Problem

Democracies worldwide face increasing challenges, including difficulty scaling up participation, the spread of misinformation and fake news, and the growing influence of generative AI in elections and public opinion formation. Traditional majoritarian voting methods are identified as a key vulnerability, often failing to represent diverse societies accurately, leading to polarization, extremism, and the suppression of minority voices. This over-reliance on simple majority rule has rendered democracies unprepared for modern complexities and emerging threats, particularly regarding the integrity of collective decision-making processes.

Method

This research introduces fair voting methods as a solution to enhance democratic resilience, drawing on real-world evidence from democratic innovations, primarily in participatory budgeting in Switzerland (specifically Aarau). The study rigorously tested two fair voting methods: cumulative voting (multi-choice, point-based ballot format) and equal shares (proportional representation ballot aggregation). A systematic approach was employed, involving pre- and post-voting surveys, to compare outcomes and citizen preferences between fair and standard majoritarian methods. The research also constructed over 50,000 AI voting personas, based on various large language models and predictive machine learning, to assess the resilience of fair voting methods against AI biases and inconsistencies in scenarios involving AI voting assistance or representation of abstaining voters across 363 participatory budgeting and national election scenarios.

Results

The findings present a "trilogy of democratic upgrades" achieved through fair voting methods: Firstly, they act as a "legitimacy incubator," resulting in 23.8% more winning projects for the same budget, greater representation of citizens' preferences (21.9% average increase), and improved spatial fairness. Citizens, including those whose preferred projects didn't win, preferred and found these methods fairer after voting, indicating a strengthening of democratic values like altruism and compromise. Secondly, they serve as a "novel impact accelerator." Projects proposed under fair voting methods were, on average, 48.4% lower in cost, leading to the selection of more diverse winning projects focused on welfare, education, and culture, areas often underrepresented by traditional methods. This shift discourages costly infrastructural projects and promotes cost-effectiveness. Thirdly, fair voting methods function as an "AI safeguard." Studies showed they are more resilient to biases and inconsistencies from generative AI in scenarios of AI voting assistance or representation of abstained voters. Outcomes from fair methods maintained consistency despite inaccurate AI estimates, unlike majoritarian methods, and were more effective in preserving winning projects facilitated by AI representation of abstained voters.

Implications

The study highlights the potential of fair voting methods to rebuild democratic resilience, particularly in democracies facing crises, such as Greece, where citizens perceive a significant democratic deficit. These methods align with citizens' desires for more proportional representation and direct democratic innovations like participatory budgeting, digital participation, e-voting, and referenda. While citizens often prioritize top-down systemic changes (e.g., combating corruption, ensuring justice), the research suggests a bottom-up approach through fair voting methods at a local level can be a practical starting point for broader democratic transformation. Furthermore, given citizens' low awareness of AI risks but high expectations for AI's positive role in democracy, fair voting methods are crucial safeguards to ensure that AI integration meets these expectations without undermining electoral integrity. The evidence suggests a growing momentum for these methods to scale up and wide, influencing various budgeting and governance processes beyond local participatory initiatives.