Proportional Participatory Budgeting with Projects Interaction

TL;DR

The paper addresses proportional participatory budgeting when projects interact, a setting where dependencies challenge traditional additive-utility guarantees. It extends the Method of Equal Shares with Interaction Equal Shares (IES) and Partition IES (PIES) to guarantee proportional welfare under substitution dependencies, introducing EJRI-1 and EJRI-z notions. IES provides polynomial-time EJRI-1 guarantees for substitute interactions, while PIES offers a more relaxed EJRI-z guarantee for arbitrary interactions, with simulations showing higher social welfare than interaction-ignorant baselines. The authors also extend the pabutools library to support project interactions and demonstrate that, in practice, IES improves welfare on average, while PIES trades welfare variance for broader proportionality. These results advance proportional budgeting in realistic scenarios with dependencies, offering practically computable mechanisms and a path toward real-world deployment.

Abstract

Participatory budgeting (PB) is a democratic process for allocating funds to projects based on the votes of community members. PB outcomes are commonly evaluated for how they reflect voters preferences (e.g., social welfare) and the extent to which they are fair (e.g., proportionality). Due to practical and computational reasons, voters are usually asked to report their preferences over projects separately, possibly neglecting important dependencies among projects, which causes the outcome to no longer be proportional and achieve lower satisfaction. This work is the first to suggest a polynomial-time aggregation method capable of guaranteeing proportional outcomes under substitution dependencies. The method is a variant of the Method of Equal Shares, and we further provide another variation that can guarantee a more relaxed notion of proportionality for any type of dependency, and is FPT rather than polynomial. Through simulations, we demonstrate that these aggregation methods achieve, on average, higher social welfare than their counterparts that ignore the dependencies.