Affordance & direct-perception: What we learn are affordances

This short piece summarises some recent research on perception, links through to arguments against direct-perception as proposed by James Gibson, and proposes that affordances are the basic building blocks of learning.

Watching my young son play, and reflecting on the work done by Piaget and others on how people become who they are, I think it's clear that play and learning is a process of combining what is known about the world in terms of existing affordances with actions that cause new events (either internally or externally initiated), from which new affordances can be derived. Gibson (and others) have argued that affordances are nested, quite a few people have even argued for meta-affordances (not sure I like that term though, one person's meta is another person's normal :-), but either way it seems to me it's possible to outline a conceptual model of learning that uses affordances as the core. Affordances can't perhaps explain how we learn, but they may perhaps describe at a very basic level what we learn. I'm struck by Andrew Ng's work on AI, and the idea that human intelligence might derive from a single algorithm. Affordance may lie at the heart of that algorithm.

A particular problem with affordances that has been highlighted it the notion of direct perception of meaning. As Kroes & Franssen’s (2013) summarise, “the basic problem of Gibson’s theory of affordances is how it is possible for an animal to directly perceive what its environment ‘has in store’ for it or affords it. According to Gibson’s theory, what the environment affords an animal is an objective feature of the environment.”. The idea of direct perception of meaning is a key concept of the overall theory of affordances. As Jones (2003) summarises, at the heart of the theory is an attempt to understand how individuals understand the meaning of objects they encounter within the world. So how problematic is the idea of direct perception for affordances?

Until Gibson’s theory of affordances, meaning was understood to take an “in-direct cognitive route”, i.e. meaning was established through a process of encoding and decoding visual and other perceptual information, much like a computer might do. Gibson, however, proposed a much more direct route - he proposed that there was form of match between the perception of objects in the world and the brain, so that meaning was inherent within perception itself. This radical proposal is challenging to understand for a culture firmly entrenched within an information processing paradigm. On face value, the idea that the meaning of complex objects and people within the world can be directly perceived seems almost absurd; surely some form of computation or processing is necessary in order for meaning to be deduced.

Ignoring the obvious homunculus problems that a theory of indirect perception suggests, it is perhaps possible to argue that the theory of direct perception can be rationalised and appear less preposterous if we consider that humans do not exist as beings independent of time. We have history, at any one point in time we have an existence and a mental structure that predates our current situation and experience. Even in the womb our cognitive development and our perceptual development are intrinsically tied together. Considered from an historical perspective, the idea of direct perception does perhaps become easier to understand. Arguments against direct perception assume the animal arrives fully formed, whereas in actual fact affordances have been structured over many years of development. My young son plays with wooden blocks, he plays with his tractor, he plays with his food, all the time using many basic affordances which have already been established to do with weight, firmness, bendability, etc.. It’s a nonsense to assume it’s all new, it’s very carefully layered within his developing mind. Perhaps more critically, the affordances are all personal and internal, they are not objective, they simply connect him to the world. What is objective are the invariants in the world which he has internalised as affordances for him. The invariants of a cooked french bean exist in its length, its hardness, its sugar levels, etc. all in some way objective properties which could be measured. The affordances from Max’s perspective are all personal, but related to these invariants, which mean he can put in his mouth, chew it, and it tastes good, it provides a nice thing to eat which he gets pleasure from. Exactly the same invariants might not provide that affordance for another child, and would certainly not provide those affordances for my cat. Invariants are objective, affordances are not.

Watching Max use the ‘Gallery’ icon on the tablet to start Timmy Time, it occurred to me that ‘affordance’ from an HCI perspective becomes irrelevant when someone has learnt the affordance itself. The gallery icon is a weird colourful square, with no physical similarity to any type of video or image, the only vague similarly it has is to a picture frame - something which has no meaning for Max. He has constructed his own meaning - or affordance - by watching others use it and by experimenting with what is invariant about it. It has consistent properties, invariance if you will, which allows him to connect to a source of pleasure - Timmy Time movies. The design itself, the graphic, is totally irrelevant. It's objective invariants have been recoded into affordances in my sons mind, seeing the invariant characteristics instantly triggers those affordances through direct perception. 

Later research has something of a habit of adding support for Gibson’s theories (cf. Warren, 1984; Costantini & Sinigaglia, 2011; Sanguinetti et al, 2014), and in my opinion the concept of direct perception is no exception. Recent work by Potter et al (2013) seems to confirm that individuals can perceive the meaning of objects presented visually in only 13 milliseconds. which is too fast for existing theories of information processing. In fact even the 13 millisecond limited may be incorrect, as this was the limit imposed on the researchers by the technical equipment they were using for their experiments. Potentially meaning could be understood even quicker. It seems likely that direct perception is the mechanism that links objects to meaning. If so, then I propose that affordances must be, to all intents and purposes, the building blocks of all learning.

Popular Posts