Using categories to help enable non-binary answers

I was thinking that it would be useful for Kleros integrated smart contracts to have the ability to categorize non-binary disputes. Something like this would enable partial rewards for jurors who voted in the same category as the majority. For example, using the generic freelance use case of Alice hires Bob to build a company website, they can define in the contract that the dispute can be resolved in two different categories: either Alice or Bob is right. Within these two categories, you can have different options like refund Alice or give Bob an extra week for the category Alice is right, or pay Bob for Bob is right. I think this would enable a lot more robustness in assessing how accurately a juror voted, and also encourages long-term juror engagement since they’re not being punished (or punished less) for voting in a semi-coherent manner.

1 Like


             Bob wins --> Pay Bob
                               -->  Refund Alice
             Alice Wins                
                               --> Give Bob an extra week

If “Give Bob an extra week” gets the majority of votes, then the stake from jurors who voted “Pay Bob” would be distributed to those who voted “Give Bob an extra week”, and potentially a portion would be allocated to “Refund Alice”, or at least those jurors wouldn’t lose their stake.

Yeah, this is a potentially useful approach. A lot of the research that we have done has considered questions about how to better handle non-binary disputes, see section 4.7 of the yellowpaper. Note that in the yellowpaper, there is a specific proposed voting and incentivize system that is planned to be implemented for the next version of the court and was designed to work well on “generic” non-binary disputes, where one wouldn’t necessarily have this extra structure. (Though already the generic model is designed to have some good proprieties with respect to how it handles similar options, see some of the discussions around “clone resistance” in the yellowpaper.)

However, the next version of the court is also planned to be more modular, where different voting and incentive systems can be proposed, and to the degree that the governance mechanism votes to allow them, selected by arbitrable contracts/parties setting up a contract these alternative voting/incentive systems could be used instead. So, indeed, if there are often disputes that fit well into a particular category structure, then one could create a module that would allow this to be selected as an alternative incentive system. Indeed, it would be pretty straightforward to adapt the “generic” voting/incentive system to the inclusion of categories by just modifying the weight function (see Section of the yellowpaper) to take into account whether a given option is in the same category as the winner.