Thursday, 1 March 2018

General Ecological Information Does Not Support the Perception of Anything

One common critique of the ecological approach is how can we use perception to explain behaviour that is organised with respect to things in the world that aren't currently in our area? How do we plan for future activities, or how do we know that the closed fridge has beer? 

A recent attempt to get ecological about this comes from Reitveld & Kiverstein (2014) who propose a relational account of affordances that enables them to talk about opportunities for more complex behaviours. This account has developed into the Skilled Intentionality Framework (e.g. Bruineberg & Rietveld, 2014), where skill is an 'optimal grip' on a field of task-relevant, relational affordances. 

I have always had one primary problem with this programme of work - I don't believe that they can show how these affordances create information and thus can be perceived. I discuss this here and here, and there's comments and replies for Rietveld and Kiverstein there too. You can indeed carve the world up into their kind of entities, but if they don't create information then they cannot be perceived and they are irrelevant to behaviour. 

I was therefore excited to see a new paper from the group called 'General ecological information supports engagement with affordances for ‘higher’ cognition' (Bruineberg, Chemero & Rietveld, 2018; hence BC&R). There is a lot of excellent work in here; but their proposal for a general ecological information is, in fact, neither ecological nor information. It is a good way of talking ecologically about conventional constraints on behaviour, but it doesn't make those perceivable and so the main thesis of the paper fails. 

General Ecological Information...

The typical ecological information story is that the properties of dynamical objects and events in the world interact with energy media and those dynamical properties are projected as specifying kinematic patterns into this media. Because of the ecological laws that govern that projection, the specifying pattern can be used as information for the dynamical properties and direct perception is possible. The laws constrain the pattern to be a certain way when a certain property is present and this constraint supports perception of the property.

BC&R need to identify a way in which non-lawful regularities can create information. For example, the presence of a beer can can be used as informative about the presence of beer in the can, because of the (non-lawful) socio-cultural regularity that beer cans typically contain beer. It's not a law because it is physically possible for something other than beer to be in the can. We seem to be able to organise our beer-seeking behaviour using the presence of the beer can, though, so how do we access the Reitveld & Kiverstein style affordance of the beer?

They propose General Ecological Information, which they define in the following way:
General ecological information is any regularity in the ecological niche between aspects of the environment, x and y, such that the occurrence of aspect x makes the occurrence of aspect y likely. Because of the regular relation between the aspects of the (sociomaterial) environment x and y, general ecological information allows an animal to couple to a distal (i.e. not sensorily present) aspect of the sociomaterial environment.
General ecological information is not limited to aspects of the environment that are sensorily present: something (say a bird of prey, aspect y) does not need to be sensorily present to get me ready to act on its affordances, because its shadow (the shadow of the bird, environmental aspect x) can reliably inform me about the presence of aspect y, even though in exceptional cases the shadow (aspect x) might be caused by a different aspect of the environment than aspect y (say for example aspect z, a kite). This example of the bird and its shadow also shows that the case of such general ecological information—due to the regularities in our ecological niche—is such that an aspect of the environment constrains (but does not necessarily specify lawfully) another aspect of the environment.
In their example, the fairly stable regularity that birds make bird shadows constrains what the shadow can possibly mean, and this (weaker than a law but still useful) constraint can also support perception. Not Ecological Information

The problem here is a common one - BC&R mix up 'the world' and 'information' levels of analysis. All of the constraints they discuss exist in the world; birds and shadows, beer cans and beer. This might exist, but unless they alter, say, light, they have not created information, nor can they themselves BE information. 

Let's step back to specification. The relevant constraint that allows the pattern in light to serve as information about the dynamic property that caused it is the ecological laws governing how light responds to interacting with surfaces, etc. The constraint exists between the dynamical property and the energy media. All of the BC&R constraints exists between dynamical properties; so no information.

So what is it?

Their analysis is close to the right answer, I think. They are right to identify the world-world constraints that make the presence of beer cans informative about the presence of beer, etc. I also think that the broad Skilled Intentionality Framework that gets them to these increasingly complex constraints is a really good way to structure the search for behaviourally relevant ones. But as I've just described, they can't get these constraints to the perceiver because those constraints, while real, aren't things that structure energy media.

General ecological information is therefore just a fancy way of labelling the situation semantics constraints that Chemero has already used in his theories (see this post and Chapter 6 of his book). I quite like the situation semantics approach and I think BC&R are using it rigorously and usefully here to point out that because of the various real socio-cultural and material constraints that exist, the presence of a beer can is informative about the presence of beer. But these constraints are not ecological information nor do they create any, and therefore cannot support perception of the constraints and the complex, relational affordances they want to be able to use.

Are We Screwed?

No. Sabrina has already solved this problem, in her 2015 paper. That paper introduces the idea of the conventional (non-lawful) use of ecological information. 

For perception. organisms only interact with patterns in energy media, and these patterns were all lawfully created and are thus constrained in a way that allows them to specify the underlying dynamics that caused them. Organisms, however, have to learn how to use those patterns as information about the dynamics (see this post and the Wicklegren & Bingham (2001) paper I discuss there). This creates a possibility, namely that while you can learn to use the information to be about the property that caused it, you don't have to. It also means that you're still just interacting with information; no special learning process is required.

Sabrina splits information use into two categories, law-based and convention-based: 
When an organism uses law-based information about affordances, it organizes its behavior with respect to the property in the world that causes the information. When an organism uses conventional information, it organizes its behavior with respect to evolutionarily, experimentally, ecologically, or socioculturally maintained conventions that link the information to properties or events. 
She (and I) then reserve the term affordances for properties that create information that gets used in a law-based manner to support the online coordination and control of actions (e.g. detecting the beer can and using that information to organise your reaching and grasping behaviour). Using information in a conventional manner (e.g. detecting the beer can and using that to organise your behaviour with respect to the presence of beer) is not affordance-based, but it is still information-based. In both cases the organism detects the information pattern and learns to use it to support some behaviour, and that works to the extent the relevant constraint supports that learning. In neither case does the organism need to 'know' about the constraint. 

The other consequence that Sabrina identifies is that the law-based use of information can support both action control and action selection. If I couple my reaching and grasping behaviour to the information caused by the size and shape and location of the beer can, I can use that information throughout the reach to go to the right place and grasp appropriately. Convention-based use of information can only support action selection, however; I can choose to grasp that can rather than this can because the former is a beer can and the latter a Coke can and there is a constraint that means when I grab beer cans I tend to get beer. That constraint does not support real time action control in anyway; the fact beer cans typically have beer has no bearing on how I actually control my arm other than providing a target (action selection). 


The main problem with this paper remains the old one - the concept of affordances as perceptible opportunities for behaviour simply cannot be stretched to support 'higher' cognition without becoming just a word in a functional description, as opposed to a real part in a mechanism. Sabrina is right - the only way to be ecological about things other than sensorimotor control is to expand what organisms are able to learn to do with patterns in energy arrays. 

So the main thesis of the paper fails. But the rest of the paper is an excellent analysis of conventional behaviourally relevant constraints and it has confirmed my growing sense that the Skilled Intentionality Framework is a powerful and valuable tool in the ecological kit. It's just not about affordances; but luckily, not everything has to be. 


  1. I haven't read the BC&R paper (but I've put it on the list!) but what you've pulled from it sounds vaguely Brunswikian. Would you consider that an accurate assessment?

    1. I wouldn't want to commit these guys to anything but I think yes, it's all in the same ballpark