Embedded Files
"Why, so often, do people build what nobody wants? Why, so often, do engineers optimise their solution based only on physical capabilities and fail to consider the stakeholders’ desirabilities?"
van Heukelum, Binnekamp, Wolfert
Course syllabus
Allocation decisions regulate access to any scarce resources like highway lanes, public transit, road space, water, energy, grid capacity, user attention, right to pollute, etc. With further automation of these decisions that are relevant for society, the control engineers need to make increasingly decisive assumptions regarding what society wants. Some design choices may be driven by norms and conventions rather than conscientiously responsible and ethical design. This topic would propose a welfarist control theory and new certificates for control systems. Similarly to how we certify stability, robustness, performance of a control system, we need a theory to certify that the control system that we design fulfills the mandate received from society (fair / equitable / efficient use of a shared resource).
Lecturers
Saverio Bolognani
Heinrich Nax
Schedule
Friday 12 September 2025 (Lausanne, MFX 1) - Opening
Friday 3 October 2025 (Zurich, ETL K25) - Lecture 1
Friday 24 October 2025 (Lausanne, ME C2 405) - Lecture 2
Friday 21 November 2025 (Zurich, ETL K25) - Lecture 3
Friday 19 December 2025 (Lausanne, ME C2 405) - Lecture 4
Friday 16 January 2026 (Zurich, HG D 16.2) - Closing
Time: 13:30 to 17:30
Session 1 - Social Choice Theory Basics and Impossibility
"So if one person gains while everyone else loses, we are not allowed to declare this change to be a deterioration, if we seek only Pareto efficiency. This remarkable reticence, it seems fair to guess, would have appealed to Emperor Nero, who evidently enjoyed playing his music while Rome burned and all other Romans were plunged into misery."
Amartya Sen
Core reading
Jump-start paper:
Welfare and Cost Aggregation for Multi-Agent Control: When to Choose Which Social Cost Function, and Why?
https://arxiv.org/abs/2503.20772
This paper translates into control language some very seminal readings such as:
A. Sen. (2008). Social Choice. In: Durlauf, S.N., Blume, L.E. (eds) The New Palgrave Dictionary of Economics. Palgrave Macmillan, London. PDF
Kevin W. S. Roberts. (1980). Interpersonal Comparability and Social Choice Theory. The Review of Economic Studies, 47(2), 421–439. https://doi.org/10.2307/2297002
C. d’Aspremont and L. Gevers (2002). Social Welfare Functionals and Interpersonal Comparability. Handbook of Social Choice and Welfare, K. J. Arrow, A. K. Sen, and K. Suzumura (Eds.), Elsevier, Vol. 1, Ch. 10, pp. 459–541. https://doi.org/10.1016/S1574-0110(02)80014-5
Opposing positions
Position A: We should optimize the weighted sum of costs (efficiency) in control engineering problems. Reaching the necessary level of comparability is just a measurability requirement that can be satisfied in the deployment/engineering of the solution.
Position B: Never assume comparability unless you have solid reasons to do so. Many applications incorrectly employ social cost functions without foundational support, which can lead to misleading or invalid conclusions.
Session 2 - Who decides what is fair?
Core reading
Engineering moral machines
Michael Fisher, Christian List, Marija Slavkovik, and Alan Winfield
https://eprints.lse.ac.uk/68212/1/List_Engineering%20moral_2016.pdf
Additional readings
Control
Fair-MPC: a control-oriented framework for socially just decision-making
Eugenia Villa, Valentina Breschi, Mara Tanelli
https://arxiv.org/pdf/2312.05554Machine learning
DECAF: Learning to be Fair in Multi-agent Resource Allocation
Ashwin Kumar, William Yeoh
https://arxiv.org/pdf/2502.04281Optimization
A guide to formulating fairness in an optimization model
Violet Xinying Chen, J. N. Hooker
https://link.springer.com/article/10.1007/s10479-023-05264-yEnergy
Review on fairness in local energy systems
João Soares, Fernando Lezama, Ricardo Faia, Steffen Limmer, Manuel Dietrich, Tobias Rodemann, Sergio Ramos, Zita Vale
Applied Energy, Volume 374, 2024, https://doi.org/10.1016/j.apenergy.2024.123933
Opposing positions
POSITION A: The appropriate notion of fairness should be selected by the same agents affected by the decision. Fairness is something endogenous. Here is how it can be done.
POSITION B: Fairness is a design choice. It’s exogenous to the decision process, it is part of the specifications and it is decided by the designer or part of the mandate that the designer receives. Here is how it is expressed.
Session 3 - Mechanism design and extracting preferences
Core reading
Efficient Mechanisms for Bilateral Trading
Myerson, Roger B.; Mark A. Satterthwaite (1983).
Journal of Economic Theory. 29 (2): 265–281. doi:10.1016/0022-0531(83)90048-0. http://www.kellogg.northwestern.edu/research/math/papers/469.pdf
Additional readings
Strategy-proofness and Arrow's conditions: Existence and correspondence theorems for voting procedures and social welfare functions
Satterthwaite, Mark Allen (April 1975).
Journal of Economic Theory. 10 (2): 187–217. CiteSeerX 10.1.1.471.9842. doi:10.1016/0022-0531(75)90050-2
Manipulation of Voting Schemes: A General Result
Allan Gibbard
https://www.jstor.org/stable/1914083
Mechanism Design Theory in Control Engineering: A Tutorial and Overview of Applications in Communication, Power Grid, Transportation, and Security Systems
I. V. Chremos and A. A. Malikopoulos
doi:10.1109/MCS.2023.3329919
Opposing positions
POSITION A: Control engineers do not need to extract preferences. Preferences are part of the modeling process, and if any private information needs to be collected, then it can be done via protocols and assume truthfulness.
POSITION B: Without a mechanism that ensures truthfulness, preferences are useless and should not be trusted, even in control engineering problems where agents are automated.
Session 4 - Dynamic Social Choice Theory
"It’s impossible to use the word, dynamic, in a pejorative sense. Try thinking of some combination that will possibly give it a pejorative meaning. It’s impossible."
Richard Bellmann
Core reading
A Vision for Trustworthy, Fair, and Efficient Socio-Technical Control using Karma Economies
Ezzat Elokda, Andrea Censi, Emilio Frazzoli, Florian Dörfler, Saverio Bolognani
https://arxiv.org/abs/2506.17115
Additional readings
Achieving Fairness in Multi-Agent Markov Decision Processes Using Reinforcement Learning
Peizhong Ju, Arnob Ghosh, Ness B. Shroff
https://arxiv.org/pdf/2306.00324
Fairness in Multi-Agent Sequential Decision-Making
Chongjie Zhang and Julie A. Shah
https://proceedings.neurips.cc/paper_files/paper/2014/file/5556d1d6ca0d004accf36cc2db73e736-Paper.pdf
Socially Fair Reinforcement Learning
Debmalya Mandal, Jiarui Gan
https://arxiv.org/abs/2208.12584
Fair and Efficient Social Choice in Dynamic Settings
Rupert Freeman, Vincent Conitzer, Seyed Majid Zahedi
https://www.rupertfreeman.com/fair_social_choice_full.pdf
Social Choice with Changing Preferences: Representation Theorems and Long-Run Policies
Kshitij Kulkarni, Sven Neth
https://arxiv.org/abs/2011.02544
Folk theorem (repeated games)
https://en.wikipedia.org/wiki/Folk_theorem_(game_theory)
Opposing positions
POSITION A: The design of dynamic resource allocation with specific fairness metrics only makes sense if it is compartmentalized into specific engineering problems.
POSITION B: Assuming that we can extract preferences and agree on a fairness metric, the next natural step is to extend the scope of the dynamic resource allocation scheme to smart cities, multiple resources, etc.
Google Sites
Report abuse