I have thought some time now that we need a model for determining the difference between the actions of a person, the actions of software acting for the person,

I have thought some time now that we need a model for determining the difference between the actions of a person, the actions of software acting for the person, or software acting for it's own internal use. It's a bit of a tricky issue when technically all things people do on a computer are software assisted, but there is a clear divide between editing a file in a text editor and a program generating a thumbnail image for it's own use. Similarly there's a distinction on sending an email by pressing send and a bot sending you an email about a issue update. All in all, I would be ok with AIs being able to create issues if they could clearly do so through a mechanism that supported something like "AGENT=#id acting for USER=#id" People could choose whether or not to accept agent help.