The "context window" analogy for human minds

post by Ruby · 2024-02-13T19:29:10.387Z · LW · GW · 0 comments

Contents

No comments

Years ago I already modeled (and wrote about [LW · GW]) that emotions are mediated by attention. You have emotions about what's on your mind (or recently on your mind) rather than having emotions constantly about everything that you technically know and believe.

Recently, I've gotten a good deal of mileage from porting over the concept of "context window" to my own mind. It's kind of the same idea, but somehow makes use of more intuitions.

I see my current emotions, and more than that–my current evaluations and leanings on questions of what to believe and do (of course, emotions inform these) to be quite dependent on what's loaded into my context window.

A useful intuition is that my context window is only so large, and for sufficiently large and complex topics, I'm not capable of loading everything into my context at once (at least not without chunking, external memory at aids). This means there are now classes of decision that previously I might have attempted to resolve while walking somewhere, showering, falling asleep, that I'll now refrain from making any conclusion on until I've sat down with exobrain, in good circumstances, and tried to load all considerations into my context window in a short space of time. 

I've found that some known techniques achieve this, namely anything "inner parts" related. Internal Double Crux, or just "writing each perspective back and forth on an issue" seems to help, at least in part via the load-it-all-into-context mechanism.

A different principle that I get motivation for downstream of this model is that taking my time with complex decisions. "Sleeping on it." This helps because it allows for different info to naturally cycle through my context window. (Especially relevant is that your present emotional state is part of the context window that feeds into beliefs, and that varies over time.)

The context window analogy assists with holding emotions as object and not necessarily strong evidence about reality, as I see strong emotions often being the result of just the current context window contents, and sure enough, as my attention shifts, so do emotions. (Which doesn't mean the emotions are informative, just not the whole picture.)

Some things are simple enough that they're easy to load into memory at once and just make a quick(ish) decision. Before I was using the context window analogy, I had briefly updated against making any big decision without "proper process", but for simpler, less internally fractious matters, I worry less.

Credit goes to @Steven Byrnes [LW · GW] for this analogy. Somehow his example in his Valence series [LW · GW] made this click to me. He describes that your valence associated with going to the gym is dependent on what you're thinking about it ("keeping new year's resolution" vs "it's smelly". Somehow that made it crisp for me just how much judgments are tied to context window.

Hmm, a thought I'm having just now is my guess for what happens for many people on many topics, is the topic itself whenever raised, comes with a mix of cached affect and cached relevant info that gets loaded/primed for the context window whenever it comes up. This will result in a feeling of consistency (and reliability) of judgment and emotion about a topic. (And the more times you've felt/judged that way, the more confident you feel.) It might be interesting to sometimes flush the cache on repeat-topics and load alternative data into the context window.

-

0 comments

Comments sorted by top scores.