Strategic Reasoning in Automated Mechanism Design

Logics in Artificial Intelligence, 2021

In this paper, we investigate strategic reasoning in the context of auctions. More precisely, we establish an explicit link between bidding actions and bounded rationality. To do so, we extend the Auction Description Language with an epistemic operator and an action choice operator and use it to represent a classical auction where agents have imperfect information about other agents' valuations. We formalize bounded rationality concepts in iterative protocols and show how to use them to reason about the players' actions. Finally, we provide a model checking algorithm.

IISc Lecture Notes Series, 2014

Aaai, 2007

Over the past half decade, we have been exploring the use of logic in the specification and analysis of computational economic mechanisms. We believe that this approach has the potential to bring the same benefits to the design and analysis of computational economic mechanisms that the use of temporal logics and model checking have brought to the specification and analysis of reactive systems. In this paper, we give a survey of our work. We first discuss the use of cooperation logics such as Alternating-time Temporal Logic (ATL) for the specification and verification of mechanisms such as social choice procedures. We motivate the approach, and then discuss the work we have done on extensions to ATL to support incomplete information, preferences, and quantification over coalitions. We then discuss is the use of ATL-like cooperation logics in the development of social laws.

2006 IEEE/WIC/ACM International Conference on Intelligent Agent Technology, 2006

Game theory is popular in agent systems for designing auctions with desirable properties. However, many of these properties will only hold if the game and its properties are common knowledge among the agents. For example, in an auction where truthful bidding is an equilibrium strategy, unless this is common knowledge, it may not be rational for an agent to bid truthfully. It is currently not clear how this state of common knowledge can be achieved, especially in open agent societies where agents may encounter previously unseen auction specifications. We need a method for communicating the rules of the game to the agents, and the agents need to be able to determine its properties. We present a machine-readable language in which the rules of the game can be written. We show that it is not feasible for an agent to determine the properties of any arbitrary specification, unless information about the properties is communicated and/or certain restrictions are placed on the specification. We look at two special cases where common knowledge is achievable: auctions with identical players where the two highest bidders determine the price, and Groves mechanisms with a restriction on the pricing rule.

… papers= Documentos de …, 2004

In this paper we develop a differentiable approach to deal with incentives in a, possibly small, subset of a general domain of preferences in economies with one public and one private good. We show that, for two agents, there is no social rule which is efficient, nondictatorial and strategy-proof. For the case of more agents the same result occurs when nondictatorship is replaced by Individual Rationality or by Envy-Freeness. Journal of Economic Literature.

2014

In traditional algorithm design, no incentives come into play: the input is given, and your algorithm must produce a correct output. How much harder is it to solve the same problem when the input is not given directly, but instead reported by strategic agents with interests of their own? The unique challenge stems from the fact that the agents may choose to lie about the input in order to manipulate the behavior of the algorithm for their own interests, and tools from Game Theory are therefore required in order to predict how these agents will behave. We develop a new algorithmic framework with which to study such problems. Specifically, we provide a computationally efficient black-box reduction from solving any optimization problem on "strategic input," often called algorithmic mechanism design to solving a perturbed version of that same optimization problem when the input is directly given, traditionally called algorithm design. We further demonstrate the power of our framework by making significant progress on several long-standing open problems. First, we extend Myerson's celebrated characterization of single item auctions [Mye81] to multiple items, providing also a computationally efficient implementation of optimal auctions. Next, we design a computationally efficient 2-approximate mechanism for job scheduling on unrelated machines, the original problem studied in Nisan and Ronen's paper introducing the field of Algorithmic Mechanism Design [NR99]. This matches the guarantee of the best known computationally efficient algorithm when the input is directly given. Finally, we provide the first hardness of approximation result for optimal mechanism design.

2009

In this paper we tackle the verification of basic auction models that are at the heart of many marketbased multi-agent systems. Specifically, we program auctions in a BDI-based programming language and then use agent model checking to verify logical properties concerning time, beliefs and goals within the multi-agent system. The basic auction model is then extended with coalition formation and trust, and verification of these aspects is carried out. * Work supported by EPSRC (UK) grant EP/D052548.

2020

Optimal auctions maximize a seller's expected revenue subject to individual rationality and strategyproofness for the buyers. Myerson's seminal work in 1981 settled the case of auctioning a single item; however, subsequent decades of work have yielded little progress moving beyond a single item, leaving the design of revenue-maximizing auctions as a central open problem in the field of mechanism design. A recent thread of work in "differentiable economics" has used tools from modern deep learning to instead learn good mechanisms. We focus on the RegretNet architecture, which can represent auctions with arbitrary numbers of items and participants; it is trained to be empirically strategyproof, but the property is never exactly verified leaving potential loopholes for market participants to exploit. We propose ways to explicitly verify strategyproofness under a particular valuation profile using techniques from the neural network verification literature. Doing so req...

ACM Transactions on Economics and Computation, 2013

There is only one technique for prior-free optimal mechanism design that generalizes beyond the structurally benevolent setting of digital goods. This technique uses random sampling to estimate the distribution of agent values and then employs the Bayesian optimal mechanism for this estimated distribution on the remaining players. Though quite general, even for digital goods, this random sampling auction has a complicated analysis and is known to be suboptimal. To overcome these issues we generalize the consensus and profit extraction techniques from Goldberg and Hartline [2003] to structurally rich environments that include, for example, single-minded combinatorial auctions.

Journal of Mathematical Economics, 1982

Following the Gibbard-Satterthwaite impossibility theorem, the incentives literature has explored restrictions on the domains of unknown characteristics which enable the implementation of some social choice rules by dominant strategy incentive compatible mechanisms. This paper makes precise the intuitive idea that, even when restrictions on preferences take the extreme form of iinitedimensional parametrizations, implementable social choice rules display a lack of robustness and in a sense made explicit are rare.