I have been reading the sequence Multiagent Models of Mind by @Kaj_Sotala, and the Internal Family System concepts it discusses resemble something I have personally experimented with.

The sequence and its sources seem to focus on the use of IFS as a therapy/psychoanalysis tool, and I do not disagree with that use case and its importance. However, I have ended up taking the ideas of an IFS in a different direction: coaxing the internal agents into an architecture where their actions and reactions harmonize to produce system-wide benefits. I have not seen any posts here regarding this idea, and thought it worth sharing.

It would be quite a rabbit hole to go too far into where I have taken those experiments, so I have selected an architectural concept to represent the conceptual idea. Keep in mind that this sort of system takes a large amount of awareness regarding your personal symbolism, worldview, and computational preferences. It is also difficult to yield major benefits without applying patches to your System I programming.

From an outsider's perspective, I would describe my IFS as being based upon an Agile programming team, with roles from Scrum inspiring some of the specialization of the internal agents and XP (Extreme Programming) serving as the normal operating state. The agents perform actions based upon the XP development cycle, developing tests and designing solutions to pass those tests before prototyping the ones that seem most promising. If the solution passes the tests, it gets stripped of any unnecessary components and goes to integration, after which it becomes part of the baseline.

Also, I should note that I am still unfamiliar with much of the normal terminology around here. Feel free to ask if any of my phrasings or word choice seem strange.

New to LessWrong?

New Comment