Ethical and Legal Challenges of AI and Data Science
Practical Seminar Block Course: 14.-25. September
(For our virtual seminar we use Webex Meetings. In case you want to join a meeting please drop us an email.)
Instructors:
- Prof. Dr. Christoph Benzmüller (FU Berlin, Dep. of Mathematics and Computer Science)
- Prof. Dr. Bertram Lomfeld (FU Berlin, Dep. of Law)
Further contributors:
- PD Dr. Jürgen Altmann (Universität Dortmund, Dep. of Physics): Expert reporting on the current state and trends of military AI applications
- Prof. Dr. Sabine Ammon (TU Berlin, Dep. of Philosophy): Expert on ethics of technology and technology assessment
- Dr. Vaishak Belle (University of Edinburgh, UK, School of Informatics, Partner in the UNA Europa University Alliance): Expert on explainability & ethics in AI, fairness in machine learning
- Dr. Huimin Dong (Zheijang University, China, School of Humanities): Expertise in logic for law and ethics, logic for social networks
- Prof. Dr. Philipp Hacker (HU Berlin, Dep. of Law): Legal expert reporting on discriminating algorithms
- Prof. Dr. Volker Roth (FU Berlin, Dep. of Mathematics and Computer Science): Expert on secure identity
- Marcus Soll (Universität Hamburg): Expertise on adversial atttacks
Course poster: see here
Target group: Master’s students (Bachelor’s students from the fifth semester onwards)
Prerequisites: None
Block course format: The main part pf the seminar will be held in the form of a two week intensive block course (9:00-12:00 and 13:30--15:00) to be held from 14.-25. September. A mixture of lectures (including invited lectures by experts), student presentations and student projects/group work will be offered. The seminar includes an intensive preparation and postprocessing phase.
Student deliverables: Presentation (30 + 15 min), report (about 10 pages), further contributions depending on selected topic (surveys, programming, modeling, empirical studies), active participation in discussions
Description: Ethical and legal challenges in AI and Data science will be identified and options to resolve or control them will explored and discussed. The list of topics that will be addressed include:
- Introduction, survey and delineation on core notions, including "AI", "Data Science", "Machine Learning", "Ethics and Law in AI and Data Science"
- Survey and discussion of international positions and recommendations on ethical and legal regulation of AI and data science applications
- Clarification and discussion of relevant notions, including "Trustworthy AI", "Explainable AI", "Transparent AI"
- Elaboration of a spectrum of critical and non-critical applications of AI and data science technology
- Exemplary discussion of selected, critical application areas, including e.g. military applications, automated financial markets, criminal profiling, etc.
- Adversial attacks, and potential countermeasures
- Bias in data science and machine learning, and potential countermeasures
- Means to explain and assess decision making in data science and AI
- Means to enforce ethical and legal control in data science and machine learning
- Means to enforce ethical and legal control in large, integrated AI systems
- Ethics and security
Preliminary Schedule (changes my still apply)
Day 1 (14.9. Monday)
- Morning (starting at 9:00):
- Welcome (Christoph Benzmüller/Bertram Lomfeld)
- Brief Introduction: The Need for "Ethical" Intelligent Systems (Christoph Benzmüller)
- Regulation for AI and Data Science (Bertram Lomfeld)
- Reading:
- On Artificial Intelligence - A European approach to excellence and trust, European Commission, White Paper, COM(2020) 65 final, 2020. https://ec.europa.eu/info/sites/info/files/commission-white-paper-artificial-intelligence-feb2020_en.pdf
- Expert opinions on AI and Robotics: https://www.bundestag.de/ausschuesse/ausschuesse18/a23/anhoerungen/fachgespraech-494576
- Afternoon (starting at 13:00):
- students introduction
- organisation of interdisciplinary
- student teams
- discussion of course organisation and assessment
Day 2 (15.9. Tuesday)
- Morning (starting at 9:00):
- Moral Machines Experiment and its Implications (Bertram Lomfeld)
- Reasonable Machines (Christoph Benzmüller)
- Reading:
- The Moral Machine experiment, Nature, 563, pp. 59–64, 2018.
- Reasonable Machines: A Research Manifesto, KI 2020, 2020.
- Afternoon (14:00-16:00):
- Technological impact assessment (Sabine Ammon), eventually related exercises/group work/student presentations
Day 3 (16.9. Wednesday)
- Morning (starting at 9:00):
- Towards Interpretable, Fair and Responsible AI: A Probabilistic Logical Approach (Vaishak Belle)
- Reading:
- Fairness in Machine Learning with Tractable Models. AAAI Workshop: Statistical Relational Artificial Intelligence, 2020.
- The quest for interpretable and responsible artificial intelligence. The Biochemist, 2019.
- Tractable Probabilistic Models for Moral Responsibility. NeurIPS Workshop on Knowledge Representation & Reasoning Meets Machine Learning, 2019.
- Philosophical foundations on the notion of fairness: Normative Principles for Evaluating Fairness in Machine Learning. AIES 2020.
- Symbolic Logic meets Machine Learning: A Brief Survey in Infinite Domains. SUM, 2020.
- Reading:
- Towards Interpretable, Fair and Responsible AI: A Probabilistic Logical Approach (Vaishak Belle)
- Afternoon (starting at 13:00): related exercises/group work/student presentations
Day 4 (17.9. Thursday)
- Morning (starting at 9:00):
- Adversial attacks (Marcus Soll)
- Reading:
- Marcus Soll: InformatiCup Competition 2019: Fooling Traffic Sign Recognition. KI 2019: 325-332, 2019. https://doi.org/10.1007/978-3-030-30179-8_29
- Marcus Soll, Tobias Hinz, Sven Magg, Stefan Wermter:
Evaluating Defensive Distillation for Defending Text Processing Neural Networks Against Adversarial Examples. ICANN (3) 2019: 685-696, 2020. https://doi.org/10.1007/978-3-030-30508-6_54
- Afternoon (starting at 13:00): Group exercises on adversial attacks (these could extend over the WE into the next week)
Day 5 (18.9. Friday)
- Morning (starting at 9:00):
- Military applications of AI (Jürgen Altmann)
- Reading:
- V. Boulanin (ed.): The Impact of Artificial Intelligence on Strategic Stability and Nuclear Risk - Volume I, Euro-Atlantic Perspectives, Solna: SIPRI, 2019. Suggested chapters:
- M. C. Horowitz: Artificial intelligence and nuclear stability, p. 79-83
- F. Sauer: Military applications of artificial intelligence: Nuclear
risk redux, p. 84-90 - J.-M. Rickli: The destabilizing prospects of artificial intelligence for nuclear strategy, deterrence and stability, p. 91-98
- J. Altmann: Autonomous Weapon Systems - Dangers and Need for an International Prohibition. KI 2019. (see also mycampus resource folder)
- (Further background reading)
- V. Boulanin, et al.: Artificial Intelligence, Strategic Stability and Nuclear Risk, Solna: SIPRI, June, 2020
- Stanley Center for Peace and Development/United Nations Office for
Disarmament Affairs/Stimson Center: The Militarization of Artificial Intelligence, 2019/2020,
- V. Boulanin (ed.): The Impact of Artificial Intelligence on Strategic Stability and Nuclear Risk - Volume I, Euro-Atlantic Perspectives, Solna: SIPRI, 2019. Suggested chapters:
- Afternoon (starting at 13:00):
- Secure identity, Software Licensing and AI (Volker Roth)
Day 6 (21.9. Monday)
- Morning (9:00-10:30):
- Discriminating algorithms (Philipp Hacker): This talk will address (i) technical foundations and causes for discrimination, (ii) short legal assessment from the perspective of anti-discrimination law, (iii) technical measures towards algorithimc fairness, (iv) legal boundaries of algorithimc fairness
- Reading:
- Teaching Fairness to Artificial Intelligence: Existing and Novel Strategies Against Algorithmic Discrimination Under EU Law. Common Market Law Review 55: 1143–1186, 2018. Preprint
- Matching code and law: achieving algorithmic fairness with
optimal transport. Data Mining and Knowledge Discovery 34:163–200, 2020. Preprint - Suggested further reading: Big Data's Disparate Impact, Why Fairness Cannot Be Automated: Bridging the Gap Between EU Non-Discrimination Law and AI, Incomprehensible Discrimination
- Afternoon (starting at 13:00): Active participation (with our student teams) at the student day of KI 2020 conference.
Day 7 (22.9. Tuesday)
- Morning (starting at 9:00):
- Logic of Legal Rights and Normative Positions (Huimin Dong)
- Reading:
- Sergot, M.: Normative positions. In: Gabbay, D., Horty, J., Parent, X., van der Meyden, R., van der Torre, L. (eds.) Handbook of Deontic Logic and Normative Systems, vol. 1. College Publication, London (2013). https://collegepublications.co.uk/handbooks/?00001
- Kanger, S.: Law and logic. Theoria 38(3), 105–132 (1972). https://link.springer.com/chapter/10.1007/978-94-010-0500-5_14
- Makinson, D.: On the formal representation of rights relations. J. philos. Log. 15(4), 403–425 (1986). https://www.jstor.org/stable/30226364
- Afternoon: Designing normative theories for ethical and legal reasoning: LogiKEy framework, methodology, and tool support (Christoph Benzmüller)
- Reading:
- Designing Normative Theories for Ethical and Legal Reasoning: LogiKEy Framework, Methodology, and Tool Support. Artificial Intelligence, 237, pp. 103348, 2020. Preprint
- LogiKEy Workbench: Deontic Logics, Logic Combinations and Expressive Ethical and Legal Reasoning (Isabelle/HOL Dataset), 2020. http://logikey.org
- Encoding Legal Balancing: Automating an Abstract Ethico-Legal Value Ontology in Preference Logic, 2020. Preprint
- Reading:
Day 8 (23.9. Wednesday)
- Morning (starting at 9:00):
- PhD projects on ethical reasoning (David Fuenmayor, Sarah Hiller, etc.)
- Afternoon (starting at 13:00):
- Exercises/group work/student presentations
Day 9 (24.9. Thursday)
- Morning (starting at 9:00):
- Exercises/group work/student presentations
- Afternoon (starting at 13:00):
- Exercises/group work/student presentations
Day 10 (25.9. Friday)
- Morning (starting at 9:00):
- Summary
- Activities in Berlin: Ethics Lab
- Final discussion (Christoph Benzmüller/Bertram Lomfeld)
- Afternoon (starting at 13:00):
- (not part of the official seminar) Meet for a drink in the real world?
Exercise Topics as discussed on 14.9.:
TOPIC 1: AI Definition and Superintelligence
TOPIC 2: AI Ethics Principles (Guidelines)
TOPIC 3: Moral Machines (Ethical AI decisions)
TOPIC 4: AI Fairness, Discrimination & Biases
TOPIC 5: ML Security (Adversial Attacks, secure identity)
TOPIC 6: Law & Formal Logic (Symbolic AI)