Grant: Discovery Grants Program - Individual
The University of British Columbia
$59,000.00
Learning Models of Strategic Behaviour
FY 2023-24
Summary
The proposed research will develop new techniques for learning models of strategic behaviour. A strategic situation is one in which multiple, self-interested agents interact and where an agent's satisfaction with the resulting state of the world depends in part on other agents' decisions. A model of strategic behaviour is a prediction about how an agent will behave given the information available to it. Such models are important for reasoning about counterfactuals (describing how agents would act in different hypothetical scenarios), which in turn enables both (i) predictions about how agents would behave in novel settings and (ii) optimization over such settings to maximize design objectives like social welfare. Classical game theory is based on a simple model of strategic behaviour: that agents choose actions that maximize expected utility. This model (called "economic rationality") has a lot going for it: it's relatively straightforward to reason about mathematically; it makes a sharp prediction in every strategic setting; and if an agent doesn't follow it, in some sense that agent is making a mistake. As a result, game theory is one of the foundational pillars of modern economics, and is widely used in computer science, national defense, business, political science, psychology, and many other fields. The model of the economically rational agent is often a good description of sophisticated actors like corporations and governments; it can be a poor description of agents that exhibit cognitive biases (e.g., unsophisticated humans) or computational limitations (e.g., agents facing settings that are so complex that optimal behaviour becomes infeasible). The field of behavioral game theory fills this gap. To date, most models in behavioral game theory are constructed by hand based on one or two insights from behavioural psychology and validated in controlled laboratory experiments. The long-term vision driving the proposed research is that modern AI/ML techniques can become a valuable tool for identifying behavioral game theory models from real-world data. To advance this vision I will work on foundations, advancing general techniques for constructing and validating models from data. I will also show how domain-specific models can be used to solve highly complex, real-world problems in four concrete application domains: combinatorial clock auctions; mobile games; peer review; and AI-enabled market design for social impact.
Program
Discovery Grants Program - Individual
Recipient
Location
British Columbia, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
RGPIN-2023-04546
Name nom
Leyton-Brown, Kevin
Department département
Computer Science
Organization id
2
Institution établissement
The University of British Columbia
Province en
British Columbia
Province fr
Colombie-Britannique
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2023
Award amount
59000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
1207
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Information, computer and communication technologies
Area of application fr
Technologie de l'information, des ordinateurs et des réseaux de communications
Research subject code
2805
Research subject group en
Artificial intelligence
Research subject group fr
Learning and inference theories
Research subject en
Intelligence artificielle
Research subject fr
Théories de l'apprentissage et l'inférence
Application title
Learning Models of Strategic Behaviour
Keyword
AI for social impact; artificial intelligence; behavioural models; combinatorial auctions; game theory; machine learning; mobile gaming; peer review
Application summary
The proposed research will develop new techniques for learning models of strategic behaviour. A strategic situation is one in which multiple, self-interested agents interact and where an agent's satisfaction with the resulting state of the world depends in part on other agents' decisions. A model of strategic behaviour is a prediction about how an agent will behave given the information available to it. Such models are important for reasoning about counterfactuals (describing how agents would act in different hypothetical scenarios), which in turn enables both (i) predictions about how agents would behave in novel settings and (ii) optimization over such settings to maximize design objectives like social welfare. Classical game theory is based on a simple model of strategic behaviour: that agents choose actions that maximize expected utility. This model (called "economic rationality") has a lot going for it: it's relatively straightforward to reason about mathematically; it makes a sharp prediction in every strategic setting; and if an agent doesn't follow it, in some sense that agent is making a mistake. As a result, game theory is one of the foundational pillars of modern economics, and is widely used in computer science, national defense, business, political science, psychology, and many other fields. The model of the economically rational agent is often a good description of sophisticated actors like corporations and governments; it can be a poor description of agents that exhibit cognitive biases (e.g., unsophisticated humans) or computational limitations (e.g., agents facing settings that are so complex that optimal behaviour becomes infeasible). The field of behavioral game theory fills this gap. To date, most models in behavioral game theory are constructed by hand based on one or two insights from behavioural psychology and validated in controlled laboratory experiments. The long-term vision driving the proposed research is that modern AI/ML techniques can become a valuable tool for identifying behavioral game theory models from real-world data. To advance this vision I will work on foundations, advancing general techniques for constructing and validating models from data. I will also show how domain-specific models can be used to solve highly complex, real-world problems in four concrete application domains: combinatorial clock auctions; mobile games; peer review; and AI-enabled market design for social impact.