Highlights of the VEIL

In this Virtual Ethical Innovation Lecture, Bernd C. Stahl, Professor of Critical Research in Technology at the University of Nottingham, previously De Montfort University, presented insights from his research on AI ecosystems. He is the author of the open access book “Artificial Intelligence for a Better Future – An Ecosystem Perspective on the Ethics of AI and Emerging Digital Technologies”

Bernd C. Stahl commences by introducing the SHERPA Project, its 10 case studies of current smart information systems (SIS) spread over a variety of fields (among others: IoT, government, smart cities, science, manufacturing, etc.) as well as five scenarios of future SIS (predictive policing, warfare, mimicking technologies, education, self-driving cars). His presentation focuses on the challenges and insights from the project and outlines the path towards an ecosystem of AI for human flourishing.

Five Challenges of Smart Information Systems

Bernd C. Stahl identifies five challenges of SIS:

1. The first challenge consists in agreeing on the ethics and human rights implications of SIS, which, in his research, involved the analysis of ethical impacts and social tensions of the case studies by identifying what people are concerned about.
2. The second challenge deals with the acknowledgement of the range of different perceptions and viewpoints. Within the project, several stakeholders and experts had been interviewed, surveyed and studied to collect competing interests.
3. The third challenge comprised the identification of possible approaches to solutions, such as guidelines for research and innovation of SIS, regulatory options, the assessment of the potential of standardisation, technical options and interventions, and terms of reference for regulators.
4. As the fourth challenge, these approaches needed to be tested and validated through stakeholder consultation to develop a prioritisation and finalized recommendations.
5. Eventually, challenge five comprises the implementation in terms of the dissemination, exploitation and advocacy of outcomes. In this context, Bernd C. Stahl highlights how he reached out to, e.g., policy makers to understand their needs for making proposals in current debates.

The Field of AI and Ethical Issues

Bernd C. Stahl emphasizes that the many attempts to define AI may not be helpful from the point of view of ethics. Instead, he proposes to adopt a perspective of AIs intended purpose, which may be found in terms of increasing efficiency, further optimization, maximizing profits as well as social control. Bernd C. Stahl identifies a broadly embedded ethical perspective in terms of AI contributing to human flourishing, which he regards as neither necessarily in conflict or congruence with the aforementioned purposes.

Project SHERPA also involved literature research to identify both academic as well as vernacular concepts of AI. In academic terms, it turned out that AI concepts feature no common core, but rather have a broad family resemblance. In terms of what people usually refer to as AI, both narrow and general AI were identified to raise rather disjoint sets of ethical considerations, while a more integrated socio-technical perspective was also identified, in which AI methods converge with contextual applications that result in ethically relevant impacts (such as manipulation or undermined autonomy). Within this perspective, questions about living in a digital world take center stage in the ongoing ethical discourse.

Potential solutions to the domain of narrow AI specific issues have been related to particular technical, or legal remedies (such as addressing explainability or algorithmic biases), while solutions to issues in the broader socio-technical domain are rather political in nature. Questions about general AI, in turn, are fundamentally related to philosophical questions.

Some explicit mitigation measures on the regulatory and policy level include human rights law, clarifications of liability, or the introduction of a new regulator. On the organizational level measures like risk management, impact assessments, quality assurance, data governance and Corporate Social Responsibility (CSR) can contribute. Some technical measures are data protection, standardisation, design guidelines, etc. In addition, on the individual level ethics frameworks, professionalism, and rigorous methodologies are avenues to address ethical issues. Bernd C. Stahl emphasizes that the vast array of options creates a complex and overlapping field.

Findings and Insights — An Ecosystem Approach

The SHERPA project provided one overarching insight: There is no magic bullet, but rather, what is required, is an intelligent mix of options that provides answers to the following questions: What to do about this epistemic complexity? Who is responsible for what? How to govern increasingly rapid technological progress? Bernd C. Stahl argues that in order to appropriately address such complex settings, the metaphor of an “ecosystem” is useful. In this sense, an ecosystem is characterized by its unclear boundaries and an inherent openness to new influences and actors. The metaphor stems from the notion that evolution acts as a driver of change, where interdependent actors co-evolve and benefit from mutual learning and competition, both of which marking a complex and potentially dynamic relationship between them.

Within this construct, supporting human flourishing benefits from adhering to the following requirements for interventions:

• Interventions should be clear about the geographical, conceptual and technical delimiters of the ecosystem at stake.
• Interventions should consider the knowledge that has to exist in the ecosystem, the technical, conceptual and procedural knowledge base and what capacity building is necessary.
• Interventions should be founded on appropriate governance structures, i.e., they need to be adaptive, flexible, there needs to be an ability to learn and there needs to be a higher level of responsibility focused on synergies instead of conflicts.

For each of these three requirements, the protagonists of the SHERPA project devised a set of recommendations. Please consider the open access book “Artificial Intelligence for a Better Future – An Ecosystem Perspective on the Ethics of AI and Emerging Digital Technologies” for further details.

Conclusion

Bernd C. Stahl concluded by highlighting the importance of the recommendations for a European AI ecosystem that contributes to human flourishing. In addition to the conceptual aspects, Bernd C. Stahl also briefly mentioned a set of case studies published in open access format, meant to provide details about what the implementation of the recommendations might look like in practice.

Q&A

The question and answer session first revolved around the appropriateness of the concept of an ecosystem (e.g., as opposed to the concept of an “innovation ecology”). Part of the answer was comprised by the insight that ethical issues with AI can only rarely be addressed at the lower level, but that, e.g., legal or economic implications also ensue. Further questions tackled approaches — or the difficulty thereof — to measure the effects of capacity building as well as the underlying notion of ethics within the conceptual approach that Bernd C. Stahl presented.