two groups: data-related challenges and model-related challenges. The concept noted in most articles was “personal and group harm,” a model-related challenge. The next most-noted concept was “privacy and anonymity.” The literature search found “newness of field” fairly often, a concept that did not appear in the codes.

The paper, with its list of 39 references, would certainly be useful for someone planning research or developing a course in data science ethics. It is accessible to a nonspecialist and encourages further reading.