When is positive self-talk not worth it due to self-delusion?

It's not about pumping up the positive side to counteract the negative side, this just leads to an arms race. It's about investigating the aversive side to understand what its original positive intent is, its motivation stack, to see the ways that those goals are not incompatible with other goals and that the confusion that they were incompatible was due to means-ends confusions (some strategies are incompatible, but the original goals are not). I recommend Core Transformation (by Connierae Andreas) for the best process I know of for doing this. CT teaches manually stack tracing the generators of thoughts and feelings, so that you can eventually do it more automatically.

I think human values emerge [LW · GW] from the multi agent dynamics associated with an ongoing inner alignment failure in the brain. In that context, it matters a lot what your different parts think of each other and of you as a whole.

We might make an analogy between positive self talk for an individual and patriotism / high social trust for a society. Such beliefs seem to pull on a lever we might call “tendency towards internal cooperation”.

There are good and bad aspects to pulling such a lever. More internal cooperation can mean less resources wasted on internal conflicts and greater focus on achieving external goals. However, internal competition can drive efficiency improvements or prevent stagnation, and sufficiently intense patriotism eventually devolves into delusional nationalism.

Psychology research often suffers from only being able to study entities that don't have much complexity. Positivity and negativity are both simple concepts. 

If you look at someone like Elon Musk, the lense of positivity and negativity muddies what's going on. When starting Tesla and SpaceX Elon gave both a 10% chance of success. You could deduce from that, that Elon is s pessimist instead of an optimist.

When it comes to your self-concept, having a self-concept that relates to reality is good. If you don't, you are likely flinch away from important data. The late Steve Andreas describes in Transform Your Self a good framework of how to actually build a positive and reality aligned self-image.

Transform Your Self is a theory with a bunch of moving parts that is more complex then the simple positive/negative and that's likely part of why there's no academic engagement with it.

Even if you would want to stay with the positive psychology interventions, "tell yourself you are awesome" is inferior to "have a gratitude journal". One of the aspects of a gratitude journal is also to relate to what actually happened in the real world.

Does it lead to action that accomplishes one's goals? Then it has worked. Does it lead to action that fails to do that? Then it is become a snare and a delusion.

Personally, I prefer to see things as they are and may be, and act to bring about how I want them to be.

A partial answer: 

• Your emotions are more negative than granted if, for instance, it's often the case that your anxiety is strong enough that it feels like you might die and you don't in fact die.
• Your emotions are more positive than granted if it's often the case that, for instance, you are excited about getting job offers "more than" you tend to get job offers.

These answers still have ambiguity though, in "more than" and in how many Bayes points your anxiety as a predictor of death actually gets.