I was reminded at work this week of the quip amongst developers and programmers, ‘It’s not a bug, it’s a feature.’
Image from I’m a Programmer http://www.improgrammer.net/not-bug-feature/
So, I wanted to take a look at two particular features of the brain, and the “features” (hereafter, bugs) in thinking that these lead to. Unlike Kahneman’s biases and heuristics program, I am speaking about features of the brain as a whole.
The first feature is the neural cascade (see figure 1, a depiction of an Evolving Cascade Neural Network a simulation of human cognition used in Machine Learning, Artificial Intelligence, and so on). The neural cascade is descriptive of the fact that the top-most neuron (y) might be an output that equates to a label or concept. Each such neuron has two or more neurons as inputs (the so-called “hidden” neurons z3 and z2), which in turn have two or more (hidden) neurons as inputs (z2 and z1), all the way back to the originating sensory neurons (xn). The bug that derives from this is Platonic Idealism and related ways of thinking.
From Schetinin, V. (2003). A learning algorithm for evolving cascade neural networks [ECNN].
Neural Processing Letters, 17(1), 21-31.
The second feature, which I’ll discuss next week, is Kahneman’s cognitive ease, also called cognitive fluency. So, let’s discuss these brain bugs in a bit more detail, starting with a bug that arises from the neural cascade, Platonic Idealism.
We all know that the label is not the thing so labelled. And we know that what we hold in mind is, likewise, the representation, not the thing represented. This is why Hume said that we never experience cause-and-effect we merely infer it. On the one hand, this is obvious; having a 3D printer in your head would be closer to having the thing, not the representation, but it would still be a representation. On the other hand, it’s not quite so obvious, experientially, because many instances of cause and effect are so commonplace (I drink coffee in the morning, I feel less homicidal, etc.), it feels as though we experience the interaction.
We have labels for things. Sometimes the fact of a clear and distinct label can give the impression that the thing is somehow more important, or more “real”, than another thing with no label (or a label that is difficult to spell or pronounce; see cognitive ease, next week). I think we can agree, though, that an apple is a thing.
In being a “thing” and one with a well-understood label, it is the peak of a neural cascade, made up of a collection of “hidden” neurons, for most of us. The specific content of the intervening hidden neurons we can only guess at. However, the content is made up of the outputs from sensory neurons, tuned to detect edges, shape, colour, texture, and smell.
Platonic Idealism would suggest that the constituent elements of our experience of an apple are also things in their own right – like ‘circle’ and ‘red’ – the apple, in effect, is an interaction between these things (which flies in the face of Hume’s contention…and I’d probably side with Hume). Clearly, circularity and redness are features of apples (and endless discussions about qualia), but Platonic Idealism would have it that they exist in a “real” sense as well, just not on this profane plane of reality.
I want to suggest that the only reason Plato thought that Ideals existed was due to the fact that labels (words) had been created for these things, and thus they are peaks of their own neural cascades. Words, in the first instance, seem to exist because we want to (and can) distinguish between the things so labelled. A circle is not a square, red is not blue. And we thus have mental representations for prototypical circles, squares, red, and blue.
An apple, as represented in the mind, is a confluence of neural peaks for the aspects that go into what is an apple: redness (or greenness, depending upon the apple, and your preferences); roundness; crunchiness; sweetness (or tartness, again, depending upon the apple and your preferences); and thus “appleness”.
Our minds have tolerances for variations from these centrally defining aspects, as intimated above. We are comfortable calling something an apple, whether it is red or green (or even sort of yellow, in the case of Golden Delicious). We are comfortable calling something an apple, depending upon how round it is, how crunchy it is (before we qualify by saying that it is ‘stewed’, ‘over-ripe’, or just plain ‘gross’), how sweet it is, and so on. We are comfortable calling something an apple based on taste-related factors, many of which we would actually be hard-pressed to put into words (aside from sweet or tart, or possibly floury, we might fall back on “apple-y”).
“Appleness” seems to be a combination of these necessary (but not sufficient) aspects, as witness the degree to which these aspects can vary without creating too much difficulty in finding the label ‘apple’. However, when one single feature is completely absent the degree to which the others contribute to whether something is an apple can be irrelevant, as in the case of the nashi, which looks like an apple, but which you know is not an apple the moment you taste it.
Nashi (Asian pear), image from Wikimedia
These aspects (redness or greenness, roundness, crunchiness, sweetness, tartness, etc.) have their own cascade, capped by an output-neuron (y) that is (or, more likely, is connected to) the neuron that is the label. Indeed, the output neuron might reasonably be called the concept-neuron. I want to suggest that the label-neuron (in the language area of the brain) is the peak of its own cascade, one that includes the physiological process of pronunciation, the sound of that pronunciation (and tolerances for accent), the appearance of the written word and letters (and tolerances for variations in written and printed representation), and so on. This neuron is associated with the concept neuron that it labels.
The Jennifer Aniston & Halle Berry Neurons
A little over a decade ago, neuroscientist Rodrigo Quian Quiroga published a paper (pdf) detailing his discovery of the “Jennifer Aniston neuron.” Looking for the areas of the brain that cause epileptic seizures, Quiroga discovered that one subject had a neuron that steadily fired whenever she was shown a photo of Jennifer Aniston. It didn’t fire for other celebrities, but seemed linked to the concept of Jennifer Aniston. Another subject had a Halle Berry neuron, and another had one that fired in response to Bill Clinton.
We don’t all have Jennifer Aniston neurons, nor is there one neuron whose sole job is to recognize Jennifer Aniston. (Quiroga says there are likely many neurons that fire in response to the actress, and that this neuron likely also responded to other concepts he didn’t test for.) But the discovery of a neuron that is linked to a particular concept is a major milestone in understanding how the mind works. – Olivia Goldhill in Quartz
Jennifer Aniston (Image from Starschanges.com)
To tie this back to the neural cascade, here’s an amusing commentary on the Jennifer Aniston Neuron in particular:
Maybe the neuron flashing “Jennifer!” isn’t acting alone. Maybe it’s the beacon atop a summit of associated neurons, each of which is flashing little bits of Jen. After all, when I say Jennifer Aniston, what comes to mind? Her blue eyes. Her chin (I’m slayed by her chin), her hair of course, her voice, her laugh, her fictive romance with Ross on “Friends”, her dresses at the Oscars, her marriage to Brad Pitt, the divorce, her movies. And more fundamentally, those Jen-parts are themselves built from rawer bits: the blue of her eyes, her smooth skin, her flowing hair, her facial patterns (especially chins), which are in turn made of even smaller bits: units of light, dark, color, shading, form. Those lower-down neurons are no doubt doing other things in my brain (blue-ing up skies, lakes, and other people’s blue eyes) when they’re not building Ms. Aniston, but when Aniston calls, they come. – Robert Krulwich in NPR
So we have a single neuron that represents the concept of Jennifer Aniston (depending on how much you watched Friends, I guess). It’s likely that certain pictures, when used as stimuli, will be more consistent in causing an Aniston neuron to fire (if you have one). On the other hand, if Ms. Aniston is just another fair-haired American actress to you, then you may have a ‘generic fair-haired American actress’ neuron that fires equally to Aniston, Christina Applegate, Jennifer Lawrence, and Jessica Alba. Of course, Alba is a brunette, but had iconic roles as a blonde (e.g. Sin City), and sadly for Mr. Krulwich (or rather his Aniston neuron), Aniston has brown eyes and hair, but wears blue contacts, and lightens her hair.
Speaking of Jessica Alba, just as relative “appleness” is a concatenation of multiple features with tolerances around a mean (average) definition of apple, so it is with what we consider beautiful in human faces. In the case of female faces those who are often considered amongst the most beautiful, in the West, have features that represent the average across Western society. As with an apple’s redness (or greenness) and roundness, so it is with eyeness (how the eyes sit horizontally on the face, between the ears) and lipness (how the lips sit on the face with relation to the eyes).
Jessica Alba apparently makes the grade on the horizontal and vertical location (see image below, displaying proportions discussed here), as does Shania Twain and Elizabeth Hurley. These women happen to have faces that are proportioned in the same way that the average across a sizeable sampling of women in the West. However, Jessica Alba probably beats those two on the other key features (“large eyes, prominent cheekbones, thick lips, thin eyebrows, and a small nose and chin” (discussed by Baudoin and Tiberghien (2004).
Jessica Alba – Averageness
What we consider to be the appliest apple is what sits right in the middle of our expectations for colour, shape, texture, smell, taste, and so forth. This will likely vary depending upon preference (Granny Smith, Braeburn, Golden Delicious, etc.). Equally, our preferences for faces are also tuned by exposure. What happens if we derive a mean from a number of contributory faces (see image below)?
You can rely on Pinterest for anything, like this “average” face made of
Alison Brie, Jennifer Connelly, Jessica Alba, Mila Kunis, and Scarlett Johansson.
In the case of apples, Anistons, and Albas, it is possible to refer to the source material to update one’s mental representation of them. Of course, with Photoshop and CGI it’s only the apple that you can be fairly sure about. In some ways, that’s the point, the representation of these actually real actresses that form part of how we understand what they look like is often impacted by someone else’s idealization of them. Those things that we experience often, like apples, people we know, and so on, are consistently recalibrated with exposure to stimuli. Not so with Platonic Ideals (which are updated by our experience, but we have no prior instantiation with which to compare the ideal itself), and not so with the supernatural or certain pseudoscientific ideas.
Platonic Idealism, Pseudoscience, and the Supernatural
An individual’s experience of an apple, or a female face, is made up of multiple interacting expectations, primarily based on the sum total of all (or at least most) personal exposure, and a degree of preference. As such “beauty”, even in the case of the Venus de Milo or the Mona Lisa, is not a perfectly encapsulated “thing”, it has atomic parts.
Speaking of atomic parts, both sodium and chlorine are dangerous elements depending on how you encounter them, and whilst table salt can be deadly, too, it is on an entirely different basis than either the sodium or the chlorine (and the exposure needs to be significant and ongoing). Salt is incredibly stable (with an electron arrangement similar to Argon, which is almost completely inert), unlike sodium or chlorine.
Image from VisionLearning
…and the point of this impromptu chemistry lesson that no one needed?
Plato extrapolated from experiences of things, extracted not just their atomic elements, but the prototypes upon which those elements were based, and claimed that they existed in reality. He extracted from the human level to the macro level. However, just as a ton of table salt is entirely different from the stack of sodium and the cloud of chlorine required to make it, so too would the perfect Platonic apple be quite different from a combination of prototypical red, a perfect circle, and so on.
In a similar vein, if we extrapolate from our experiences of human kindness, knowledge, power, and presence, the resultant super-kind, super-smart, super-powerful, able to be anywhere in the blink of an eye Super-man might make sense as an idea, but he would be just that, an idea (and I don’t think that DC or Nietzche meant otherwise). We can mentally glue together an ideal face from our idea of perfect eyes, lips, nose, cheeks, and eyebrows, but that doesn’t mean that we have created a coherent or real being (see Anime, Hentai, and Picasso). Equally, a being glued together from maximally extrapolated kindness, knowledge, power, and presence (i.e. omnibenevolence, omniscience, omnipotence, and omnipresence) is incoherent, and for the same reason.
We should maybe look at the impact of Photoshop and the ever more unrealistic depictions of female perfection that appear in our media, note the effect on our (particularly, but not exclusively) female youth, and our culture in general. Then we might wonder what ideas of perfect (and unattainable) kindness, knowledge, power, and presence have on those that aspire to that. (Controversially, I’d like to suggest that this maybe one reason behind the creation of an attainable perfection, in the form of Jesus, as compared to the unattainable “perfection” of God.)
The dehumanising influence of idealism (and Photoshop) from Digital Synopsis
The “macro” direction of perfection and Platonic idealism is not the only direction that extrapolations from experience can take. If redness or greenness, roundness, crunchiness, sweetness or tartness can be detached from the apple, and then manipulated and dissected, it can seem like the atomic elements of experience are being contemplated at the subatomic level. They aren’t of course, at least not in any sense other than conceptual, but also not in any sense that is constrained by reality.
The subatomic particles in sodium and chlorine are also in table salt. It takes quite a bit of understanding of the relevant physics to understand how the behaviours of those subatomic particles are expressed. You are unlikely to come to understand sub-atomic physics, or quantum physics by exclusively studying the behaviour of table salt. Equally, you are unlikely to come to an understanding of quantum physics by studying endocrinology or the human experience, a fact that Brian Cox attempted to illustrate to Deepak Chopra.
If Plato attempted to derive a multiverse or macro-level understanding of existence from human experience, then Deepak Chopra is attempting to understand the micro-level of existence in much the same way, and with about the same level of success. Doing the same thing some 2,500 years later, and expecting different results, really is a sign of madness, or at least pseudoscience.
So, neurons that sit at the peak of neural cascades represent whatever the concatenation of contributory sensory neurons, and intervening “hidden” neurons represent. This may be singular, indivisible primary sensations for which we have a prototype – like “red” or “sweet” – that we compare all other things to. In that sense Platonic Ideals exist, but only as derivations from experience in our minds, and Plato’s ideal “red” or “sweet” will be different to yours and mine, and ours will be different from each other’s, and thus they have no singular unified existence in some ideal plane.
Whilst we do glue these elements together in our minds, we also update our response to the source stimuli regularly, thereby updating the representation (and the elements that make them up), but not so with pseudoscience and the supernatural. There is no stimulus to refer back to for an update, no way of remaining grounded.
Concatenations of atomic elements of experience are not a good basis upon which to attempt to understand the world at the micro- or macro-level, but this what some forms of pseudoscience, and almost all supenatural beliefs are doing. When elements interact in the real world, they can lose constituent and definitional properties as sodium and chlorine do when they become table salt. The constituent properties of idealized elements, however, are not altered when they interact with one another, and this is a strong indication that they are not real.