From AI ethics principles to data science practice: a reflection and a gap analysis based on recent frameworks and practical experience

Abstract

In the field of AI ethics, after the introduction of ethical frameworks and the evaluation thereof, we seem to have arrived at a third wave in which the operationalisation of ethics is central. Operationalisation is required, since ethics frameworks are often not suited to be used by data scientists in the development of AI-based services or products. Therefore, in this paper, we aim to contribute to this third wave by mapping AI ethical principles onto the lifecycle of an AI-based digital service or product and combining it with an explicit governance model to clarify responsibilities in operationalisation. We then discuss practical, conceptual, and political implications of this analysis to end with key challenges around operationalising AI ethics.

Data Ethics Emergency Drill: A Toolbox for Discussing Responsible AI for Industry Teams

“…Ethical literacy is a process, not an end goal and products should continually be monitored for value alignment. Similarly to Georgieva et al, we see the future of responsible AI as a "landscape of methods, standards and procedure" [23].…”

Artificial Intelligence for Surface‐Enhanced Raman Spectroscopy

“…As for security and privacy, this is a common challenge for a shared database with a large population or different data sources. [ 339 ] If the open‐source database on Raman big data is developed, especially one containing human sample test data, it is necessary to ensure the reasonable use of the data, the privacy of the people involved, or to counter the possible attack on the database. Reasonable solutions, such as defending and validating AI models against adversarial attacks, are expected to address growing trust concerns about AI, especially its security in mission‐critical applications.…”

Defining organizational AI governance

“…There is a growing body of research acknowledging the importance of governed AI. Georgieva and her colleagues [8] call this the "third wave of scholarship on ethical AI," which focuses on turning AI principles into actionable practice and governance. The third wave aims at promoting practical accountability mechanisms [24].…”

“…To reap the benefits and manage the risks, there is widespread consensus that AI systems need to be governed to operate in line with human and societal values [5,6]. However, current AI governance work faces the challenge of translating abstract ethical principles, such as fairness, into practicable AI governance processes [7,8]. In a global overview of AI governance, Butcher and Beridze [9] conclude that "AI governance is an unorganized area."…”

“…Second, we position AI governance as part of an organization's governance structure, together with corporate, IT, and data governance. In doing so, we advance the body of knowledge on implementing AI ethics (e.g., [7,8,13,23]) through AI governance (e.g., [12,18]). Our contributions clarify the significance of AI governance as part of organizational governance that helps align the use of AI technologies with organizational strategies and legal and ethical requirements coming from the operating environment.…”