I can't figure out if this is a good blogpost topic or not, but I've been thinking about how many conversations I see about human behavior in software overindex on like, differences between people* and not within-individual variation**
Overall malleability of our own traits and states over time is fascinating and underexplored in a very essentialist kind of culture***
* "all managers are like x"
** "some days I am like x and some days I am like y"
*** I find tech to be very essentialist
I would really like to read something on this topic, esp. if it was written in a way friendly to people who have a reductionist approach to reality.
I wonder where would you put within-individual variation that's a result of slow but consistent drift over time (it's within-individual, but is not "some days are xy and some are zy").
I meant the difference between variability (i.e. noisy/oscillating changes) vs changes that have a trend. Developmental changes over significant fraction of lifespan are one example of the latter; the example I was thinking of was changes due to social environment one enters (that, I'd anecdotally estimate, happen on the scale of months to years). (BTW. I think that not all of the former -- variability -- kind of changes are quick: for example depression episodes -- if I understand their typical progression correctly -- are examples of slow variability).
I think I am such a person, so let me answer these questions about me. First, let me respond in a way that's topic-agnostic (I think it's helpful, because most of my learning experience was about subject that don't describe human behaviour). I hope that reading about a phenomenon will let me develop a model of it and lets me associate pieces of that model with evidence for their correctness, or will show me that the model I already have for it is incorrect (and will also help me create a correct, though often less predictive, model)[1]. That means for me that I can (at least for somewhat idealized situations) predict observations and have some handle of (at least relative) accuracy/certainty of such predictions.
Often (even in much less noisy areas than ones that study humans) having a reasonably complete model is infeasible. The second best thing then is having what physicists calls symmetries: sets of changes in the modelled phenomenon that, when applied together, do not change outcomes. (BTW. Even if we can have a ~complete model, such symmetries are often part of a nicely structured argument for its correctness.)
For some parts of the phenomenon, neither of the two is feasible. Then we often can still make "directional" statements: that changes in one parameter in some direction change some outcome in a particular direction.
[1] Most insightful things I consciously learned about human behaviour was from the model breaking bucket (e.g. that people ~never communicate the literal meaning of what they say, or that gender is not purely social).