Traditionally, technology development has typically revolved around the functionality, usability, efficiency and reliability of technologies. However, AI technology needs a broader discussion on its societal acceptability. It impacts on moral (and political) considerations. It shapes individuals, societies and their environments in a way that has ethical implications.

The interpretation of ethically relevant concepts can change with technologies (consider what “privacy” meant before social media). Furthermore, when new technologies are introduced, users often apply them for purposes other than those originally intended. This reforms the ethical landscape, and forces us to reflect and analyze the ethical basis of technology over and over again.

Ethical frameworks are attempts to build consensus around values and norms that can be adopted by a community – whether that’s a group of individuals, citizens, governments, businesses within the data sector or other stakeholders.

Various organisations have participated in developing an ethical framework for AI. Naturally, their views differ in some respects, but there’s also been an emerging consensus to them. According to a recent study (Jobin et al 2019), AI ethics has quite rapidly converged on a set of five principles:

• non-maleficence
• responsibility or accountability
• transparency and explainability
• justice and fairness
• respect for various human rights, such as privacy and security

The five principles of AI ethics answer different questions and focus on different values:

1. Should we use AI for good and not for causing harm? (the principle of beneficence/ non-maleficence)
2. Who should be blamed when AI causes harm? (the principle of accountability)
3. Should we understand what, and why AI does whatever it does? (the principle of transparency)
4. Should AI be fair or non-discriminative? (the principle of fairness)
5. Should AI respect and promote human rights? (the principle of respecting basic human rights)

The rest of this course will focus on these principles of AI ethics. We will analyze what these concepts imply and how they can be interpreted, in the fashion of traditional philosophy: concept analysis. We will also look at how these concepts are being applied in practice, discuss their problems and mention some open questions regarding these principles.

In the last section of the course, we will look at the project of AI ethics as a whole. We will be asking the “cui bono” question: who is AI ethics for, and who or what is left out?

Lastly, we want to note that when speaking of AI and the social implications, AI ethics is the first on the list. But there are other theoretical frames for looking at ethical codes for algorithmic, data-driven systems. For example, questions of the social implications of AI come up in fields like algorithmic cultures, gender studies and media studies, amongst numerous others. Correspondingly, the cognitive and psychological aspects of human-machine interaction shapes the question of appropriate ethical framework for AI. Simply, there is a lot more to AI ethics than just data or algorithm ethics.