https://github.com/aindilis/audience/tree/main/Audience/Dialog
is a quickly-written conversation simulator.
It allows you to model/edit/save/load a simulated conversation as a game with moves (and responses), and possible transpositions between states. Dialog in general is a well studied area of AI, but this system is super simplistic and was just used to train myself in prepraration for difficult dialogs, such as job interviews, or in this case, pay negotiation.
There are much more interesting AI dialog systems we're working on. LogicMoo has one, and I've recently begun trying to set up this contingent dialog executor system:
https://github.com/IBM/contingent-plan-executor
But in general, the Audience system, which I've released today
https://github.com/aindilis/audience
was supposed to centrally handle all of my communication, however, it kind of fell short of that a long time ago. It was supposed to say who I was allowed to talk with and what subjects were permissible. For instance, I used to talk about Artificial Intelligence too much, so for many people AI would be a prohibited topic. Audience was supposed to augment my communication with more social intelligence. Having a tap on all my comms would provide more discipline to my communication. I will probably revive it at some point and enhance it a lot.
There are a few components of Audience I still currently make regular use of. I often write a memcon (memorandum of conversation) after an important conversation, or I can author letters to contacts quickly. My IRC software additions, such as text to speech, are part of Audience, although I released them separately on github. Audience has a talking point system that primes me for what to talk about, and the FreeLifePlanner version includes software to remind me to ask any unanswered questions I have for the person. Audience includes an XMPP instant messaging client, and sends messages to my phone for the FreeLifePlanner.
There was a useful Audience program I used to use before I switched to gmail called ball-in-court which figured out whose turn it was in ongoing back-and-forth conversations by email. I also planned to use natural language processing to codify communications into a political event ontology like TABARI, but enhanced for more interpersonal communciation styles, in order to make better judgments about how best to respond. I'd also like Audience to realize who knows what, based on communications, and to help me reason with that information better.
I will probably write here about its current and future capabilities here in the future.
See Also: