“The way that human minds arrange particulars is a skill so deeply hidden in the human soul that we shall hardly guess the secret trick that Nature here displays” – Immanuel Kant
This is the first part (of three) of the talk ‘What I Know and Why I Know It’. A fair proportion of it covers ground already covered in various previous posts but pushes the argument a bit further along.
The talk will hop between bouts of Neuroscience, Philosophy and trying to fuse the former two disciplines together. We’ll start with some neuroscience…
1. Big / Small
We have a good understanding of the large-scale view of the brain and of the (very) small-scale view of the brain’s components, but we lack an understanding of how the two are related.
The large-scale view is understood through observing the behaviour of others (either as an everyday activity or through psychology) and, uniquely as a method, introspectively, of ourselves.
The small-scale view is understood through basic neuro-science – the functioning of neurons (and the glial cells), how they connect to one another via synapses so that when one neuron fires, other neurons that it is connected to may subsequently fire.
2. Cortical Columns
The small- and large-scale views are seemingly related through there being 16 billion-odd neurons in the cortex within the brain, with thousands of neurons together forming cortical columns and millions of those columns forming the cortex – sheets of these columns that are then scrunched up to give the distinctive wrinkly appearance in order to fit within our skull.
Besides the general astonishment of such a ‘hypothesis’, it is also remarkable that the structure that gives rise to such varied functionality and behaviour is so uniform. Vernon Mountcastle famously observed this in ‘An Organizing Principle for Cerebral Function’ (1978; note from this and subsequent dates how recent much of this work is). There are phylogenetically older regions of the cortex that are not ‘neocortical’, with fewer that the 6 layers observed by Santiago Ramon y Cahal, such as the cingulate cortex. There are regions such as the Hippocampus in which there is a seemingly more chaotic organisation between grey matter (un-myelinated neurons) and white matter (myelinated neurons). There are regions in which the size of the cortical columns have a significantly of a different size that elsewhere (such as the large columns within the lower visual regions e.g. ‘V1’. But generally, the brain is remarkably uniform across regions traditionally with visual sensation, auditory sensation, motor control, higher level cognitive functions and so on.
Above the level of the physical ‘building block’ that is the cortical column, we may discern a hierarchy of higher-level components (groups of many many columns) such as those famously mapped within the visual system of Macaque monkeys by Felleman and Van Essen (1991). At the bottom (marked ‘RGC’) are the retinal ganglion cells – the light-sensitive neurons with the retina of the eyes. These connect via the Lateral Geniculate Nuclei (‘LGN’) within the Thalamus up into the primary visual areas of the Cortex, such as ‘V1’. There are then many layers of groups of columns up eventually to the Hippocampus (‘HC’). There is a hierarchical structure in that columns typically connect to columns either in the same layer, the layer above or the layer below but not further afield.
How could all this interconnectivity between groups of neurons produce the advanced behaviour of the whole?
4. Adaptive / Predictive Hierarchical Models
A Previous talk (‘Intelligence and The Brain’) introduced Karl Friston’s ‘Variational Free Energy’ theory (2005) of the what the brain might actually be doing in order to produce advanced behaviour from assembling billions of relatively simple components. To recap at an almost-absurdly simplified level:
- There is a hierarchy (as noted above) from low (connected with the environment via senses and motor controls) to high.
- Within each group of columns, a model is formed of the behaviour of everything below it in the hierarchy.
- The behaviour of the group of columns is modified through a combination of action (motor output to lower levels) and adaptation (modification of the model in response to sense inputs to lower levels. Overall, the behaviour is said to be that of ‘minimizing surprise through action and perception’.
- Downward-going signals are predictions. Upward-going signals are prediction errors.
- When something happens in the outside world, prediction errors will propagate up the hierarchy. Predictions and errors will rattle around the many levels of hierarchy until behaviour settles. (In reality, something is always happening and prediction never settles.)
Within the above framework, learning is the process of going from being unable to predict to be able to predict. These may be equated with ‘not knowing’ and ‘knowing’ respectively. This process may be seen a (locally) counteracting entropy. In the previous talk (see ‘Entropy, Intelligence and Life’), this was explained with an example of the process of learning what the capital city of Australia is, with reference to boxes containing different densities of gasses, in order to make the connection to entropy. In its most abstract form, intelligence may therefore be seen akin to biological life.
In this talk, we do not need to concern ourselves with any connection to entropy. It does not matter what the underlying technology of our ‘groups of columns’ is – be it neurons, electronic transistors or whatever. In making this point, John Searle famously referred to using ‘beer cans and windmills’. So we could view those dots in boxes representing gas molecules in the different chambers alternatively as empty beer cans used as voting tokens. The group of columns is trying to move from a state of indecision/ignorance (beer cans distributed across many candidate boxes) to a state of knowledge (the majority of beer cans concentrated on one candidate).
6. A Unified Brain Theory
The above described behaviour can alternatively be presented as the simultaneous operation of two orthogonal processes – a ‘horizontal’one and a ‘vertical’ one.
The horizontal process is the ‘pushing-together’ (of beer cans) in order to formulate an opinion. In the absurdly simple capital-of-Australia example, the options are obviously mutually exclusive. In more realistic scenarios, it may not be that they are logically mutually exclusive. It may seem fairly arbitrary at this stage – what is important is that a single choice is being made and it is irrelevant whether it is the right choice or even vaguely a good choice.
The vertical process is the ‘pulling-together’ across the many levels of hierarchy. This helps ensure that the ‘opinion’ being formulated at any one level in the hierarchy fits in with other levels – and ultimately with the external environment.
Simultaneously, the horizontal and vertical processes push and pull towards forming good opinions, good models of the outside world which allows the brain to make predictions of the environment.
(The pushing and pulling of a network of rubber bands, pulleys and rods may be a better mechanical analogy than Searle’s ‘beer cans and windmills’.)
Epistemology is a major branch of Philosophy alongside Metaphysics, Ethics and Political philosophy and concerns problems surrounding knowledge and what it means to know something.
Within epistemology, knowledge has traditionally been defined as ‘Justified True Belief’ (and commonly credited to Plato). Knowledge is belief that is deemed to be true by virtue of some justification. There are three main theories of truth – differing in how the truth is justified:
- The correspondence theory of truth
- The coherence theory of truth
- The pragmatic theory of truth
The Correspondence theory of truth is the dominant theory and the most common-sensical. An example: I see a glass of water on the table in front of me. I claim to know that there is a glass of water on the table because:
- I see the glass and it corresponds to my understanding of what a glass is inside my head.
- I see the table and it corresponds to my understanding of what a table is inside my head.
- I see the relationship between the glass and the table and it corresponds to my understanding of what ‘on’ is inside my head.
This seems so obvious, how could anyone believe otherwise? Well, the same argument could be applied to our observations of the Sun:
- I see the Sun low down in the sky in the East in the early morning.
- I see the Sun high up in the sky around midday.
- I see the Sun low down in the sky in the West in the evening.
These observations all correspond to the idea in my head that the Sun is moving around the Earth. And yet, we do not believe that the Sun moves around the Earth but it is the other way around.
According to the Coherence theory of truth, we believe the Earth moves around the Sun and not the other way around because we have learnt that this idea coheres with other ideas in our heads, such as the idea that apples fall down towards the Earth. The separate ideas form a coherent ‘story’ of gravity.
But the Coherence theory is also problematic. Consider a situation in which someone is hynoptized into believing their left arm does not belong to them. They are then asked to empty a bottle of water into a glass and they proceed to do so by trying to take the lid off the bottle with their right fingers whilst awkwardly holding the bottle in their right hand. When asked why they are doing it like that, they reply that they slept awkwardly on their other arm last night and so are giving it a rest. In short, the confabulate a story: they try to build the most plausible coherent story they can from the circumstance they find themselves in. Yet their story clearly bears no relationship with what’s happening in the outside world, which is patently obvious to the hypnotist’s audience.
The Pragmatic theory of truth can be summarized by saying that truth ‘is what works’. But this sounds rather strange. I might believe that your nice house and nice car is actually mine, since it ‘works for me’. But this is expediency rather than truth. This is not what is meant by ‘works for me’ and I shall return to the Pragmatic theory later on.
8. A Better Theory
Before that, I want to look at improving on the Correspondence and Coherence theories.
The Foundationalist theory of truth is an attempt to improve on the Correspondence theory of truth by accepting that there are so-called ‘basic beliefs’ which have no justification of correspondence and are therefore axiomatic. These beliefs form the foundation for other beliefs which are justified by correspondence. I mention this theory not because it solved a previously mentioned problem of the Correspondence theory but just to note of its existence for what follows…
In her 1993 book ‘Evidence and Inquiry’, Susan Haack introduced the Foundherentist theory of truth which is an attempt to synthesize a better theory from the Foundationalist and Coherentist theories such that it takes the advantages of both without their drawbacks. Hence the rather unwieldy name. It is thus indirectly a synthesis of the Correspondence and Coherence theories of truth. Not surprisingly, the Foundherentist justifies knowledge by both correspondence to evidence and coherence with other knowledge. But the theory is best illustrated by a very good analogy created by Haack: the crossword puzzle.
In a crossword puzzle, we try to find answers to questions to fit into the grid. If we find that an answer fits in with an already completed question, we gain confidence that the answer is right. If it does not, doubt is then raise either about our proposed answer or the answers that conflict with it in the grid. In this analogy:
- The answers to the clues are analogous to knowledge corresponding to evidence.
- The grid provides the framework for building a coherent set of answers.
If we ignore the grid, we are left with just a quiz: a list of questions to which you are expected to provide answers where each question is quite independent of the others. This is analogous to the Correspondence theory of truth.
If we ignore the clues, we do what you might have done when you’ve got so far with a crossword and got stuck – try to fit any words into the remaining spaces in the grid. This is analogous to the Coherence theory of truth.
Note that Haack is also a Pragmatist. The Pragmatic theory of truth sees truth as an ongoing process of inquiry, building better predictive models of the environment. Her philosophy ths unites the 3 basic approaches to epistemology.
9. Neuroscience and Epistemology
You may now see where I’m going with this argument. I started off by presenting a physicalist description of how the brain works in terms of the simultaneous combined effects of:
- The horizontal ‘pushing-together’ process in order to formulate an opinion, and
- The vertical ‘pulling-together’ process across the many levels of hierarchy, linking predictive models to the outside environment.
and I finished with a philosophical theory of knowledge creation being an ongoing process of the combination of:
- a (horizontal) coherence with other knowedge
- a (vertical) correspondence with evidence
The two visions look rather similar! What has been presented is a physically grounded theory of knowledge – a ‘natural epistemology’.
(Furthermore, there is the similarity between the hierarchical nature of Foundationalism and the hierarchy in the brain.)
I seem to be going beyond the ‘cooperative naturalism’ position within ‘naturalized epistemology’– where science (and in particular, neuroscience) can help inform epistemology (‘cooperative naturalism’) such as when Haack says (‘Evidence and Inquiry’, p. 118):
“ … the results from the sciences of cognition may be relevant to, and may be legitimately used in the resolution of traditional epistemological problems”
…to an all-out ‘replacement naturalism’ which rejects the philosophical approach in favour of the scientific.
Re-iterating the quote given at the start (with bold emphasis added), Kant said (Critique of Pure Reason, 1781):
“The way that human minds arrange particulars is a skill so deeply hidden in the human soul that we shall hardly guess the secret trick that Nature here displays”
With progress in neuroscience following on from that in psychology, it can be argued that we will increasingly be in a position where we no longer need to guess. Nature’s secret trick is (gradually) being uncovered.