Attention All Attention Skeptics

by rsbakker

Wayne Wu, this month’s featured scholar on Brains, has another fascinating round of posts up on the topic of attention. In “What is Attention?” he mentions how he seems to be bumping into more and more attention skeptics, people who either doubt the cognitive science research community’s ability to find any consensus definition of attention, or doubt that attention exists at all.

“How in the world could that possibly be?” he asks.

As it turns out, this question is largely rhetorical for Wu, so I thought I would take a preliminary run at an answer. I should note that the research and literature on this topic runs very deep, and that I’m only conversant with the broad issues. But Wu’s argument amounts to what Anthony Chemero calls ‘Hegelian arguments’ in cognitive science: an nonempirical attempt to regiment empirical priorities. It is philosophical, through and through.

Wu begins with what he takes to be a consensus starting point: the tasks performed by attention in various modalities of perception. These tasks, he points out, provide the experimental paradigms generally used to research attention, be it visual search or spatial cuing or object tracking and so on. As he writes, “for each of these attention paradigms, there is a fundamental assumption, namely that the task defines some relevant targets, and that where the subject selects those targets to perform the task, the subject is attending to the target.” What he wants to argue is that an implicit definition of attention is built into the very structure of these experimental paradigms. Adopt any one of them, Wu argues, then you are tacitly endorsing a sufficient condition of the form, If S selects X for task T, then S attends to X.

This, Wu thinks, warrants prioritizing task-based approaches to attention over neural-based approaches, simply because, as he puts it, “to investigate each of these phenomena, neuroscientists must use task-based attention paradigms where by defining a task and a target, they can carefully control their subjects’ behavior and, where the subject correctly performs the task, they can infer that their subjects are attending as they should.” In other words, they must endorse his sufficient condition to even get their research off the ground.

I admit, it sounds pretty ironclad, but only, as I hope to show, because of the role played by neglect.

What fascinated me reading this particular argument for the first time was the way it recapitulates the structure of so many arguments against so-called ‘eliminativist’ approaches in cognitive science. Wu’s terms obscure this fact, but one need only review the concepts he employs to describe his ‘task-based’ approach to see that he is giving us a broadly intentional construal of the ‘general experimental paradigm,’ one where experimenters ‘control’ behaviour such that subjects perform as they ‘should.’ What he calls, the ‘neural based’ approach to attention, on the contrary, is primarily mechanical in emphasis, bent on describing and explaining what is actually happening in our brains when we ‘attend to X.’ What Wu is arguing, in effect, is that any mechanical approach to the question to attention conceptually and operationally presupposes the intentional structure of his general experimental paradigm–and obviously so.

But this claim is far from obvious. Given Blind Brain Theory (BBT), for example, the ‘obvious’ way to approach the question of attention is directly opposite Wu’s. On BBT, our intuitive conception of attention is blinkered to the degree it turns on metacognition. Our brains are all but opaque to our brains, thanks to their astronomical complexity, among other things. Given considerations such as these, it is almost certainly the case that the intentional characterization employed by Wu in arguing his task-based approach involves drastic heuristic simplifications of what is actually going on. Sure, these intentional characterizations of the ‘general experimental paradigm’ suit the needs of scientists and subjects alike in any number of communicative contexts, but their applicability to questions like ‘What is attention?’ is by no means clear. What if it’s the case that the very information neglected to facilitate Wu’s ‘task stance’ is the very information required to answer the question of attention?

The heuristic nature of the task stance means neural-based approaches do not presuppose task-based approaches either conceptually or operationally–any more than this blog post presupposes stenographer’s shorthand. Sure, short-hand efficiently discharges a number of functions within a comparatively restricted domain, namely, those problem ecologies (such as court proceeedings) it is adapted to solve. This blog post, however, is not one of those ecologies. Precisely the same, I think, can be said of what I’m calling Wu’s ‘task stance,’ the big difference being that its ‘applicability conditions’ are nowhere near so clear. The task stance is the most economical way to conceive the experimental scene because it the most economical way to conceive human action. But why should either of those economies apply to the empirical question of attention?

This is where neglect comes in. You see, what makes Wu’s task-based interpretation of the ‘general experimental scene’ seem the obvious ‘only game in town’–what makes it paradigmatic–is simply the fact that the neural-based description of that same experimental scene remains unknown. As the only way Wu can think of to describe the scene, it seems to become the only way for the scene to be conceived, or ‘paradigmatic.’ The issue of its heuristic applicability to the question, What is attention? is accordingly lost. As the operational kernel of every attempt to understand attention, applicability seems to be implicitly given.

So the reason Wu’s argument leapt out for me, why the task stance, far from seeming the only game in town, struck me as a parochial means of understanding the experimental scene, is simply because BBT has, for quite some time now, had me looking at psychological experimentation mechanistically as kinds of information extracting meta-systems consisting of the regimented interactions of various subsystems, what we intuitively think of as ‘researchers’ and ‘subjects’ and ‘experimental apparatuses.’ As a result, I now generally look at intentional characterizations like Wu’s against this baseline, as information-neglecting heuristics, not so much accurate descriptions of what is going on as economical ways to navigate what is going on given certain problem contexts.

From this biomechanical standpoint, Wu is attempting to tackle the question of attention ‘on the cheap.’ But unless one wants to argue that intentional characterizations are not heuristic, or that they are heuristic but somehow remain applicable to the theoretical question of what attention is (despite neglecting, as intentional heuristics seem to do, the very causal information so much scientific explanation requires), then it would seem that the biomechanical way of understanding attention is the only way of knowing what it is apart from our nebulous experiences. And since knowing what it is within our experience turns on what attention is apart from that experience, it would seem that even this ‘intuitive,’ or ‘phenomenological’ aspect of attention, requires the priority of neural-based approaches to be understood.

If it exists at all.

My guess is that like other folk faculties, attention will be progressively revealed to be more fractionate, that its intuitive lack of internal structure–it’s simplicity–will be shown to be a byproduct of neglect. Attention, as we think we presently know it, is actually quite easy to doubt when you look at research into other faculties like memory and reason. The more we learn, the more complicated attention becomes, and the more informatically impoverished intuition is shown to be. Memory isn’t an aviary. Reason isn’t a charioteer battling unruly moral and immoral horses. Odds are, attention isn’t a selective spotlight. We should expect fractionation, surprises–continuous complication. And even if you have faith in the theoretical accuracy of metacognition, the bottom line is you simply don’t know where those intuitions sit on the information food chain. Nothing need be accurate about our intuitions of brain function for the brain to function. Given this, using them to conceptually and operationally anchor an empirical research program smacks less of necessity than a leap of faith.