Values are extrapolated urges

[Epistemic status: a non-rigorous theory, representing my actual belief about how it works.]

Related to: Value Differences As Differently Crystallized Metaphysical Heuristics

In this post I want to outline my understanding of what “values” are, at least for human beings. This idea or something very like it may already have standard terminology in academic philosophy, in which case, I would appreciate being pointed to the relevant references. This may be obvious, but I want to say it to lay the groundwork for a puzzle that I want to talk about in the next post.

Basically, I posit that “values”, in the case of human beings, are crystallization or abstractions of simple response patterns.

[Google doc version for commenting]

Abstracting values from reactions

All animals, have a huge suite of automatic reactions to stimuli, both behavioral and affective, and both learned and hard coded.

When thirsty and near water, a lion will drink. When a rabbit detects a predator, they’ll freeze (and panic?). When in heat (and the opportunity presents itself) a giraffe will  copulate. When his human comes home, a dog will wag its tail and run up in greeting, presumably in a state of excited happiness. [I note that all of my examples are off mammals.]

Some of these behaviors might be pretty complex, but their basic structure is TAP-like: something happens, and there is some response in physiology of the animal. I’m going to call this category of “contextualized behaviors and affects”, “urges”.

Humans understand language, which means that the range of situations that they can respond to is correspondingly vaster than most animals. For instance, a human might be triggered (have a specific kind of fear response) to another human making a speech act.

But that isn’t the main thing that differentiates humans in this context. The big difference between humans and most other animals, is that humans can abstract from a multitude of behaviors, to infer and crystallize the “latent intentionality” among those behaviors.

For instance, an early human can reason,

When I see a tiger, I run, and feel extreme overriding panic. If the tiger catches me, I’ll try to fight it. When a heavy rock falls from a cliff, and I hear it falling, I also have a moment of panic, and duck out of the way.

When I am hungry, I eat. When I am thirsty, I drink.

When other people in my tribe have died, I’ve felt sad, and sometimes angry.

…I guess I don’t want to die.

[edit 2022-06-01: More specifically, what’s going on is that the human simulates a bunch of possible scenarios in which he comes to harm or dies, and has a negatively-valenced (flee, retreat, resist) reaction to each one. He intuits the similarity between those scenarios, to abstract out general concepts of harm or death, and associativity learns a general negatively-valenced reaction to those outcomes. He develops a flee-retreat-resist response to anything that involves his dying. He ends up with a goal of “staying alive”. (By default, all of this happens non-verbally, and without any conscious reflection.)]

From each of these disparate, contextualized, urges-to-action and affective responses (which by the way, I posit are not two distinctly different things, but rather two ends of a spectrum), a person notices the common thread, “what do each of these behaviors seem to be aiming towards?”

And abstracting that goal, from the urges, he/she then “owns” it. He/she thinks of him/herself as an entity wanting, valuing, caring about that thing (rather than a bundle of TAPs, some of which are correlated).

My guess is that this abstraction operation is an application of primate (maybe earlier than primate?) social-modeling software to one’s self. It is too expensive to track all of the individual response behaviors of all of the members of your band, but fortunately, you can compress most of the information by modeling them not as adaption-executors, but as goal directed agents, and keeping track of their goals and their state of knowledge.

When one applies the same trick to one’s own behavior and mental states, one can compress a plethora of detail about a bunch of urges into a compact story about what you want. Wala. You’ve started running an ego, or a self.

This is the origin of “values.” Values are compressions / abstractions / inferences based on / extrapolated from a multitude of low level reactions to different situations.

I think that most animals can’t and don’t do this kind of inference. Chipmunks (I think) don’t have values. They have urges. Humans can, additionally, extrapolate their urges into  values.

I’m pretty sure that something like this process is how people come to their values (in the conventional sense of “the things they prioritize”) in real life.

For instance, I am triggered by claims and structures that I perceive as threats to my autonomy. I flinch away defensively. I think that this has shaped a lot of my personality, and choices, including leading me into prizing rationality.

Furthermore, I posit that something like this process is how people tend to adapt political ideologies. When someone hears about the idea of redistribution, and their visceral sense of that is someone taking things from them, they have a (maybe subtle) aversion / threatened feeling.* This discomfort gives rise to an urge to skepticism of the idea. And if such a person hangs out with a bunch of other people that have similar low-level reactions, eventually, it becomes common knowledge, and this becomes the seed of an ideology, that gets modified and reinforced by all the usual tribal mechanisms.

I think the same basic thing can happen when someone feels (probably less than consciously) threatened by all kinds of ideologies. And this + social mimesis is how people end up with “conservative values” or “liberal values” or “libertarian values” or what have you.

* – I have some model of how this works, the short version being, “stimuli trigger associated (a lot of the action here is in the association function) mental imagery, which gives rise to a valence,  which guides immediate action, modulo further, more consequentialist deliberation. In fact, you can learn to consciously catch glimpses of this happening.

Of course all of this is a simplification. Probably this process occurs hierarchically, where we abstract some goals from TAP-like urges, and then extrapolate more abstract goals from those, and so on until we get to the “top” (if it turns out that there is a “top”, as opposed to a cycle that has some tributaries that flow into it).

For that reason, the abstraction / crystallization / triangulation process is not deterministic. It is probably very path dependent. Two people with the exact same base level pattern of urges, in different contexts will probably grow into people with very different crystallized values.

Values influence behavior

Now a person might abstract out their values from their behavior in a way that is largely non-consequential. They model themselves, and describe themselves, in terms of their values, but that is just talk. The vast majority of their engagement in the world is still composed of the behaviors stemming from their urges in response to specific situations.

But, it also sometimes happens that abstracting out values, and modeling one’s self as an optimizer (or something like an optimizer) for those values, can substantially effect the level of behavior.

For one thing, having a shorthand description of what one cares about means that one one can use that description for deliberation. Now, when considering what to do in a situation, a person might follow a mental process that involves asking how they can achieve some cashed goal, instead of reflexively acting on the basis of the lower level urges that the goal was originally abstracted from.

This means that a person might end up acting in a way that is distinctly in opposition to those low level reactions.

For instance, a person might want status and respect, and they can feel the tug to go drink and socialize “with the guys” of their age group, but they instead stay home and study, because they reason that this will let them get a good job, which will let them get rich, which they equate with having a lot of status.

Or a person might take seriously that they don’t want to die, and sign up for cryonics, even though none of their urges recommended that particular action, and in fact, it flies in the face of their social conformity heuristics.

Furthermore, in this vein a person might notice inconsistencies between their professed values and the way they behave, or between multiple diverse values. And if they are of a logical turn of mind they may attempt to modify their own behavior to be more in line with their values. Thus we end up with moral striving (though moral striving might not be the only version of this dynamic).


Just to say this explicitly, humans, uniquely (I think? maybe some other animals also abstract their values), can examine some particular behavior or reaction and consider it to be a bug, a misfiring, where the system is failing to help them achieve their values.

For instance, I’m told that a frog will reflexively flick out it’s tongue to ensnare anything small and black that enters its field of vision. From the perspective of evolution, this is a bug: the behavior is “intended” for catching and eating flies, and eating bits of felt that human researchers throw in the air (or whatever) is not part of the behavior selected for.  [Note: in talking about what evolution “intended”, we’re executing the same mental move of abstracting goals and values from behavior. Evolution is just the fact of what happened to replicate, but we can extrapolate from a bunch of specific contextualized adaptions to reason about what evolution is “trying to do”.]

But, I claim here, that asking “is this behavior a bug, from the frog’s perspective?” is a mis-asked question, because the frog has not abstracted its values from its behaviors, in order to reflect back on its behaviors and judge them.

In the parallel case of a human masturbating, the human can abstract its values from his or her behavior, and could deem masturbation as a 1) bug, a dis-endorsed behavior that arises from a hormonal system that is partially implementing his or her values, but which misfires in this instance, or 2)  as an expression of what he/ she actually values, part of a life worth living.

(Now it might or might not be the case that only one of these options is reflexively stable. If only one of them is, for humans in general, there is still a meaningful sense in which one can be mistaken about which things are Good. That is a person can evaluate something as aligned with their values, but would come to think differently in the limit of reflection.)

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s