🤖 AI Summary
To address the underestimation of weak players’ contributions and insufficient coalition stability in big boss games, this paper proposes the Projection Shapley Value (PSV): an orthogonal projection of the Shapley value onto the diagonal of the core, yielding an optimal approximation that balances the big boss’s payoff demands with weak players’ stability requirements. Key contributions include: (i) the first systematic characterization of structural differences between the τ-value and the Shapley value in big boss games; (ii) the introduction and uniqueness proof of PSV as the core-constrained optimal solution; (iii) an equivalent characterization of convexity for big boss games; and (iv) theoretical demonstration that PSV minimizes statistical distance to the Shapley value—ensuring robustness and representativeness even when the Shapley value is computationally intractable. Empirical evaluation confirms that PSV significantly enhances coalition stability.
📝 Abstract
Big Boss Games represent a specific class of cooperative games where a single veto player, known as the Big Boss, plays a central role in determining resource allocation and maintaining coalition stability. In this paper, we introduce a novel allocation scheme for Big Boss games, based on two classical solution concepts: the Shapley value and the $ au$-value. This scheme generates a coalitionally stable allocation that effectively accounts for the contributions of weaker players. Specifically, we consider a diagonal of the core that includes the Big Boss's maximum aspirations, the $ au$-value, and those of the weaker players. From these allocations, we select the one that is closest to the Shapley value, referred to as the Projected Shapley Value allocation (PSV allocation). Through our analysis, we identify a new property of Big Boss games, particularly the relationship between the allocation discrepancies assigned by the $ au$-value and the Shapley value, with a particular focus on the Big Boss and the other players. Additionally, we provide a new characterization of convexity within this context. Finally, we conduct a statistical analysis to assess the position of the PSV allocation within the core, especially in cases where computing the Shapley value is computationally challenging.