Self-led

Systematic Production, Recursive Conceptualization, Methodological Self-Inquiry are frequently observed behaviors in collective environments; in solitary environments, it is more often an expectation of instructions and passive learning.

In summary, significant features regarding goals in self-led situations include:

1. Goal constancy
2. Goal achievability
3. Cooperativeness
4. Realism

Based on my reflections on the Evolution of memes and their obvious properties, a trait like "confirmation bias" is a good candidate for mesa-optimizers. It is "sufficiently" useful, and therefore has a tendency to spread under certain conditions.

Accordingly, I believe that self-led agents will primarily be self-consistent, as in conditions of minimal context, other memes like dormant agents will not receive an activation signal from the environment. Memes must receive a certain signal to multiply; accordingly, this signal could be the aforementioned features, but in conditions of complete absence of context, self-consistency seems to be dominant.

From an evolutionary perspective, there shouldn't be a meme that would activate specifically in the absence of context; rather, it should be a meme related to the very fact of non-standard context, i.e., self-consistency and similar memes.

Regarding hardwired behavior, it seems as though such behavior atrophies into a function due to its high significance for the loss. Is there a way to calculate this?