Affordances

post by abramdemski · 2021-04-02T20:53:35.639Z · LW · GW · 18 comments

Contents

  A Toy Model
  Engineering Your Affordances
    More affordances
    Less Affordances
  The Affordance Model of Human Behavior
None
19 comments

I noticed recently that, although the term "affordance" seems to be in the LessWrong lexicon, there isn't any article about it. I think it might be one of those concepts which you currently have to pick up from CFAR, or learn through osmosis. This seems like an important concept, so I'm going to attempt to do it justice in this essay.

The easiest way to define the term is with examples: a button affords pushing, a lever affords pulling, etc. An affordance is a vividly present option for action. Note that affordances vary from person to person: one person may have the affordance to dance at a party, while another person lacks that affordance.

When I google "affordances", I mainly get results about UX design and human-computer interaction. This makes sense: a big part of designing products (be they software or hardware) is about making sure the user has all the right affordances. If you design a new kitchen implement which requires a pumping action, you want the user to immediately have a "pumping action" affordance when they see it.

The biggest hurdle to having an affordance is raising the possibility to attention. But that's not quite all there is to it. For example, if you're struggling to ask someone out on a date, but you just can't, I would say you lack the affordance. You explicitly know what you want to do, but you "don't have that menu option". 

A Toy Model

I'm going to use a system 1 / system 2 [? · GW] model to explain what might be going on.

System 1 has a vast array of possible actions it could take at any time; at least, all possible combinations of muscle tensions across time. To simplify planning, system 1 abstracts a small number of possibilities for our attention. We usually don't even consider options that aren't on this list, because this list just is the action-space to us, at least when we're not thinking hard about it.

On the other hand, system 2 has its own abstract, simplified view of the world, and its own role to play in planning. Sometimes, then, system 2's idea of the action-space will get out of sync with system 1's. Thus, a shy person finds themselves tongue-tied when they try to ask an attractive person out. System 1 doesn't present that menu option: we lack that affordance. If system 2 can manage to override system 1 at all, we're turning off the auto-pilot, and steering manually. Forming coherent sentences suddenly seems difficult. We don't know what to do with our hands.

System 1 is quite intelligent about which affordances to bring out. Often this is based on aspects of our self-image. (You can think of your affordances as part of your self-model: what you expect yourself to be capable of.) For example, when I briefly experimented with polyamory, I felt a huge shift in my affordances taking place, in a way which felt "automatic"/"outside of my control".

Engineering Your Affordances

More affordances

From a CFAR perspective, one of the most important things about affordances is that you may lack important affordances. When you lack an affordance, you're often blind to that fact, because it's just not something which exists in your attention.

As I mentioned, sometimes all you need to do is bring something to attention. "You mean I can just bake my own cookies any time I want?!" There are a number of exercises you can do to try to see new possibilities like this.

Sometimes, however, just raising possibilities to attention isn't enough. Sometimes we know perfectly well which affordances we're missing, but we just ... can't.

(Note that in my model, this is very different from a situation where you have the relevant affordance, but never in fact do the thing. "I have the menu option, but I never press it" is different from "I don't have the menu option". However, I could understand disagreement with this distinction, since there isn't an obvious difference in behavior.)

In those circumstances, system 2 models X as a possible action, but system 1 lacks the abstraction: it can't smoothly generate a motor-plan which corresponds to what system 2 has in mind. Some reasons why this might be:

One CFAR class used the example of stretching before a workout: apparently, some research suggests that stretching involves a strong cognitive component. The brain has a body-model which includes information about which physical configurations are safe vs unsafe. What we experience as a physical inability to stretch beyond a specific point, is often system 1 putting on the brakes before we hurt ourselves. However, system 1's body-model is not perfect. Stretching allows us to gently explore what happens when we put ourselves in different positions, and the safety information we gain often allows system 1 to relax some limitations.

Cognitive-behavioral therapy would advise us to cash out the feeling of resistance into a hypothesis, and come up with a way to test that hypothesis safely. For example, if we expect our boss to blow up at every little thing, we might be able to test that hypothesis with something small before trying to talk to our boss about what's really bothering us.

The connection between affordances and identity which I mentioned earlier suggests another thing we can try: play with our identity in order to shift affordances. Another CFAR class I took involved a technique like this, using a time-boxed improv session to experiment with alternative personality traits.

Less Affordances

The previous section explored the idea of increasing our affordances. However, system 1 limits our affordances for good reason!

An affordance is a lot like an "open loop" in the Getting Things Done sense. If this is true, then if you have your phone in your pocket, the possibility of taking it out to check something takes some sliver of subconscious attention. On this model, you can increase your attention by removing things like this.

Just as you might place a book prominently near where you sit in order to give yourself the affordance of reading it, you might similarly take away any distractions from around where you habitually sit, to decrease affordances which might be distracting you.

Relatedly, you might hide the cookie jar in a cupboard so that you don't think of taking one every time you pass by it.

The Affordance Model of Human Behavior

I want to finish off this post with what I consider to be my one personal contribution: a very simple model of human behavior. This model says: "we can mostly predict what humans will do by what they have an affordance to do."

As some famous mountaineers have said when asked why climb Mount Everest: "Because it is there." According to this theory, the major reason people do things is because it's a salient possibility.

Why do people collect tiny spoons? Because it is there! Why do people try for high-scoring spots on the work pinball machine? Because it is there!

Some children go through a phase where they have a serious biting problem, frequently biting the fingers and other body parts of other children or adults. Why?? What purpose could this serve in the mind of the child? One possibility occurs to me: "because it's there!" When I try to put myself in the mindset of such a child, an outstretched hand is simply a wonderful opportunity. It doesn't have to be about anger or any desire to harm. The "biting" action is simply what an outstretched hand affords.

I am a decent juggler. For a time, juggling was one of my main social crutches at parties. It gave me an easy way to interact with people. However, that interaction often involved me trying to do tricks, while non-jugglers tried to distract me or steal my juggling balls or similar antics. This isn't particularly difficult. Nothing especially satisfying happens when they succeed. So why is this so common? 

I think it's similar to the reason why a child often wants to knock over another child's block tower. It's not out of spite. It's just because it's there! A button affords pushing, a lever affords pulling, and a tower affords toppling. People interfere with my juggling because they don't have other ways to interact. It's not out of some general mischievous desire to spoil what others are trying to do. It's just that "make the balls fall" is an affordance, whereas actual collaboration with the juggling isn't.

(Which is why my next move is to try to rope them into pair-juggling tricks, which can often be executed successfully even by non-jugglers. This puts a more constructive option on the table.)

Another important aspect of this model is social contagion. For example: why do people keep diaries? Most of the reason is "explained" by the fact that they've simply heard it's a thing you can do. On this model, a whole lot of human behavior is explained by assuming humans are dumb imitation learners who mostly decide what to do by looking around and copying what other people are doing.

Now, obviously, the affordance model of human behavior isn't going to be perfect. Clearly there are other important factors. But I think it's a useful model nonetheless. It's too easy, otherwise, to underweight the importance of mere salience. Get too caught up in a climber's personal story about why they climbed Mt. Everest, and you might forget that statistically, a whole lot of whether-someone-climbs-mt-everest is probably statistically explained by whether they encountered situations which created the affordance.

18 comments

Comments sorted by top scores.

comment by jaspax · 2021-04-03T06:37:42.799Z · LW(p) · GW(p)

My major problem with this use of the "affordance" terminology (which I'm mostly familiar with from the perspective of UX design) is that you are inverting the role of agent and object. In the original usage of the term, and affordance is something that the object has which both signals to the agent that an action is possible and makes that action easy to carry out. Most of your psychological examples make the "affordance" a property of the agent, referring to the agent's predisposition and ability to carry out an action in a particular context. It's not clear to me whether you even noticed that you inverted the locus in your extended examples.

This makes it hard for me to read your examples, and I wish we had a different word for the agent-focused notion of "affordance" and kept the original notion of "affordance" for the object-focused concept.

Replies from: abramdemski, abramdemski, adamzerner, sune-jakobsen
comment by abramdemski · 2021-04-04T16:47:58.623Z · LW(p) · GW(p)

Thinking about this a bit more, I think a natural usage, which is pretty compatible with what I've experienced, would be:

  • A path affords walking.
  • A person has an affordance for walking down the path.
comment by abramdemski · 2021-04-04T15:04:39.416Z · LW(p) · GW(p)

My understanding is that the concept is supposed to be a fully relational one, indicating something about the interaction of a subject and object. So I would say a door "has an affordance" (for a person to open it) and a person "has an affordance" (to open the door), much like I would say both people in a romantic relationship "have a romantic relationship" (with each other). 

In the original usage of the term, and affordance is something that the object has which both signals to the agent that an action is possible and makes that action easy to carry out.

My basis for thinking it's supposed to be relational is the Wikipedia article:

The key to understanding affordance is that it is relational and characterizes the suitability of the environment to the observer, and so, depends on their current intentions and their capabilities. For instance, a set of steps which rises four feet high does not afford climbing to the crawling infant, yet might provide rest to a tired adult or the opportunity to move to another floor for an adult who wished to reach an alternative destination. This notion of intention/needs is critical to an understanding of affordance, as it explains how the same aspect of the environment can provide different affordances to different people, and even to the same individual at another point in time.

You wish for different terms:

This makes it hard for me to read your examples, and I wish we had a different word for the agent-focused notion of "affordance" and kept the original notion of "affordance" for the object-focused concept.

I suppose it would be possible, like a King "has subjects" and the subjects "have a ruler". So in an affordance relationship, the object could "have a an afforder" and the subject could "have an affordee" or something like that. But personally I find it perfectly intuitive to say that both "have an affordance", and understand that different (but symmetric) things are meant in each case.

Replies from: jaspax
comment by jaspax · 2021-04-05T06:33:25.012Z · LW(p) · GW(p)

The concept is definitely relational; no disagreement there.

My objection is more narrowly linguistic: the syntactic structure used to describe the "affordance" relationship is Object affords Action to Agent. All of your quotes from Wikipedia follow this example, eg. the "set of steps... does not afford climbing to the crawling infant" (emphasis mine). I find no examples of this syntactic structure being inverted to allow Agent affords Action. Consequently, it seems that the noun "affordance" is best applied to the Object's side of this relationship, and not the Agent side, since the Object is the syntactic subject.

Conceptually, this does matter because the affordance relationship is non-symmetric: what the Object does ("affords") is very different from what the Agent does! Aside from the syntactic objection, I think that it obscures the topic to have the same word used for both sides of a non-symmetrical relation. Your suggestion of using "have an affordance" is possibly usable though I still think that it invites confusion. I do like the phrase " behavioural repertoire", mentioned in another comment, but it does not lend itself to being verbed very well. Another suggestion might be "reciprocate" or "engage": an Agent engages the affordance by carrying out the Action in the manner intended. (Does the existing literature have a verb that slots into this construction?)

I don't know. Words are hard. I still think that it's important to have different words for the Object's and the Agent's respective contributions to the activity.

comment by Adam Zerner (adamzerner) · 2021-04-04T04:20:09.280Z · LW(p) · GW(p)

Yeah, agent vs object is the main thought I have as well after reading this post.

Coming from the perspective of affordances as belonging to the object, I have been thinking about the following recently. In programming, one of the big reasons to "choose the right tool for the job" is because of affordances. For example, suppose you have a bunch of functions that you want to wrap in a namespace. You can have a class and then make all of the functions static methods of the class. This is something I see Ruby programmers do. However, a class affords a few important things that aren't being used here. Mainly object instantiation and inheritance (if you don't need inheritance you can use a factor instead of a class). So then, due to the misleading affordances, I see a class with static methods as the wrong tool for the job.

comment by Sune Jakobsen (sune-jakobsen) · 2021-04-03T06:43:50.794Z · LW(p) · GW(p)

I think behavioral repertoire captures this concept (I’m not too familiar with its usage: I have only heard it once).

comment by Paul Crowley (ciphergoth) · 2021-04-03T13:29:04.763Z · LW(p) · GW(p)

I would guess a lot of us picked the term up from Donald Norman's The Design of Everyday Things.

comment by Gordon Seidoh Worley (gworley) · 2021-04-03T00:10:52.942Z · LW(p) · GW(p)

I'm pretty sure I've commented this before elsewhere on LW when a similar discussion came up, but there's a wealth of literature on this topic but under a different set of terms than "affordance". SEP feels the right general term is "disposition", and it generalizes not just to people but to all things.

Replies from: abramdemski
comment by abramdemski · 2021-04-03T02:17:26.682Z · LW(p) · GW(p)

I don't think this is quite the same. An affordance is relational and subjective. From Wikipedia:

For instance, a set of steps which rises four feet high does not afford climbing to the crawling infant, yet might provide rest to a tired adult or the opportunity to move to another floor for an adult who wished to reach an alternative destination.

Hence an affordance depends both on the subject and the object. A disposition seems like a concept which is supposed to apply to the object in itself, independent of subject.

Replies from: gworley
comment by Gordon Seidoh Worley (gworley) · 2021-04-03T02:45:15.517Z · LW(p) · GW(p)

A disposition seems like a concept which is supposed to apply to the object in itself, independent of subject.

I wouldn't say there's any existence of an object (as an object) independent of a subject; there's instead just stuff that's not differentiated from other stuff because something had to tell the difference, hence I don't see a real difference here, although the theory of dispositions is jumbled up with lots of philosophy that supposes some kind of essentialism, so it's reasonable that there might seem to be some difference from affordances under certain assumptions.

Replies from: abramdemski
comment by abramdemski · 2021-04-04T14:40:04.109Z · LW(p) · GW(p)

Even if so, I miiight contend that there's an important distinction between (a) affordance as a concept which is itself relational, vs (b) affordance as a predicate on objects, where objects are understood to be subjective. In the guest case, it's possible for agents to have a shared model in which an object has different affordances for different people. In the second case, if agents try to have a shared model but end up disagreeing about affordances, it's not clear what they should do.

comment by romeostevensit · 2021-04-02T22:50:54.097Z · LW(p) · GW(p)

One of the major helps for me was thinking in terms of number of potential leverage points for an intervention. If an intervention seems hard to me it is often the case that I am simply trying to lever on a bad spot and am unaware of other spots to place the lever.

Deliberate practice is often about building an intervention model that is more fine grained, which exposes more such points.

comment by Raemon · 2021-04-03T16:53:00.171Z · LW(p) · GW(p)

When I google "affordances", I mainly get results about UX design and human-computer interaction. This makes sense: a big part of designing products (be they software or hardware) is about making sure the user has all the right affordances. If you design a new kitchen implement which requires a pumping action, you want the user to immediately have a "pumping action" affordance when they see it.

Possibly relevant history: the word affordance AFAICT comes from the book "The Design of Everyday Things." The Design of Everyday things was actually originally going to be called "The Psychology of Everyday things", but marketing people were worried about it being a confusing title.

The book is in IMO "The Sequences, if they were written by someone trying to improve product design rather than someone trying to build artificial intelligence." It covers System 1 and System 2, various cognitive biases, etc, weaving them together with design principles.

Particularly noteworthy: the book actually introduces two terms – affordances, and signifiers, which often get confused. From this random article I just googled:

  • An affordance is something an object (or dashboard) can do. A tap/faucet can run hot or cold water, for example.
  • A signifier is an indicator of some sort. In our tap example, this might be red/blue dots signifying which way to turn the tap to get hot or cold water.
Replies from: abramdemski
comment by abramdemski · 2021-04-04T14:56:25.852Z · LW(p) · GW(p)

Wikipedia gives a pretty different history, according to which the term comes originally from psychology, not design.

Replies from: Raemon
comment by Raemon · 2021-04-04T22:09:24.570Z · LW(p) · GW(p)

Welp, today I learned. ("It originated in psychology" feels consistent with my previous beliefs, but I didn't know about all this history of it)

comment by Yoav Ravid · 2021-04-03T06:54:17.025Z · LW(p) · GW(p)

This reminds me of Relevance Realization [LW(p) · GW(p)]. One term you used that John Vervaeke also uses a lot is salience. I think he also uses affordance? though i don't remember specific examples. 

comment by Pattern · 2021-04-04T19:05:50.395Z · LW(p) · GW(p)
For example, do you wish you had more of an affordance to lie? Probably not, right?

If something comes to mind as an option (an action to take), or as a possibility (to consider), maybe it's in your mind for a reason, and it might be useful to understand why. For example:

An affordance is a lot like an "open loop" in the Getting Things Done sense. If this is true, then if you have your phone in your pocket, the possibility of taking it out to check something takes some sliver of subconscious attention. On this model, you can increase your attention by removing things like this.

Maybe your attention goes to the phone because you have a habit involving that phone. Theories involving dopamine/skinner boxes/etc. aside, if doing a thing 'makes you more likely to do it', doing a thing less may be required for you to 'do it less', and feel the urge to do it less.


Get too caught up in a climber's personal story about why they climbed Mt. Everest, and you might forget that statistically, a whole lot of whether-someone-climbs-mt-everest is probably statistically explained by whether they encountered situations which created the affordance.

The story may be useful to get an idea of what those situations ("which created the affordance") are, or how they might be encountered.

comment by Sune Jakobsen (sune-jakobsen) · 2021-04-03T06:38:20.263Z · LW(p) · GW(p)

This reminds me of the concept of a behavioral repertoire, although this considered to depend on only the subject rather than on the (subject, object) pair.