Aims
This course aims to:
• introduce you to the essence and the social impact of Big Data and AI
• provide you with knowledge and understanding of the fundamental ethics principles that underpin the governance of Big Data and AI
• enable you to reflect on, reason about and reconcile tensions among risks and benefits of Big Data and AI through the application of these fundamental ethics principles
Content
This course provides an introduction to the main ethical implications of Big Data and AI, and to the fundamental ethics principles for their governance. The widespread use of Big Data analytics and AI in various social domains has surfaced a range of attendant risks and benefits. Existing and newly adopted regulations are unable to fully address and balance the prospects of harm with those of benefits. This course explores the role of ethics in this endeavour. We will first examine the essence of Big Data and AI and the sociotechnical harms, risks and benefits associated with their use. You will be then introduced to the origins, meaning and role of ethics and to the key features of the main normative ethics theories. In a separate session, we will further discuss the significance, utility and limitations of ethics principles as a whole. You will next comprehend the essence of the fundamental ethics principles of human autonomy, non-maleficence, beneficence, and justice. In the domain of Big Data and AI, these principles have been further specified as the tenets of human agency and oversight, privacy, technical robustness and safety, social and environmental well-being, diversity, fairness and non-discrimination, transparency and accountability. By exploring these specifications, we will uncover the core of the fundamental ethics principles and the normative guidance that they provide in the areas of Big Data and AI. In the final session of the course, we will delve into the practical application of the principles through examination of specific frameworks and relevant methods for ethics-based governance of Big Data and AI.
Presentation of the course
D. Fessenko teaches the course by combining the traditional, podium-based format of lecturing with narrative, interdisciplinary and practice-oriented approaches to content delivery and knowledge development and retention. In particular, building upon real-life examples in an interactive way, we will delve into fundamental theoretical concepts and their practical application. Throughout each session, you will be invited to hone your critical thinking and problem-solving skills by engaging with case studies, specific questions and other tasks that would challenge your moral intuitions and perspectives. You will also have the opportunity to ask questions in the course of each session.
Course sessions
1. The essence of Big Data and AI, and the main associated risks and benefits
We will review the course program, objectives, and outcomes. You will then get acquainted with the essence of Big Data and AI, and the various conceptions of the latter as a technology, a form of agency, and social and business practices. We will also examine the main categories of sociotechnical harms, risks and benefits that Big Data analytics and the development and use of AI give rise to.
2. What is ethics and what role it has to play in governing Big Data and AI
This session will introduce you to the origin, the concept and the objectives of ethics, as well as to its significance in the governance of Big Data and AI. We will discuss the distinctions between right and wrong, and good and bad, and what moral judgements ultimately purport to achieve. You will be introduced to the key features of the main normative ethics theories of utilitarianism, deontology, virtue ethics and sentimentalism, and what pathways to ethical use of data and AI they have inspired.
3. What are ethics principles for?
During this session, we will explore the significance, utility and limitations of ethics principles as a whole. In particular, we will discuss how principles help rationalize, test, unify and systematize our moral judgements, and help achieve their ultimate objective(s). Ethics principles also have their limitations, such as the potential to lead to insolvable conflicts, a lack of a sufficient epistemic basis, and insufficient action guidance, which we will survey in the last part of the session. We will conclude the session with observations and contemplations on the advantages and constraints of using ethics as data and AI governance tools.
4. The principle of privacy and its demands for self-determination and self-governance
This session will first examine the essence and the difference between privacy and the right to privacy. We will then delve into the various conceptions of privacy and how they have evolved with the advent of Big Data analytics and AI. You will understand how privacy relates and enables autonomy, which we will discuss in the next session. We will approach all these aspects through the lenses of a specific case study from the healthcare domain.
5. The principle of human autonomy and its demands for human agency and oversight
We will discuss the core and the various philosophical conceptions of the principle of autonomy. The session will further tackle the meaning and logic behind one specific formulation of the principle – of human agency and oversight – in the areas of Big Data and AI given their impact on human ability for independent decision-making and self-governance. We will draw on specific examples from behavioural advertising and healthcare to illustrate the various points.
6. The principle of non-maleficence and its demands for technical robustness and safety
Harm prevention is the lead consideration and driver behind most governance efforts in the realm of Big Data and AI. This consideration emanates from the principle of non-maleficence or “do no harm”. We will explore its essence and postulates, and how they translate into the demands for technical robustness, reliability, safety and security given the harmful potential of data- and AI-driven technology solutions. Case studies from the domains of biosecurity and social media will help us highlight and clarify these various aspects.
7. The principle of beneficence and its demands for social and environmental well-being
This session will explore the meaning and various interpretations of the principle of beneficence. We will discuss what social and environmental wellbeing entails specifically in the context of data and AI. Examples from child care and education, as well as environmental protection, will illustrate the meaning and application of these ethical tenets.
8. The principle of justice and its demands for diversity, fairness and non-discrimination
The session will provide an overview of the main theories of justice and their conceptions of the principle. Drawing on real-life examples from data- and AI-enabled healthcare and recruitment, we will examine various formulations of justice as fairness and non-discrimination metrics in data analytics and machine learning, along with their broader implications. You will better understand the role of diversity (e.g. in data, teams, stakeholders) in ensuring equal regard to everyone’s needs and in overcoming structural disparities.
9. The demands of all four principles for transparency and accountability
We will consider another set of demands of the four fundamental principles, namely for transparency and accountability. We will discuss what they mean in practice. Examples from automated decision-making in banking and recruitment will illuminate the significance of transparency and accountability for the proper exercise of human agency, harm prevention and human thriving, responsibility and trust building.
10. Ethics-based governance of Big Data and AI in practice
In this final session of the course, we will explore the common frameworks and practices for implementing ethics-based governance of Big Data and AI, such as industry self-regulation, value- and principle-based state regulations, industry standardization, and corporate policies. As some of the key ethical considerations and principles may at times appear at odds with each other, the adoption and implementation of these forms of governance would require (some) reconciliation of the underlying moral tensions. We will look at established methods for this, such as value sensitive design, deliberative participatory co-creation, red teaming and iterative risk and impact assessments.
Learning outcomes
You are expected to gain from this series of classroom sessions a greater understanding of the subject and of the core issues and arguments central to the course.
The learning outcomes for this course are to:
• demonstrate your grasp of the essence and the ethical and social implications of Big Data and AI
• understand and be able to explain the core and significance of the fundamental ethics principles governing Big Data and AI
• be able to identify, introspect, reason about and reconcile tensions among risks and benefits of Big Data and AI through the application of these fundamental ethics principles
Required reading
The following publications and textbook are particularly helpful and thus required readings in preparation for the course.
OECD, OECD Recommendation of the Council on Artificial Intelligence, (2019, May 22)
Section 1: Principles for responsible stewardship of trustworthy AI https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0449
High-level Expert Group on AI, Ethics guidelines for trustworthy AI (2019, April 8)
https://digital-strategy.ec.europa.eu/en/library/ethics-guidelines-trustworthy-ai
Dignum, V, Responsible Artificial Intelligence: How to Develop and Use AI in a Responsible Way (Springer, 1st ed. 2019 edition, 2020) ISBN: 10303030373X or 9783030303730
