Seminal
The magazine and atlas of EU-funded researchA new publication from the web and AI team at Scott Charnley

Atlas / France

Marie Skłodowska-Curie Action · Running

LIAISON Algorithmic Contract Design for AI Markets: Learning, Information Design, and Robustness to Strategic Manipulation

Coordinator
Institut National De Recherche En Informatique Et Automatique, France
Runs
1 May 2026 to 30 April 2028
Status
Running
EU funding
€226k
Total cost
—
Partners
1 organisations in 1 countries
Type
Marie Skłodowska-Curie Action
Grant agreement
101272753

What the project is doing

In the project's own words, from its summary on CORDIS.

Artificial Intelligence (AI) markets, such as outsourced data labelling and model training, are expanding rapidly but face incentive problems. Service providers may reduce effort to cut costs, while the complexity and randomness of machine-learning (ML) outputs make it difficult to verify quality or effort. While contract theory offers tools to align incentives, a gap remains between theory and practice. Classical models assume that game parameters are commonly known, whereas in reality, much of this information is private.

The learning efficiency and computational tractability of finding good contracts under such incomplete information remain largely unexplored. The irregular structures often lead to hardness results, while positive findings are scarce. Furthermore, when learning algorithms are deployed, agents with information advantages may manipulate the process, undermining the reliability of ML in economic interactions. This project develops an algorithmic framework for contract design with incomplete information, with applications in AI markets.

It focuses on two fundamental questions: efficient learning and strategic use of information. Three objectives guide the work: (1) characterizing when optimal contracts are efficiently learnable under realistic conditions; (2) studying how committed information disclosure (information design) shapes contract choices and payoffs; and (3) analyzing how informed agents manipulate a learning principal and designing robust countermeasures.

Combining the applicant’s expertise in algorithmic game theory with the host’s strengths in statistical learning and optimization, the project will advance contract theory along computational, learning, and economic dimensions, develop techniques that extend to general ML theory and Stackelberg games, and contribute to responsible AI markets. It supports Work Program’s priorities on human-centric, trustworthy AI and on strengthening Europe’s competitiveness in the digital transition.

Topics

Who is involved

OrganisationCityCountryRole
Institut National De Recherche En Informatique Et AutomatiqueLe Chesnay CedexFranceCoordinator

Source: CORDIS, European Commission. Data of 5 Nov 2026, reused under the Commission's reuse policy. View on CORDIS · Grant DOI