This post from Paul Graham, Y Combinator founder, is fantastic. I came across it two years ago and it’s stuck with me every since. Imagine how better things would be if people learned to disagree better.
The thesis is that, online, there’s bound to be more disagreement than agreement because there’s not much to add when you agree with someone. So, if that’s the case, we better start to learn to have better disagreements.
Enter the Disagreement Hierarchy (DH).
In short the disagreement hierarchy proceeds from DH0, the most base and least value adding form, to DH6, the most powerful form of logical refutation.
DH0: Name Calling
DH1: Ad Hominem
DH2: Responding to Tone
DH6: Refuting the Central Point
In talking with a colleague today, I realized that the disagreement hierarchy would be a wonderful think to try to code. It could be used to generate DH scores on popular review sites and blog comments, giving readers (and writers) a sense of how well reasoned a review or comment was and how on target it was with respect to the original argument. In the process, thoughtful reviewers could earn authority for delivering sound responses focused on issues while trolls, flamers and less thoughtful reviewers would be exposed by an objective measures.
Perhaps more importantly, such a capability could provide valuable feedback that might help well-intentioned people (not trolls, etc.) learn to compose more relevant responses, make their points more convincingly and ultimately raise the level of discourse online.
I have no doubt that building such a system with acceptable accuracy, range and nuance would be extremely difficult. There are plenty of automated paper graders (somewhat dated overview of here). These might be good places to build from.