Mark Ryan and Bernd Carsten Stahl, "Artificial intelligence ethics guidelines for developers and users: clarifying their content and normative implications"

Bibliographic info

Ryan, M., & Stahl, B. C. (2021). Artificial intelligence ethics guidelines for developers and users: clarifying their content and normative implications. Journal of Information, Communication and Ethics in Society, 19(1), 61–86. https://doi.org/10.1108/JICES-12-2019-0138

Commentary

What I found most valuable about this paper is that it moves beyond discussing abstract ethical principles and instead examines what existing AI ethics guidelines actually expect developers and organizations to do. By analysing 91 different guidelines, the authors identify recurring themes and translate them into practical recommendations. This makes the paper especially useful because it connects broad concepts such as fairness, transparency and accountability with concrete responsibilities for those designing and deploying AI systems.

I also liked that the paper distinguishes between AI developers and AI users. Ethical responsibility does not end once a system has been built; organizations that adopt AI also have obligations regarding oversight, monitoring and responsible use. This perspective aligns well with topics we discussed throughout the Digital Ethics course, particularly the importance of governance and accountability across the entire lifecycle of AI systems.

One limitation I noticed is that the paper intentionally maps existing guidelines rather than evaluating them. As a result, it presents a comprehensive overview of ethical recommendations but offers little guidance on how to deal with situations where principles conflict. For example, maximizing transparency may reduce privacy, while improving explainability can sometimes come at the expense of predictive performance. The paper successfully identifies these principles, but it leaves the difficult task of balancing them largely to practitioners.

Excerpts & Key Quotes

Mapping existing guidelines rather than proposing new ones

• Page 62

"It must be made clear here that we are not providing prescriptive recommendations, but rather, are mapping the prescriptive recommendations found in these guidelines."

Comment

I think this quote is important because it clarifies the purpose of the paper from the beginning. The authors are not arguing that every recommendation is necessarily correct or sufficient. Instead, they analyse what current AI ethics guidelines have in common. This makes the paper a useful reference for understanding the current state of AI ethics, while also reminding readers that broad agreement on ethical principles does not automatically solve ethical dilemmas in practice.

Black-box AI in high-stakes domains

• Page 66

"High-stake domains (such as health care, criminal justice and welfare) should reconsider using black-box AI altogether (AI Now Institute, 2017)."

Comment

This passage stood out to me because it illustrates how demanding some AI ethics guidelines can be. In high-stakes settings, the consequences of an incorrect decision can be significant, making explainability-A and human oversight especially important. While I agree that transparency should be prioritised in these contexts, I also think the issue is more nuanced. In some situations, more complex models may provide substantially better performance, so the challenge is not simply choosing between explainability and accuracy, but deciding when the benefits of one justify sacrificing some of the other.

The responsibility gap

• Page 71

"Moral responsibility is a very important issue within AI ethics, with a fear that companies will try to obfuscate blame and responsibility onto the autonomous or semi-autonomous system. There may also be incidences where because of this relative autonomy, AI creates a 'responsibility gap', whereby it is unclear who is responsible."

Comment

This was probably the section that interested me the most. As AI systems become more autonomous, it becomes easier for organizations to attribute mistakes to the technology rather than to the people who designed, deployed or managed it. I agree with the authors that this "responsibility gap" should never be accepted as an excuse. AI systems are tools, and responsibility should always remain with individuals or organizations. This idea connects closely with the discussions we had in class about accountability, traceability and responsible AI governance.

accountability
transparency
explainability-A
explainability-B
fairness
responsibility
traceability
governance