Unit 1: AI Ethics

Unit 1 - AI Ethics


Goal of the unit: In this first unit of the module, we will explore some of the key ethical principles that emerge from modern AI systems, as well as the study of normative ethics in AI, from the perspectives of socio-technical researchers.


Learning objectives:

  • Explore the key ethical principles that are relevant to AI Ethics.

  • Review some examples of ethical challenges that emerge from real-world systems.

  • Examine the kinds of impacts that AI systems can bring about at different levels of society.

  • Become familiar with the types of technical research on Ethical AI.


Summary

With AI playing a driving role in the Fourth Industrial Revolution and its visible influence in our everyday lives, there is also increasing conversation surrounding AI Ethics. Most of the conversation revolves around questions of normative ethics - how should AI technologies behave, as to respect human values and to promote the successful and appropriate use of AI?


In this unit, we explore the broad subject of AI Ethics, examining some of the key ethical principles that apply to AI. In our first video lecture, Prof. Michael Rovatsos (The University of Edinburgh) presents a computer scientist’s perspective on the subject, exploring first some key definitions surrounding ethics and providing examples of ethical challenges that arise in real-world AI systems. He advocates for “Ethical AI,” the technical work that must go into engineering systems that respect human social and ethical norms. Prof. Rovatsos also explores what is unique about AI Ethics and how it differs from ethics in business, research and/or innovation practices.


Our second video lecture, by Prof. Joanna Bryson (Hertie School of Governance), presents several examples of socially biased AI systems / components, including her work on how gender stereotypes are reflected in word embeddings (Caliskan et al., 2017). Beyond these examples, Bryson characterizes the impacts of AI and related digital technologies, at both the micro (i.e., personal) and macro levels.


The third video lecture is presented by Prof. Bettina Berendt (TU Berlin, Weizenbaum Institute, KU Leuven) and is entitled “Deconstructing FAT (fairness, accountability and transparency)”. Prof. Berendt, a computer scientist, details her experiences in leading two interactive workshops, involving a technique called “mind scripting.” The technique enables a group of participants to explore their own implicit assumptions and values when talking about issues of FATE in AI. One of the key points explored is the role of context or situatedness (e.g., cultural context, group v. individual setting etc.) and its influence on the way we approach the study of ethical issues in AI systems.


Four suggested readings accompany the above video lectures. First, the article by Hagendorff (2020) provides an overview of the kinds of ethical issues addressed in AI Ethics guidelines. Specifically, Hagendorrf conducts a comparative analysis of 22 published guidelines, providing an overview of the normative ethical principles emphasized, but also the design recommendations provided in guidelines. Next, the article by Keyes and colleagues (2019) provides a satirical example of FATE analysis, through a “mulching proposal,” demonstrating the challenges of developing ethical systems and highlighting the limitations of FATE. Finally, the articles by Caliskan, Bryson and Narayanan (2017) and Elkin-Koren (2020) analyze examples of real-world systems/components.

Video Lectures

  1. “AI Ethics”, Prof. Michael Rovatsos (University of Edinburgh) - (slides)

2. “Personal and Transnational Economic Impacts of AI and Digital Technology”, Prof. Joahna Bryson (Hertie School of Governance) - (slides)

3. “Deconstructing FAT: using memories to collectively explore implicit assumptions, values and context in practices of debiasing and discrimination-awareness”, Prof. Bettina Berendt, (Technische Universität Berlin - (slides)

Reading List


  1. Caliskan, A., Bryson, J. J., & Narayanan, A. (2017). Semantics derived automatically from language corpora contain human-like biases. Science, 356(6334), 183-186. Link: https://science.sciencemag.org/content/356/6334/183

  2. Hagendorff, T. (2020). The ethics of AI ethics: An evaluation of guidelines. Minds and Machines, 30(1), 99-120

Link: https://link.springer.com/article/10.1007/s11023-020-09517-8

  1. Keyes, O., Hutson, J., & Durbin, M. (2019, May). A mulching proposal: Analysing and improving an algorithmic system for turning the elderly into high-nutrient slurry. In Extended Abstracts of the 2019 CHI Conference on Human Factors in Computing Systems (pp. 1-11).

Link: https://dl.acm.org/doi/abs/10.1145/3290607.3310433

  1. Elkin-Koren, N. (2020). Contesting algorithms: Restoring the public interest in content filtering by artificial intelligence. Big Data & Society, 7(2), 2053951720932296. Link: https://journals.sagepub.com/doi/full/10.1177/2053951720932296

Unit Quiz

By taking this Quiz you will be able to assess the knowledge your gain from this Unit.

You will get feedback immediately via Google Forms, once your responses are submitted.