Friday, 2 December 2016

Scarantino (2003) “Affordances Explained”

Turvey, Shaw, Reed and Mace (1981) laid out an ontology of affordances; a formal account of the kind of things they are. They described them as dispositions, properties of the world constituted by sets of anchoring properties that offered an action to an organism whose dispositions could complement the affordance. Making affordances dispositions makes them real, makes them pre-date the interaction with the organism, and accounts for their odd ‘not doing anything until interacted with’ kind of existence. I am firmly Team Affordances are Dispositions and I have yet to meet an alternative account that supports a science of affordances or even allows them to be perceived.

The literature on dispositions was somewhat limited in 1981, but in 1998 Stephan Mumford published the definitive work on what they are and how they work. I always hoped someone with the necessary philosophy chops would use this work to strengthen the foundations of affordances (I even almost talked a philosopher into doing it!) but it turns out I’m covered. Andrea Scarantino (2003) published ‘Affordances Explained’ and did much of the necessary work, and there are some very useful things in the analysis. This post is me working through this material, translating from the technical philosophy into words I can understand better.

Why Did Gibson Invent Affordances?
Gibson’s key insight was the ecological nature of our behaviour; the fact that organisms and their environments aren’t just complementary with respect to physical and biological factors, but also with respect to psychological factors. The correct level of analysis for psychology is the Organism-Environment system.

There are two consequences of this stance. First, behaviour must be explained as occurring at an ecological scale. Most of physics simply isn’t relevant to an organism’s day to day life, so we need to pick out the stuff that is relevant. Second, our explanations must be intrinsically relational; this complementarity must be baked into our science, not tacked on at the end. Affordances are Gibson’s solution to the consequences; they exist at an ecological scale and are intrinsically about the complementarity.

(Scarantino notes here that Gibson did not believe affordances were bestowed upon objects by perceivers; objects have affordances by virtue of their ecological scale physical properties. To be honest, to my mind, correctly identifying this immediately kills off relational accounts of affordances, but there are unfortunately a few Gibson quotes and other ideas that breathe life into those theories).

Affordances are only interesting to psychology if they are perceptible, and we can only figure out how this might work once we know what kind of properties affordances are.

What is a Disposition?
Mumford (1998) states that a predicate expressing a property is dispositional if that expression describes a subjunctive conditional statement that is true by conceptual necessity. Easy, right? Let’s cash this out.

Saying ‘X is fragile’ is a predicate about a property (fragility) that entails the subjunctive conditional ‘if X were suitably hit, then X would break’. X is not currently breaking; that’s not what fragile means. What fragile means is that under certain conditions, X would break. It is also conceptually necessary to the definition of fragile that this conditional applies; the definition of fragile is in part constituted by this conditional.

In contrast, saying ‘X is triangular’ is a predicate about a property (triangularity) that entails the subjunctive conditional ‘if X’s sides were suitably counted, the result would be 3’. While this is true, it doesn’t really capture the meaning of ‘triangular’ in the way it did for ‘fragile’. It’s not conceptually necessary to the definition of triangle, it’s just a consequence of the definition of triangle. This particular conditional does not constitute any part of the definition of triangular.
So for affordances to be dispositions, they must entail subjunctive conditionals that constitute part of their definition and thus be conceptually necessary. The next part is to see if this holds up.

How to Make Affordances Dispositions
As above, the predicate “X has affordance property A (at time t relative to an organism O in circumstances” is dispositional if you can flesh out a particular subjunctive conditional it entails. Annoyingly (at least for the purposes of clarity) the particular subjunctive conditional is a subjunctive conditional conditional; there is a condition nested in the condition and it takes the general form
 At time t, if background circumstances C were the case, then (if a set of triggering circumstances T were the case) then a manifestation M involving X and O would be the case with probability p.
The first ‘if’ (about C) is one condition; the second if (about T) is the nested condition. If you can cash out a subjunctive conditional conditional that is entailed by the affordance predicate by conceptual necessity, you have a disposition. To fully characterise an affordance as a disposition, you have to be able to fill in all the gaps.

So far so good; but to be useful, there have to be some constraints on how you go about filling in the gaps. X, t, T, M, and p are all fairly straightforward. A is the thing you are trying to flesh out, so it’s form is up to you; you’ll either frame it so that this all works out or not, and if it does the result will either be perceptible in principle, or not. Only when it works out and is perceptible can it be an affordance.

C (the background circumstances) and O (the organism) are where it gets tricky.

Things to Worry About C (the Background Circumstances)
The background circumstances necessary for X to have a disposition is an indefinitely large set, and even if you have that set you can always make a new one by adding a circumstance that breaks the disposition. So you cannot exhaustively specify C.

That said, you can’t just allow C to be open ended, because you can probably always create a set of circumstances where just about anything could have just about any disposition.

A solution: Scarantino states here that, with respect to affordances,
we should rely on a tacit understanding of C as the set of normal ecological circumstances
and that this understanding must include a way to identify when that normal range has been crossed. Without this last bit, you’d never know if the failure to get an affordance to manifest was because it exists but you’d gone outside the range or because you are wrong about your candidate affordance.

(Let me interject here, and add to the mix the Turvey et al (1981) notion of ecological laws and the scope those laws have. I need to think some more on this, but it seems at a first glance that ‘the scope of the ecological laws’ is precisely the kind of theory of ‘normal ecological circumstances’ required to avoid having the whole thing be vacuous.)

Things to Worry About O (the Organism)
Scarantino notes that there are a couple of different ways to manifest an affordances. ‘Climbing’ is something an organism does; ‘Falling off’ is something that happens to an organism. He thinks this is worth noting and classifies them as to whether the manifestation is goal-achieving or not, and focuses on the goal-achieving doing affordances. He then states
…if it is true that, given the background circumstances C, an organism O can at time t engage in an event that qualifies as a doing or a happening M and involves X, then X is at time t an affordance-bearer with manifestation M relative to O in circumstances C.
We’ve covered C above; we now need to worry about O. We need to hone in on what ‘O can at time t’ means. What are the relevant ability properties of O?

There are two issues
  • Current vs latent abilities – a current ability is something O can definitely actually do (e.g. I am able to sit) while a latent ability is something O could learn to do but hasn’t yet (e.g. I can’t ride a unicycle but I could learn). 
  • Reliability – if I try to sit, I am good enough to pull it off almost all the time. That is a reliable current ability. During learning, I might stay on the unicycle one time in ten. This is an unreliable current ability.
Scarantino suggests restricting “affordance-bestowing abilities” to reliable current abilities, where what counts as reliable will require some work. So the full description of affordances as dispositions is
The subjunctive conditional entailed by conceptual necessity by the ascription of goal affordances will then have the form ‘at t, if background circumstances C were the case, then if organism O were to select the goal of [climbing, grasping, letter mailing, etc.], then the manifestation of [climbing, grasping, letter mailing, etc.] would occur with positive probability p. (pg 959)
Then the only remaining question is, what counts as a sufficient probability p to ascribe an affordance? What constitutes reliability? Scarantino has no answers here, which I think is fair; this one will need empirical work.

An opportunity and a risk
Scarantino finishes with these thoughts.

First, he notes that having dispositional affordances in your explanatory toolkit provides the opportunity to have “an object of perception imbued with a kind of meaning – ecological meaning” that all organisms can potentially have in common, and that it would make sense to say a variety of animals all show access to this meaning when they ‘similarly engage in a perceptually guided behavioural discrimination of the brink of the cliff as fall-off-able.” In other words, how can psychology say we all experience the same thing? By showing we all behave with respect to the same affordances.

Second, he notes a risk of trying to explain everything in terms of affordance perception. While catch-ability should work under this account, ‘score-with-ability’ (the ability of a fly ball to score a home run) will need more pieces to work. (This accords perfectly with Sabrina’s analysis of expanding the ecological approach into, for example, language – work that tries to find linguistic affordances is stretching the concept beyond its natural scope, and the better move is to look for the conventional use of information; Golonka, 2015).

Some thoughts
This paper is excellent. It does some amazing work in nailing the dispositional account of affordances to the more general understanding of dispositions, and it points to some useful consequences.

There are some gaps that need filling. For example, while I agree that to say something affords an action to me, that action has to be a reliable current ability, we will still need a story whereby the dispositional property underpinning this 'affordance ascription' can serve to guide learning. At a first pass, it seems like you might start out calling the property just a disposition that creates information, and reserve ‘affordance’ for when it reliably supports a now-current ability. Like reliability, however, it’s not clear how to draw a principled line between the two descriptions. That said, the property is always the same thing – the issue is what we call it and when.

I can also (perhaps) perceive that something affords something for someone else but not for me; that cup is reachable by you because you are closer, or taller. This may connect to the above; am I perceiving an affordance, or do I restrict that term for actions I can do and I am instead perceiving two dispositions (the reachability of the cup and your ability to manifest that)?

The reliability issue I think can be informed by bringing perception back into the story. Remember, this paper is about the ontology of the objects of perception, and not the process of perception of those objects. Now we know what the objects of perception are, we can get into the issues of the information for those objects and the detection of that information. Learning is, in large part, perceptual learning – becoming attuned to the relevant information. Reliability will vary as a function of that attunement, and it is in the act of perception that we should probably look for ways to ground the distinction between reliable and unreliable.

I’ll flag up here that all this talk of ‘meaning’ and ‘information about’ affordances is annoying a lot of enactivists. Hutto & Myin (2013) for example are working hard to purge cognitive science of the scourge of content, and some ecological authors are all on board (e.g. van Dijk, Withagen & Bongers, 2015; Bruineberg, Kiverstein & Rietveld, 2016). Sabrina and I are currently on the other side of this (I mean, we’ve written a paper called ‘Ecological Representations’ for the love). Reading van Dijk et al, it seems clear to me that getting rid of content only works if you treat affordances as relational. Given that this isn’t right, that argument is going to go away, and I think the work Scarantino has done helps buttress that rejection. I’ll save that for the next post, though.

No comments:

Post a Comment