V. V. Abramov
The article provides theoretical-game model of voting in decentralized autonomous organization, where honest participants and evil agents meaning harm to the system strategically interact. To restrain harmful behavior mechanism of reputation tokens, i.e. intangible assets accumulated for conforming voting and lost in case of inactivity or confronting decision. The access to voting is given only when the minimum reputation threshold is exceeded. The model shows the introduction of reputation can transform one-step dilemma of participation into dynamic game. The key result is identification of two principle types of balance: mixing one, when evil agents behave like honest for a certain period of time in order to accumulate influence for the future attack and separating one, when they reveal their type quickly and are expelled from the system. Analysis shows that mechanism effectiveness depends drastically on its parameters (amount of rewards and fines) and informational structure: complete information of agents about proposal value can raise effectiveness of goal-oriented attacks. On the basis of this analysis recommendations were provided for designing sustainable systems, including the necessity to combine reputation with other mechanisms (quorum, delegation) and adjust parameters with regard to the share of evil agents.