Pages

Friday, 5 February 2016

On "The poverty of embodied cognition" (Goldinger et al, in press)

A new paper in Psychonomic Bulletin and Review (Goldinger, Papesh, Barnhart, Hansen & Hout, 2015) has taken a swing at the field of embodied cognition, claiming that it is vague, trivial and unable to add anything scientific to the investigation of cognition.
...our goal is to zoom out from specific empirical debates, asking instead what EC offers to cognitive science in general. To preview, we argue that EC is theoretically vacuous with respect to nearly all cognitive phenomena. EC proponents selectively focus on a subset of domains that work, while ignoring nearly all the bedrock findings that define cognitive science. We also argue that the principles of EC are often (1) co-opted from other sources, such as evolution; (2) vague, such that model building is not feasible; (3) trivially true, offering little new insight; and, occasionally, (4) nonsensical. 
My basic take is a) I actually agree with a lot of the criticisms in the context of the kinds of 'embodied' cognition we critique for similar reasons, but b) there is nothing new to any of these critiques, none of them are compulsory failings of the field and nothing about them makes embodiment an intrinsically empty notion. 

The authors identify three theoretical commitments of embodied cognition:
  1. cognition is influenced by the body, including its potential actions
  2. cognition is influenced by the environment, including off-loading
  3. cognition may not require internal representation
They then ask what can you do with these commitments, in terms of science? They address 9 major topics in cognitive science in detail (e.g. face perception, priming, categorisation) and see if the three commitments add anything non-trivial to our understanding of the topic generated from standard cognitive science. They also list 20 additional topics they say "appear challenging to explain from an embodied cognition perspective". These are all things that are out beyond topics like catching fly balls, the kind of topics they accuse embodied cognition theorists of ignoring empirically but claiming they can probably explain. In each section, they review the basic topic and then conclude with something along the lines of "we cannot think of how those EC commitments help you explain this topic"; 

They conclude the paper
All EC researchers can remember their mothers’ names, their favorite toothpastes, and their fifth-grade teachers. They can discriminate toucans and penguins, and appreciate that neither is very “birdy.” They should also appreciate that embodied cognition cannot logically explain these basic aspects of mental life.
Analysis
Argument: there's lots of major topics that embodied cognition simply ignores 
We get this thrown at us a lot; "explain X without representations!", where X is that person's favourite topic. Here's the thing; we can't, always. There really are huge areas of behaviour that embodied cognition theories have gone nowhere near, there really are major hurdles to getting these theories to work with things like language and the rest. It is a completely valid point that EC has not tackled everything.

One answer is simple that we just haven't tackled it yet. This is (I think) a valid response I sometimes use, although I completely understand that it is not satisfying to the other person. But EC researchers have indeed spent a lot of time on certain topics which means time not spent on other topics, and any new theory goes for the low hanging fruit first to get moving. It's not ideal, but it's not a disaster, and personally, our work right now is about contributing to the developing theoretical framework that provides the tools required to move into these new areas. In other words, as Twitter knows I like to say, we're working on it.

Argument: we cannot see how the principles of embodied cognition could ever generate an explanation for this topic
This is another common one, but I don't find it convincing. My main reply is "don't mistake your lack of imagination for my problem". It takes a long time to immerse yourself in both a topic and a theoretical framework and generate hypotheses from the framework for the topic, and when you are already steeped in a different framework your view of the other one is necessarily limited. Just because you can't do it doesn't mean no one ever will, and there are many people (us, Chemero, and more) trying.

Argument: it is impossible to generate formal models from embodied cognition accounts (bold because their claim is this strong)
How might we write an equation that expresses embodiment? How can the environment (such as the affordances of various objects) be parameterized?
I will concede that lots of embodied cognition suffers from this. People really do make wild claims about the magical powers of affordances based solely on a verbal description, and these claims have significant problems that we have discussed many times. 

But at one level this one is just embarrassing for the authors. There is a rich field of dynamical systems modelling, some of which is also ecological, and these models have been extraordinarily generative. Bingham's coordination model and Thelen and Smith's A-not-B error model spring to mind; there are many others. But, for example, my latest paper is literally a formal, quantitative description of an affordance, and the coordination model is literally an equation that expresses embodiment, and they are not isolated incidents. The fact the authors don't know any of this work reflects very poorly on them, and the fact that they think it is impossible is just amazingly wrong.

Argument: embodied cognition is logically unable to handle many topics because we need things like mental representations
This style of argument will never convince anyone who doesn't already agree. I know this because I've tried variations on it from our side and been rightly shot down. It also feeds back into the 'just because you don't know how to make it work...' answer from above.

Summary
Here's the funny thing. I basically agree that most of the field known as embodied cognition suffers from serious, perhaps even fatal flaws. We even got into a lot of this in the Wilson & Golonka paper. A lot of work really does just amount to a nice verbal description and some fancy words, and this is indeed a real problem. There really are huge swathes of cognition that none of us have tackled properly, and there really are many serious and unsolved problems such as what to replace mental representations with.  But these are all old problems and this paper does nothing new. 

More importantly, these problems are all with the state-of-the-art, and are not in-principle objections to embodiment. While we as a field have not tackled everything, we have many well worked out examples that include detailed analyses of the relevant information, formal models that generate testable and supported predictions, and serious forays into more complicated, 'higher-order' territory. The existence of the many success stories across embodied cognition is, I believe, positive evidence that this really is a hypothesis worth pursuing, so long, of course, as we up our game and science as hard as we can. 

And finally, I want to emphasise that the main thrust of this paper comes from the authors' inability to figure out how any of this might work. While I do agree it's our responsibility and not theirs to put forward positive theories, this amazingly common argument is nonsense and should never be given any weight, let alone space in a journal. I hope we can all agree on this, at least.

References
Goldinger, S. D., Papesh, M. H., Barnhart, A. S., Hansen, W. A., & Hout, M. C. (in press). The poverty of embodied cognition. Psychonomic Bulletin and Review.

2 comments:

  1. Better hope Daniel Hutto doesn’t find out about this paper. He’d be out of tenure in a hurry.

    ReplyDelete
  2. Really very happy to say, your blog is very interesting to read. I never stop myself saying something about it. You’re doing a great job. Keep it up.
    https://blog.mindvalley.com/cognition

    ReplyDelete