ENSAE Paris - École d'ingénieurs pour l'économie, la data science, la finance et l'actuariat

Ethics and responsibility in data science- group 1

Teacher

TUBARO Paola

Department: Sociology

Objective

Objectives
Building on students’ (expected) basic knowledge of data ethics and data law, this course analyzes in greater depth the concrete problems that arise in advanced data science projects, and prepares students to manage them both within the framework of their studies and with a view to their future professional integration. The course addresses the ethical and social issues of digitalization, data, and more specifically artificial intelligence (AI), from the perspective of the human choices and social structures that shape them intrinsically, and that frame the work of data scientists. To do this, students will learn to leverage the tools of critical and historical thinking from the humanities and social sciences to recognise, analyse and ideally re-imagine the human, political and economic contexts in which data is produced and used. Students in the data science tracks will thus be prepared to engage as informed and responsible professionals in the various domains of our 'datafied' economies, while the other students will be able to deepen their knowledge and enhance their ability to contribute to public debates and to participate in technology policy choices.

Learning outcomes:
At the end of this course, you will be able to:
- analyse how data science is transforming individual and collective life
- recognise the influence of representations, norms and values on the development of algorithms and data-based solutions
- identify the assumptions underpinning algorithms and 'smart' solutions
- discuss the visions of the future implicit in existing AI solutions
- reflect on desirable outcomes and ways to achieve them

Planning

Teaching and learning activities
We will meet for four three-hour sessions, during which the themes of the course will be addressed, in part, on the basis of concrete use cases, which illustrate their relevance and potential impacts. As it is impossible to cover all the possible themes exhaustively, the course will begin with a fairly general overview of the issues, followed by a focus on a selection of major current questions, and will conclude with the presentation of possible solutions.

Plan
Session 1: The social challenges of a data-driven world
- Why a growing concern for ethics in relation to data science, digital technology and artificial intelligence
- Highly publicised debates: bias, discrimination, injustice...
- Current initiatives and their limits: proliferation of ethical charters; technical solutions (e.g. research on algorithmic fairness)
- Polemics on ‘ethics washing’, incomplete or absent regulation
- Value of a holistic approach – taking into account all the socio-politico-economic systems in which the technology is integrated


Session 2: The cost of inputs
Part 1: Technology and sustainable development
- The environmental and material costs of digital technology; direct and indirect impacts
- The energy consumption of algorithms; measuring costs and benefits
- The AI supply chain in a globalised extractive industry
Part 2: Data, privacy and surveillance
- Tensions between the need to protect personal data and the business models of data-intensive technologies
- Derivatives of surveillance systems
- Limits of existing solutions, both legal and technical


Session 3: Artificial intelligence, employment and work
- Will robots replace workers?
- Data work in the service of AI: a deterioration in working conditions and remuneration, already in place
- A production system that increases inequalities


Session 4: What to do?
- End-to-end ethics
- Rethinking governance: commons systems
- Forms of ‘data-activism’, reappropriation, and dialogue

References

Ananny M. & Crawford K. 2018. Seeing without knowing: Limitations of the transparency ideal and its application to algorithmic accountability. New Media & Society, 20(3): 973-989.

Buolamwini J. & Gebru T. 2018. Gender shades: Intersectional accuracy disparities in commercial gender classification. Proceedings of the 1st Conference on Fairness, Accountability and Transparency, PMLR 81: 77-91, https://proceedings.mlr.press/v81/buolamwini18a.html

Casilli A.A. & Tubaro P. (2023, à paraître). An end-to-end approach to ethical AI: Socio-economic dimensions of the production and deployment of automated technologies. In L. Robinson & S. Rogerson (éds.) Handbook of Digital Social Science, Edward Elgar.

Dulong de Rosnay M. & Musiani F. 2020. Alternatives for the Internet: A journey into decentralised network architectures and information commons. tripleC: Communication, Capitalism & Critique, 18(2): 622-629

Hanna A., Denton E., Smart A. & Smith-Loud J. 2020. Towards a critical race methodology in algorithmic fairness. Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (FAT* '20), ACM, 501–512

Jobin A., Ienca M. & Vayena E. 2019. The global landscape of AI ethics guidelines. Nature Machine Intelligence, 1: 389–399

Miceli M., Schuessler M. & Yang T. 2020. Between subjectivity and imposition: Power dynamics in data annotation for computer vision. Proceedings of the ACM on Human-Computer Interaction. ACM, 1, 1, Article 115, https://dl.acm.org/doi/pdf/10.1145/3415186

Raji I.D., Scheuerman M.K. & Amironesei R. 2021. ‘You can’t sit with us’: exclusionary pedagogy in AI ethics education. Proceedings of the ACM FAccT ’21 Conference, 515–525.

Strubell E., Ganesh A. & McCallum A. 2019. Energy and policy considerations for deep learning in NLP. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 3645-3650. Association for Computational Linguistics

Tubaro P. & Casilli A.A. 2022. Human listeners and virtual assistants: Privacy and labor arbitrage in the production of smart technologies. In F. Ferrari & M. Graham (eds.), Digital Work in the Planetary Market, MIT Press, pp. 175-190