Decentralized systems like blockchain networks require effective dispute resolution mechanisms to maintain trust and fairness. On-chain disputes are settled via smart contracts, while off-chain disputes often rely on solutions like crowdsourced juries. Decentralized justice platforms typically use token-based weighted selection and randomness in jury formation. Since many public blockchain networks exhibit a power-law distribution in token holdings, these processes introduce unfairness by favoring entities with more tokens. At a structural level within the blockchain's social network, a few nodes are selected more frequently, introducing bias in the jury selection process.


To address this issue, we present the Fair-Spread family of algorithms designed to ensure fairness in selecting a crowdsourced jury with memory enhancement that maintains a history of jury selections and probabilistic methods. We introduce novel metrics for assessing the quality of crowdsourced juries and perform a comparative analysis with various synthetic social network graphs and public datasets exhibiting power-law degree distribution. We conduct experiments with different network and jury sizes to assess algorithm performance, offering recommendations for the most suitable jury selection algorithm based on fairness and representation requirements.