Lack of Belief Propagation as a Defense Mechanism

A problem with human beliefs - verbally professed coherent statements about the world - is that we don't automatically apply their implications to all our other beliefs and behavior, and we end up behaving inconsistently.

Note the qualifier "coherent". There are lots of claimed beliefs about the world that don't amount to a coherent truth claim, or that alter understanding about the world. Many of these have to do with signaling social afiliation, especially with religion, nationality or ethnicity, and even sports teams. But this post is not about those.

Let's take an example: someone starts telling their fellow Southern California residents that they should prepare a kit and a plan for future fire evacuations, since destructive fires occur more often than earthquakes (this is especially true in San Diego County, given the path of the major faults). However this person does not himself prepare a kit. This is not because he actively thinks he's somehow safe from fires, but rather because it just never occurs to him to propagate his new belief to his other beliefs and behaviors. (As you might be guessing, this was me.)

The most severe failure to propagate beliefs occurs in delusions, where there is a fixed belief that is given absolute weight, and conflicting beliefs are discarded or interpreted so as not to be in conflict with it - when, that is, they're compared and noted to conflict. But relative to our long-term memory, we have small short-term (working) memory - we can't call all of our long-term declarative memories (our beliefs) into short-term memory to constantly update them. This mismatch is true of all humans, not just those of us with delusions, although it's possible that one reason people have delusions is because of a working memory deficit. Consequently, there's no guarantee that we've updated all our beliefs when we develop a new one, even if we consciously attempt to sort through all beliefs that the new one may impact. This failure to completely propagate beliefs is a sin of omission, and doesn't include some of the sins of comission we find in reasoning biases that prevent us from doing so, like confirmation bias.

Of course, beliefs can be false, and lead to bad decisions that damage the agent, and all humans have some false beliefs. Because we develop false beliefs, and what's more language allows other agents with ill-intent to implant false beliefs, it's worth considering that an immediate propagation of a new belief through all our beliefs and therefore behavior might not be in our best interest. You might make the argument that complete propagation would make the agent more likely to identity the new belief as false since it could be identified as inconsistent with the rest of teh agent's beliefs. The challenge is that we still can't guarantee all the other beliefs the agent has are true, if only because they're all developed with limited information.

Therefore, it's worth considering whether rapid propagation of beliefs might actually be selected against, because it produces naive agents more likely to be taken advantage of by other agents.