Vanilla reputation systems don't optimize for "validity"; rather, they optimize for consensus--whatever opinions are popular get hyper-reinforced in tight feedback loops. If you believe that our society is irredeemably racist, then a vanilla reputation system will optimize for racism. Basically, the reputation system can't dictate what is valid--that has to be determined outside of the reputation system. In other words, set a clear, objective standard for "valid" content and use a reputation system that holds moderators accountable to that external definition of "validity".