Victor Naroditskiy

Verification in referral-based crowdsourcing

Online social networks offer unprecedented potential for rallying a large number of people to accomplish a given task. Here we focus on information gathering tasks where rare information is sought through “referral-based crowdsourcing”: the information request is propagated recursively through invitations among members of a social network. Whereas previous work analyzed incentives for the referral process in a setting with only correct reports, misreporting is known to be both pervasive in crowdsourcing applications, and difficult/costly to filter out. A motivating example for our work is the DARPA Red Balloon Challenge where the level of misreporting was very high. In order to undertake a formal study of verification, we introduce a model where agents can exert costly effort to perform verification and false reports can be penalized. This is the first model of verification and it provides many directions for future research, which we point out. Our main theoretical result is the compensation scheme that minimizes the cost of retrieving the correct answer. Notably, this optimal compensation scheme coincides with the winning strategy of the Red Balloon Challenge.

Information verification during natural disasters

Large amounts of unverified and at times contradictory information often appear on social media following natural disasters. Timely verification of this information can be crucial to saving lives and for coordinating relief efforts. Our goal is to enable this verification by developing an online platform that involves ordinary citizens in the evidence gathering and evaluation process. The output of this platform will provide reliable information to humanitarian organizations, journalists, and decision makers involved in relief efforts.

Crowdsourcing contest dilemma

Crowdsourcing offers unprecedented potential for solving tasks efficiently by tapping into the skills of large groups of people. A salient feature of crowdsourcing—its openness of entry—makes it vulnerable to malicious behaviour. Such behaviour took place in a number of recent popular crowdsourcing competitions. We provide game-theoretic analysis of a fundamental trade-off between the potential for increased productivity and the possibility of being set back by malicious behaviour. Our results show that in crowdsourcing competitions malicious behaviour is the norm, not the anomaly—a result contrary to the conventional wisdom in the area. Counterintuitively, making the attacks more costly does not deter them but leads to a less desirable outcome. These findings have cautionary implications for the design of crowdsourcing competitions.

Redistribution of VCG payments in public project problems

Redistribution of VCG payments has been mostly studied in the context of resource allocation. This paper focuses on another fundamental model--the public project problem. In this scenario, the VCG mechanism collects in payments up to

Optimal payments in dominant-strategy mechanisms for single-parameter domains

\frac{n-1}{n}

Task assignment with controlled and autonomous agents

of the total value of the agents. This collected revenue represents a loss of social welfare. Given this, we study how to redistribute most of the VCG revenue back to the agents. Our first result is a bound on the best possible efficiency ratio, which we conjecture to be tight based on numerical simulations. Furthermore, the upper bound is confirmed on the case with 3 agents, for which we derive an optimal redistribution function. For more than 3 agents, we turn to heuristic solutions and propose a new approach to designing redistribution mechanisms.

Maximizing social welfare in congestion games via redistribution

We study dominant-strategy mechanisms in allocation domains where agents have one-dimensional types and quasilinear utilities. Taking an allocation function as an input, we present an algorithmic technique for finding optimal payments in a class of mechanism design problems, including utilitarian and egalitarian allocation of homogeneous items with nondecreasing marginal costs. Our results link optimality of payment functions to a geometric condition involving triangulations of polytopes. When this condition is satisfied, we constructively show the existence of an optimal payment function that is piecewise linear in agent types.

An algorithm for the penalized multiple choice knapsack problem

We analyse assignment problems in which not every agent is controlled by the central planner. The autonomous agents search for vacant tasks guided by their own preference orders over available tasks. The goal of the central planner is to maximise the total value of the assignment, taking into account the behaviour of the uncontrolled agents. Such optimisation problems arise in numerous real-world situations, ranging from organisational economics to “crowdsourcing” and disaster response. We show that the problem faced by the central planner can be transformed into a mixed integer bilevel optimisation problem. Then we demonstrate how this program can be reduced to a disjoint bilinear program, which is much more manageable computationally.

Global Manhunt Pushes the Limits of Social Mobilization

It is well known that efficient use of congestible resources can be achieved via marginal pricing; however, payments collected from the agents generate a budget surplus, which reduces social welfare. We show that an asymptotically first-best solution in the number of agents can be achieved by the appropriate redistribution of the budget surplus back to the agents.

Computing pure Bayesian-Nash equilibria in games with finite actions and continuous types

We present an algorithm for the penalized multiple choice knapsack problem (PMCKP), a combination of the more common penalized knapsack problem (PKP) and multiple choice knapsack problem (MCKP). Our approach is to converts a PMCKP into a PKP using a previously known transformation between MCKP and KP, and then solve the PKP greedily. For PMCKPs with well-behaved penalty functions, our algorithm is optimal for the linear relaxation of the problem.