Edith Elkind

Recent breakthroughs in generative artificial intelligence (AI) and large language models (LLMs) unravel new capabilities for AI personal assistants to overcome cognitive bandwidth limitations of humans, providing decision support or even direct representation of abstained human voters at large scale. However, the quality of this representation and what underlying biases manifest when delegating collective decision making to LLMs is an alarming and timely challenge to tackle. By rigorously emulating more than >50K LLM voting personas in 363 real-world voting elections, we disentangle how AI-generated choices differ from human choices and how this affects collective decision outcomes. Complex preferential ballot formats show significant inconsistencies compared to simpler majoritarian elections, which demonstrate higher consistency. Strikingly, proportional ballot aggregation methods such as equal shares prove to be a win-win: fairer voting outcomes for humans and fairer AI representation, especially for voters likely to abstain. This novel underlying relationship proves paramount for building democratic resilience in scenarios of low voters turnout by voter fatigue: abstained voters are mitigated via AI representatives that recover representative and fair voting outcomes. These interdisciplinary insights provide decision support to policymakers and citizens for developing safeguards and policies for risks of using AI in democratic innovations.

We study two stylized, multi-agent models aimed at investing a limited, indivisible resource in public transportation. In the first model, we face the decision of which potential stops to open along a (e.g., bus) path, given agents' travel demands. While it is known that utilitarian optimal solutions can be identified in polynomial time, we find that computing approximately optimal solutions with respect to egalitarian welfare is NP-complete. This is surprising as we operate on the simple topology of a line graph. In the second model, agents navigate a more complex network modeled by a weighted graph where edge weights represent distances. We face the decision of improving travel time along a fixed number of edges. We provide a polynomial-time algorithm that combines Dijkstra's algorithm with a dynamical program to find the optimal decision for one or two agents. By contrast, if the number of agents is variable, we find \np-completeness and inapproximability results for utilitarian and egalitarian welfare. Moreover, we demonstrate implications of our results for a related model of railway network design.

Shortlisting is the process of selecting a subset of alternatives from a larger pool for further consideration or final decision-making. It is widely applied in social choice and multi-agent system scenarios. The growing demand for participatory decision-making and the continuously expanding space of candidates create an urgent need for efficient and fair shortlisting procedures. However, little principled study has been done on this problem. This blue-sky paper aims to highlight the overlooked significance of shortlisting, distinguish it from related problems, provide initial thoughts, and, more importantly, serve as a call to arms. We envision that principled shortlisting can reduce cognitive burden, enable fair collective decisions, encourage broader participation, and ultimately build trust in democratic systems.

In strategic candidacy games, both voters and candidates have preferences regarding the election results, and candidates strategically decide whether to participate or abstain, in order to sway the outcome in their preferred direction. Often, it is assumed that candidates are keen to be elected or even just show up for visibility purposes, unless their presence is harmful. However, there may be candidates who find it costly to run an electoral campaign in the first place and therefore prefer to withdraw if their participation has no effect on the election outcome (a.k.a. lazy candidates), or those who would rather not win the election due to perceived interpersonal or reputational risks, unless they see no better alternative (a.k.a. reluctant leaders). In this work, we open the study of reluctance in strategic candidacy, and extend the standard candidacy game model to incorporate the concepts of reluctant participation (lazy candidacy) and reluctant leadership. We examine the pure-strategy Nash equilibria and the outcomes of natural iterative dynamic processes in such games, both from a normative and from a computational perspective, and compare them with their counterparts in the conventional model.