In this situation, neither party need be aware that the other party is 'antisocial' in this regard. The way I see it, the following situations apply for a tool

In this situation, neither party need be aware that the other party is 'antisocial' in this regard. The way I see it, the following situations apply for a tool that produces communication indistinguishable from the user's own writing (AI)... "AI" for the following cases would be beneficial: Social -> Antisocial Social <- Antisocial "AI" for the following cases would be net neutral: Antisocial -> Antisocial Antisocial <- Antisocial And finally "AI" would be inapplicable (since neither party would use it) in the following cases: Social -> Social Social <- Social In this drastically simplified model, there aren't any cases where the existence of a sufficiently competent AI would be detrimental to any party involved, while still providing value for those who choose to use it.