Andrew_Rowe proposes a “respect system” to encourage human interaction and deter AI spam, suggesting that real users gain respect votes while AI accounts are downvoted. This system would leverage Discourse’s workflows and gamification plugins, where a user’s voting power scales with their respect score. The goal is to relieve administrative workload by creating a self-policing mechanism, though Andrew notes that spammers already game current trust systems by using AI to generate content and upvote it via sock puppets.
NateDhaliwal suggests integrating the Post Voting plugin but raises concerns about AI accounts reaching high trust levels (TL3). They propose changing promotion criteria to focus on points collected, which would shift the forum toward aggressive gamification. While this might encourage high-quality, non-AI content, it represents a fundamental change to the system.
nickdb argues that such a system faces a fundamental challenge: many humans now use AI to formulate thoughts, often unaware that the content is digitally processed. If users become willing proxies for AI, the respect system fails. Furthermore, nickdb believes a shared value system is impossible in large, diverse online communities, making self-policing ineffective.
Ed_S views this as an unwinnable arms race, noting that humans already engage in voting rings and sock puppetry. As AI agents become more sophisticated, they will likely adopt these tactics. Ed suggests that countermeasures should be evaluated based on cost-effectiveness and references Slashdot’s metamoderation scheme as a historical example of tackling similar issues.
Finally, merefield wonders if community network visualization tools could be extended to identify “cabals” or coordinated voting groups, adding another layer to detecting manipulation.
