Ai, Ethics and Law

MORAL AGENCY
Week 1
Floridi & Sanders, 'On the morality of artificial agents', 2004
Week 2
Wallach & Allen, 'Can (ro)bots really be moral?', 2008
Week 3
Wynsberghe & Robbins, 'Critiquing the reasons for making artificial moral agents', 2019
MODULE 2: MORAL RESPONSIBILITY
Week 4
Elish, 'Moral crumple zones: cautionary tales in human-robot interaction', 2019
Week 5
Hakli & Mäkelä, 'Moral responsibility of robots and hybrid agents', 2019
Week 6
Thoma, 'Risk imposition by artificial agents: the moral proxy problem', 2022
MODULE 3: VALUE ALIGNMENT
Week 7
Wallach & Allen, 'Top-down morality', 'Bottom-up and developmental approaches', 2008
Week 8

Bostrom, 'The superintelligent will: motivation and instrumental rationality in advanced
artificial agents', 2012
Week 9
Gabriel, 'Artificial intelligence, values, and alignment', 2020
MODULE 4: POLITICAL DIMENSIONS
Week 10
Danaher, 'The threat of algocracy: reality, resistance, and accommodation', 2016
Week 11
Benn & Lazar, 'What's wrong with automated influence?', 2022
Week 12
Christiano, 'Algorithms, manipulation, and democracy', 2022

LAW
Module 0 — Foundations: law, norms, and technology (2 h)

Lecture 1. Why AI designers must engage with the law. What
a legal norm is: conditional structure (operative facts →
legal effect), the gap between rule and case, the problem
of interpretation. Normative systems compared: law and
code ("code is law," Lessig); architecture as regulation.

Essential analytic concepts: validity, efficacy,
legitimacy. Introduction to the double question and to the
idea of subjecting computational power to the rule of law
(digital constitutionalism).

Module 1 — Personal data protection (6 h)

Lecture 2. The architecture of the GDPR: material and territorial scope;
key definitions (personal data, processing,
pseudonymisation/anonymisation — relevant to ML); the principles of Art.
5; the principle of accountability.

Lecture 3. Lawful bases for processing (Art. 6); special categories of data
(Art. 9 — relevant to bias); data subject rights; transparency and
information duties; the roles (controller, processor, joint controllership).

Lecture 4. Data protection by design and by default (Art. 25) as an
engineering requirement; data protection impact assessment (DPIA, Art.
35); automated decision-making and profiling (Art. 22) as a bridge to the
AI Act. Brief coverage: ePrivacy, the data ecosystem (Data Act, Data
Governance Act), international transfers.

Module 2 — The EU AI Act (8 h)

Lecture 5. Genesis, rationale, and structure of Regulation (EU) 2024/1689.
The risk pyramid. The definition of an "AI system" (OECD alignment; the

ELI three-factor reading). Scope and exclusions. The actors in the value
chain: provider, deployer, importer, distributor.

Lecture 6. Prohibited AI practices (Art. 5). Classification of high-risk
systems: Annex III and the product-safety route (Annex I).

Lecture 7. Requirements for high-risk systems — risk management
system, data governance, technical documentation, logging/record-
keeping, transparency towards the deployer, human oversight, accuracy,
robustness, and cybersecurity. Mapping the requirements onto the ML
lifecycle (the course's principal STEM bridge).

Lecture 8. Obligations of the provider vs. the deployer; conformity
assessment and CE marking; harmonised standards and the role of
ISO/IEC 42001; GPAI models and systemic-risk obligations; governance
(AI Office, national competent authorities); the phased application
timeline. The evolving picture: the Digital Omnibus.

Module 3 — Liability and attribution (2 h)

Lecture 9. Civil liability for AI-related harm. The revised Product Liability
Directive (software and AI as a "product," burden of proof, disclosure
duties). The debate on the AI Liability Directive (a proposal of uncertain
outcome). Causation and system opacity. The responsibility gap and the
"problem of many hands" (a philosophical reading). Allocating
responsibility along the value chain; notes on insurance and
administrative dimensions.

Module 4 — Fundamental rights, non-discrimination, and human-centered
governance (4 h)

Lecture 10. The Charter of Fundamental Rights: dignity, private life,
equality. Algorithmic discrimination as a legal problem (not merely
technical): direct and indirect discrimination, proxies, the burden of proof
in anti-discrimination law. The gap between technical fairness metrics
and the legal notion of equality.

Lecture 11. Transparency, explainability, and contestability as legal
requirements (across the GDPR, the AI Act, and sectoral rules).
Meaningful human oversight and control. The Council of Europe
Framework Convention on AI, human rights, democracy, and the rule of
law. Digital constitutionalism: disciplining public and private algorithmic
power. (Optional note: DSA/DMA and platform power.)

Module 5 — Synthesis: compliance-by-design, ethics and law, cases (2 h)

Lecture 12. From rules to design: governance management systems
(ISO/IEC 42001), the compliance lifecycle, documentation as a form of
accountability. The double question revisited: where ethics exceeds the
law. Concluding case study (e.g. a high-risk AI system in healthcare,
recruitment, or public administration): end-to-end analysis of obligations
across data protection, the AI Act, liability, and fundamental rights. Future
trajectories and the professional responsibility of those who design.