Learning to exercise agency
Hey, you want to read a newsletter post in which I mostly provide a bunch of semi-novel conceptual distinctions?
Well, if you don’t, you can stop reading. That’s exercising your agency, which is what this post is about.
What is agency?
Exactly what constitutes agency seems to be a subject of active debate. I’m going to try to sketch out roughly what I think of as agency, but I don’t intend this to be a super precise definition or the definitive word on it.
Roughly, agency is the capacity to act to satisfy some preference.
This breaks down into three parts:
The capacity to act.
Possessing preferences over outcomes of those actions.
Being able to choose to use the former to satisfy the latter.
This isn’t a perfect definition and not intended as such, but it’s pretty good at highlighting the key characteristics. If any of these are absent, it’s probably not agency. If all of these are present and relevant, it’s probably agency.
For example:
Choosing to stop reading something because it’s not working for you is an exercise in agency. Failing to do so because you feel like it would be improper or show a lack of respect for the author even though you can and want to and agree it would be the right thing to do is a failure to exercise agency.
Noticing that something in your environment is broken and fixing it rather than just enduring it is an exercise in agency.
Noticing that you’re confused about something important and seeking to improve your understanding of it is also an exercise in agency.
Failing to get up and do the dishes and instead writing a newsletter post about agency, even when you acknowledge that the former is the correct thing for you to be doing right now, is also a failure to exercise agency, because it is not using your capacity for action to satisfy your preferences even when you could. Uh, hang on BRB.
OK, I’m back. Another good example is a thermostat1. You could argue that it has preferences, maybe, and it certainly has the capacity to act. But it only habitually responds, it cannot actually make decisions in aid of this. You can see this because if you break the heater it controls, it will still continue doing the action despite having no effect, and if you reverse the heating and cooling controls it will never notice this and will just keep on doing its thing.
Note that agency doesn’t necessarily require you to succeed. It’s perfectly possible to exercise your agency and fail to achieve your goal, either because you’re trying something too hard or in some way lacking in skill, and that’s not a failure to exhibit agency in and of itself. If you repeatedly keep doing the thing anyway without learning from your experience (as in the thermostat example), that likely is a failure to exhibit agency.
Failures of agency
A lot of why agency is interesting is because people often don’t exhibit it in cases where they would benefit from doing so. This isn’t a judgement on those people - “those people” very much includes me, which is much of why I’m interested in this subject.
I think the most central and distinctively agency-related way to have failures of agency are in the third part of the three part characterisation: Failure to use the capabilities you have to satisfy the preferences you have. Lacking capabilities is more a practical question of how you develop skills and such. Lacking preferences is more a question of how you develop emotional awareness of the preferences you already have.2 But having preferences and the ability to act to fulfill them and not doing that is weird, right? There's something interesting there to understand.
So this is what I’ll call a “failure of agency” - a case where there is a clear preference, and a clear ability to act on that preference, but no action occurs (and possibly the idea that you could act never even occurs to you).
There are a couple obvious classes of example here that I won’t consider a failure of agency though.
When you could act, but it would be punished. e.g. if you’re in an environment where showing initiative is treated as illegitimate and an authority figure will smack you down for it, when you fail to satisfy your preferences you’re not showing a lack of agency, you are correctly satisfying your preference to not be smacked down.
I think such environments can (and often do) lead to future failures of agency, but your behaviour in them is not in and of itself a failure of agency. I’ll get back to that in a bit.
Another thing that I don’t consider a failure of agency is when you could act but it’s too hard, or you don’t have a clue where to start. e.g. if your hot water system makes a funny noise, you probably want it to not do that. You could call a plumber, but how do you find a good plumber? Also this is potentially expensive. You could learn the relevant plumbing skills yourself, but this is hard and overwhelming. Given that the problem is not that bad and the solutions are all very hard, it’s not intrinsically a failure of agency to choose not to act on it, it might just be a perfectly reasonable trad off between competing preferences (“This noise is annoying” vs “fixing it seems like a lot of work”).
I do think there’s often a temptation to overestimate how much work things are, and that this can stem from a sort of learned helplessness, so the above sort of reason isn’t always legitimate, but it is probably fair more often than not, and I will mostly be interested in cases where acting on your preferences does not require developing new capabilities for action.
Specificity of agency
One of the things that I find very confusing about how a lot of people talk about agency is that they talk about agency as something you can have more or less of, talking about “high agency” and “low agency” people. This has never struck me as terribly helpful,3 because it's very unclear whether agency is something you can compare in that way, or what the specific differences being pointed to are.
That’s not to say that it never makes sense to compare two people’s agency. If you’re in a situation where someone is exhibiting a failure of agency and another person is not, it clearly makes sense to say that that latter person is showing more agency. But what doesn’t make sense to me is to then be surprised when this reverses in another situation and the previously “low agency” person is the one showing the higher level of agency. e.g. A skilled handyman may be “high agency” when repairing some plumbing and low agency when faced with a computer program crashing on him.
I think what’s happening is that those who talk about people being high or low agency are treating “agency” as if it were a thing like “strength” or “intelligence”, where although there are many different types of strength and many different types of intelligence, they all tend to display what’s called positive manifold - being good at one tends to indicate you’re good at another, albeit imperfectly. Alex might beat Pat at power lifting, and Pat could still beat Alex at arm wrestling, but on balance you’d expect it to happen more often that Alex beats Pat at both or neither than at only one, because although the world’s best arm wrestler is probably not the world’s best power lifter, an average strength person will probably lose out at both to a strong person, a weak person will probably lose out to an average person, etc.
In contrast I think agency is more like “skill”. It makes sense to talk about someone being more or less skilled than someone in a particular domain, but there is so much domain diversity that any given skill is at best very weakly predictive of any other. Kim might be more skilled than Charlie at Chess, while Charlie is much better at cooking French cuisine than Kim, and this combination isn’t even slightly surprising. If anything, it’s slightly more surprising to have someone excel at both than it is for them to excel at only one. At the same time, you’d probably quite reasonably expect that Kim is better at Go than Charlie, and Charlie is better than Kim at cooking Italian food - probably neither of them are experts in that, but they can draw on general skills from their existing expertise and give it a go.
Even to the degree that it makes sense to say that someone has “more” agency than another person, that’s not necessarily predictive of everything, in much the same way that I clearly have more skill than a 10 year old but that doesn’t necessarily mean I can sing better than them4. The label may be meaningful, but it's not usually useful without specifying a domain.
Generality of agency
That being said, I do think there are a number of general factors affecting agency, which are likely to generalise across domains.
The first is that I think there are many things that can impair agency.
Things labelled executive function disorders can roughly be thought of as being disorders of agency. Executive function is to a large degree the capacity for exerting agency in the face of multiple competing demands, so when it gets overwhelmed agency suffers.
Anxiety can also significantly impair agency - it’s much harder to act when you’re worried that everything is going to go terribly, horribly, wrong, if you do, so anxiety in a specific area will often reduce your agency in that area (e.g. anxiety about your technical skills will make you less willing to try things at work, social anxiety will make it harder to act on a desire to make friends, etc), a tendency towards anxiety will often result in less agency in a generalisable sense.
In the other direction, I think there are skills and dispositions which increase your agency in general. Practical wisdom, the skill of knowing that to do, is generally agency-increasing because it makes it easier to find paths to act on your preferences, so gets you used to acting in general.
Reflexive agency
Part of what started me on the line of thinking that lead to this newsletter issue is noticing that there was a particular characteristic of agency that I hadn’t seen highlighted before: Sometimes you will exhibit reflexive agency - agency as an automatic response to the situation you’re in. You have a preference, you have the ability to satisfy it, so you do that. What could be more natural?
Let me give you an example. In a previous flat, when I moved in the dining room table was extremely wobbly, to the point of being basically unusable. I asked about this, and was told “Oh yeah the table is just like that, has been since I moved in”. So I got a spanner, tightened the nuts holding the table together, and it stopped wobbling.5
This was very much not an example of me possessing more DIY skill than them. Literally anyone with a spanner could have done this. Probably anyone could have done an adequate job without a spanner just getting the nuts finger tight. All that was required was treating it as a problem to be solved, and then spending a couple seconds investigating how to solve it. Further, it had clearly been bothering people other than me for a very long time. And yet, I was the one to solve it.
Why? Well, because I was the one who went “Ah, right this seems like a problem. Why don’t I solve it?”
Why didn’t the others do that? I’m not sure. I suspect not doing that is just the default state of humanity, so doesn’t require explanation. The reason I was able to do that is probably because I had my dad as a model of the sort of person who behaves that way, and a gift of a tool box from him (I believe the flat already had a toolbox, but perhaps it didn’t feel salient in the same way).
I am not, of course, habitually able to exercise agency in DIY, as witnessed by the following conversation on discord in response to sharing a draft of this post:
drossbucket: The table anecdote reminded me of when I got a new flatmate and within an hour or two of moving in she'd picked up my saucepan, noticed the handle was wobbly and immediately got a screwdriver and fixed it. Rest of us in the house of 7 had all just been using it without thinking.
DRMacIver: uh brb
(I've been meaning to fix that for months but only remember at bad times to do so)
I suspect the common thread here is that it’s much easier to exhibit reflexive agency in response to a novel situation - something you’ve known about for a while just gets treated as part of the background, while something novel you can respond to.
Another example of this sort of thing that I am constantly frustrated with is something that I experience as other people being very lacking in curiosity, but from the inside is probably more a lack of reflexive agency over their understanding of the world: It seems like people just stay confused about things that matter to them, and I find this utterly baffling. To me, it’s intuitively obvious that when you don’t understand something important, you fix that.
A trivial example of this is bay leaves. I run into so many people who are like “What do bay leaves even do ha ha idk?!”. If that’s you, go put a bay leaf in a cup of boiling water for five minutes and then take a sip. Congratulations, now you know what a bay leaf does.
This is a dumb example, but I think it’s useful because it demonstrates that this really isn’t a failure of intelligence on people’s part. It’s something more basic - a failure of agency over going out and discovering things for yourself.
As with agency in general, I think reflexivity is quite subject specific. I’m much better at it in contexts where I’m interested in or skilled at the problems - e.g. “I don’t understand this, so I’ll figure it out” is reflexive for me while, as the frying pan example shows, I’m still working on “This is broken, so I’ll fix it”.
Prompts for agency
The opposite to reflexive agency is prompted agency - agency which only becomes available given some prompt to action. “That does seem like a problem. Have you tried solving it?” is my only-half-joking prompt I use with people: It’s something that highlights the failure of agency, in a way that makes people aware that action is an option available to them.
Prompted agency doesn’t necessarily have to be external. For example, you can use journaling as a way to explore why you’re unhappy, and in the course of journaling you might prompt yourself to do something about it. Another easy prompt like this is just to talk out loud to yourself about what you could be doing now.
You can think of this as two separate exercises of agency:
I am unhappy, I should journal about that (might be prompted or reflexive depending on what leads you to that conclusion)
Having journaled, you become aware of your capacity to do something about whatever is causing you to be unhappy, and exercise agency to solve the actual problem.
In contrast, having someone external who just tells you what to do to become happy is not an exercise of agency, even if it works (at least, not an exercise of your agency), because it lacks the element of choice.
This is a fuzzy line (e.g. you could choose whether or not to do what they tell you to), but that’s OK. I’m not really trying to draw a hard and fast agency vs not-agency distinction so much as to point out a characteristic spectrum of agency: On the one end, you have fully reflexive agency in which you act spontaneously to satisfy your preferences, on the other end you have total obedience to some authority who tells you how to be happy, and in the middle is prompted agency. Self-prompted is perhaps closer to the reflexive agency end, and other-prompted (e.g. coaching) to the obedience end.
Prompts and permission structures
I think often what’s going on with prompted agency is that it gives you permission to act, and many people aren’t good at acting without permission.
We grow up in very low agency environments. Between home and school, generally you aren’t able to exercise much agency without explicit permission to do so.6 As adults, we have much more permission to act autonomously, but we first have to unlearn the habits and emotional responses that we acquired when we didn't.
As an adult, your opportunities for agency are certainly not completely unconstrained. Bosses will get annoyed if you don’t do your job, your community might get upset with you if you suddenly start talking about how the other side maybe has a point, your friends and partners will be justifiably quite annoyed with you if you disappear for a month without warning, and many more examples besides.
But also, you do have a lot of leeway, and probably nobody is going to get mad at you for fixing the dining room table or making a nice hot mug of bay leaf tea. It’s just that some part of you still thinks it needs to ask for permission.
When self-prompting works, it usually will do so because it makes you aware of capabilities you already had and already had permission to exercise, but hadn’t properly internalised. Sometimes you’ll come up with genuinely clever ideas, but most of the time it’s a “Oh yeah I can just put a coat hook there, let me get my drill” level of insight.
In theory what should happen is every time you have this type of insight, your modern permission structure should become slightly more emotionally accessible. You build the habit of exercising agency, and you learn emotionally that this is a thing you’re allowed to do without external permission.
Prompts as defamiliarization
Another reason why prompts work is that they break the familiarity that prevents us from exercising agency - as in the table and the frying pan scenarios, where it’s not that we don’t have the ability to fix them, but that we’ve gotten used to the problems so don’t see them as something that can be fixed.
Dan Luu has a good post about normalization of deviance (the tendency for objectively bad behaviour to become normal in organisations over time) in which he quotes Julia Evans’s explanation of this as follows:
new person joins
new person: WTF WTF WTF WTF WTF
old hands: yeah we know we're concerned about it
new person: WTF WTF wTF wtf wtf w...
new person gets used to it
new person #2 joins
new person #2: WTF WTF WTF WTF
new person: yeah we know. we're concerned about it.
Prompts which force you to look at the situation allow you to break out of the “old hands” mindset where you acknowledge it’s a problem but it doesn’t feel like something you can exert agency over and let you look at it with fresh eyes where it feels like something actionable again.
Learning to exercise agency
I’m a big fan of the four stages of competence model of learning. I doubt it’s literally true, but it’s often a good way of thinking about things.
The four stages are:
Unconscious incompetence: You’re bad, but you don’t really understand that you’re bad, of if you do understand that you don’t understand why. You might e.g. describe yourself as “bad at mathematics” as if that if that were an intrinsic trait.
Conscious incompetence: You’re still bad, but you can look at your own skill and identify and understand specific failings. e.g. “I always get this sort of problem wrong, because I make sign errors.”
Conscious competence: You’ve started to become good, but it takes active concentration and careful thought. Many tasks are possible but don’t come naturally and are much slower than you’d like them to be.
Unconscious competence: You’ve internalised much of the skill to the point where you can be good at it without really having to think about it.
I think you could approximately map these to four stages of developing agency in a domain:
Unconscious incompetence in agency is more or less learned helplessness. You cannot exercise agency (in a particular domain), even with external assistance. This is a failure of agency.
Conscious incompetence: You are aware that you could be exercising more agency than you are, and with external guidance you can identify actions you could take to satisfy your preferences and then act on them. This is externally-prompted agency.
Conscious competence: Although it does not come automatically to you, through some mix of journaling and other methods you can exercise agency. This is internally-prompted agency.
Unconscious competence: You relatively automatically act to satisfy your preferences without having to sit down and think about it.
This isn’t perfect and I think the middle two blur together somewhat (this is generally the case and one of the problems with the stages of competence model I think), but to the degree that prompted agency in some area is accessible and reflexive agency is not, this seems like a relatively clear path to increasing agency for most people.
So, if you want to increase your agency: Have conversations with others, or do journaling, about what you can do to improve your circumstances, and then act on the conclusions from that.
Likely a lot of the time when you do this you’ll find that regardless of what is suggested and whether it seems like a good idea, you’ll not want to do it, and there’s some underlying strong emotional reaction against the idea of improving your agency in that domain. This might be legitimate (there are genuinely cases where becoming more competent is actively bad for you because capabilities can be coerced) or it might be an emotional problem you need to debug.
Conveniently, this too is something you can try to get greater agency over, using the same tools of conversations and journaling.
All of which is to say: Lack of agency? That does seem like a problem. Have you tried solving it?
Postscript
The cover image is Tools in a cardboard box by Wikimedia user Vinne4568, made available under a creative commons license.
If you’d like to hang out with the sort of people who read this sort of piece, you can join us in the Overthinking Everything discord by clicking this invitation link. You can also read more about it in our community guide first if you like.
There are a lot of boring arguments about whether machines or other nonhuman entities can have agency. My philosophical stance on the nature of consciousness requires me to say that the answer is “obviously yes wtf why are you even asking this question?”, but if you disagree then it mostly won’t be relevant for this article.
Which is not to say that people can’t develop new preferences, but I don’t think that’s the bottleneck for most people
Even to the degree it’s true, it mostly seems to result in people self declaring themselves to be high agency and calling all the low agency people NPCs. Very few people ever identify as low agency. The internet is like this whenever you divide people into groups meaning “the good ones” and “the bad ones”.
Spoiler: I probably can’t.
At least, that’s how I remember the story. I suspect this whole saga actually took about a week, with several days of me putting up with it, then asking, and then several more days before I fixed it. But this version of the anecdote is a better example.
Much of this is correct - kids aren’t very good at using their agency in ways that don’t result in everything going horribly wrong. Arguably if you gave them more options to exercise it, they’d get better at it faster, but I try not to give parenting advice and the fact of the matter is regardless of whether kids should be in low agency environments they usually are.