Worldwide AI Ethics: A Review of 200 Guidelines and Recommendations for AI Governance

Developer Perspectives on Potential Harms of Machine Learning Predictive Analytics in Health Care: Qualitative Analysis

“…Additional studies push for more attention to organizational context [46,47], while the study by de Ágreda [58], which compared acceptance of ethical codes designed to govern development of AI technologies for use in military versus nonmilitary settings, concluded that an important characteristic for success of transferability of codes across settings is the way individuals understand the work an algorithm-driven tool is meant to accomplish. This idea is echoed in other studies cited here and suggests that future research might benefit the field by focusing on ML developers [17,24,56,59]. Unlike prior interest in individuals, directed toward assessing their knowledge of an ethics code's content [13], these proposals draw attention to the possible benefit of examining developers' understanding of the relevance of ethics to their daily work.…”

“…The effects of codes of ethics on conduct are less difficult to assess, but the results are not encouraging [55]. Of the recent studies conducted about codes of ethics in computer science and engineering, the positive effects of the ethics codes on conduct are rare [13,17,24,56]. Research examining the Association for Computing Machinery's code reported that having participants consider the code when making design decisions had "no observed effect" when compared with a control group [13].…”

“…Concerns about MLPA for health care began to garner interest within computer science in the mid-2010s [19,20] and captured public attention a few years later with the publication of studies demonstrating racial bias in algorithms used in criminal justice systems, health care services, and credit agencies [10,[21][22][23]. A steady stream of proposals to better manage artificial intelligence (AI) tools for social benefit has continued since, with a recent publication identifying 200 studies from the past decade offering guidelines to oversee the technology's development [24]. The volume of publications evinces the persistence of social concern about potential harms resulting from the rapid spread of AI tools but also the challenge their management poses.…”

United Kingdom ∙ Striking a Balance: UK's Pro-Innovation Approach to AI Governance in Light of EU Adequacy and the Brussels Effect

“…As Nicholas Kluge Corrêa and colleagues identified 17 resonating principles, such as accountability, privacy, and safety, there is hope that these principles can streamline and consolidate global AI governance efforts. 22 Clement-Jones expressed that by focusing the Summit on existential risks posed by AI, Summit attendees were not concentrating enough on the risks the world is currently facing. 23 From a stakeholder perspective, the Summit has been highly successful, bringing government representatives from the US and China together to the table as well as executives of technology companies.…”

The Ethical Risks Posed by New Technologies in Research

“…There is no shortage of documents that purport to provide governance for AI usage in a large variety of domains, created by a large variety of private and public organizations. A recent meta-analysis discovered 200 such documents, warning that, in actuality, this is a small dataset, as many more documents are available in many languages around the world (Corrêa et al, 2023). Interestingly, most of these guidelines converge, according to the meta-analysis cited, on five aggregated principles: transparency/ explainability/ auditability, reliability/ safety/ security/ trustworthiness, justice/ equity/ fairness/ non-discrimination, privacy and accountability/ liability.…”