I've spent so much of my adult life depressed that it has unfortunately formed too much of my self image. Separating it out into it's own identity, and then calling it dysfunctional and undesirable is helpful in overcoming some of that.
I said that I'm not really interested in continuing to live; I might get over this feeling if I manage to keep my depression in check for several more years. Right now, knowing that I will (statistically) soon feel like killing myself is that tiny note of discord, and makes me doubt any desire I have to affirm life.
I would create an agenda when in 'neutral' mode to accomplish, and then maximize the positive effects and minimize the negative effects that each mood has on the realization of that agenda
The primary way I achieve this agenda is to prevent either of the other agents from taking control. The fallback mechanism is the compromise I mentioned. How can Neutral modify their behavior if they won't cooperate? If they do cooperate, isn't that necessarily an agreement?
Of course, Neutral and Hypomaniac can always be insincere to depressed
Though this is a joke, we at Less Wrong often talk about agents with access to each other's source code. This is sort of a restricted example of that. If one of my agents defects, and says "my finger slipped", the other agents know immediately whether the defection was indeed accidental. Rarely is this possible at all, much less so convenient!
How can Neutral modify their behavior if they won't cooperate?
Do they exist as distinct entities or are they ideas? Neutral can unimagine them. Or, rather, you can unimagine all of them. Dictate them out of your head.
Are you actually three agents? Have you tried being four agents or two? Does that make sense as a question to ask?
Have you tried any other treatment for depression as depression, not as bipolar disorder - e.g. cognitive behavioural therapy?
Related to: Akrasia as a collective action problem and Self-empathy as a source of "willpower".
The Less Wrong community has discussed negotiating with one's conflicting sub-agents as a method to defeat akrasia and other forms of dynamic inconsistency, with some mix of reactions about how possible or effective that strategy can be. This article presents a successful example in my life, though it is probably an extreme outlier for a number of reasons.
I have been diagnosed with bipolar II disorder. It is one of the most significant challenges in my life, and certainly the one with the most dire implications. I can be fairly well modeled as three major sub-agents1:
Neutral feels it necessary to let Hypomanic take control more often to ensure that the compromise has weight to Depressed, but has started using Hypomanic to accomplish goals that are otherwise too exhausting to attain (a several-day code crunch or a need to meet and make a good impression on dozens of people). Meanwhile, Hypomanic has been more responsible lately in relinquishing control within days rather than weeks, partially because of these negotiations, but mostly because of other people in my life who have been conscripted to help monitor and rein me in.
I do not have a great deal of proven success with this strategy. I started doing this less than a year ago, and have not dealt with a full-blown major depressive episode since then. During that time I have also been more successful than ever at preventing myself from slipping into depression in the first place and treating early depression aggressively. In the end, that makes a much more significant difference, but on the two occasions when I became depressed enough to start feeling suicidal I was positively influenced by this agreement.
It seems unlikely that this approach will help many people with anything, but I feel like it is interesting in the debate about dynamic inconsistency, and I encourage others to find mutually-beneficial agreements they can make with themselves if they also feel like they deal with mutually incompatible agents from time to time. Also, this is my first post that is more than a link, so please be constructive.
Notes
1 I've never used names to refer to myself in different states, and don't think of my major sub-agents as individuals, but I felt that it was useful for didactic purposes to refer to myself in different states as different proper nouns.
2 I don't race cars, do drugs, or get in fights (except at the dojo). I do push my physical limits farther than I should (do parkour that I'm not be ready for, run 20km when I usually run 5, etc.), and I have injured myself this way, but just pulled muscles, sprains and once a broken finger.
3 I haven't heard this argument before, but this is the reason I haven't signed up for cryonics.
If it's not obvious, I was in a neutral state when I wrote this. It would have been impossible for me to do while depressed, and unlikely for me to try while hypomanic. I tried to de-bias myself, but no matter what state I'm in, I prefer my own viewpoint, and speak less highly of the others that diverge.