“Strange to wander in the mist, each is alone. No tree knows his neighbor. Each is alone.” Hesse
Cognitive science can’t settle questions of right and wrong. But it can set the table for theories of rational decision-making and ethics. And the one thing we know for sure about how we think is that, at the most fundamental level, our brains are connection systems that are changed by experience. This matters; transformational experiences and self-altering decisions exist because of the way we think. Our minds change. They change in the shallow sense of changing what we remember, the facts we know or the beliefs we hold…but they also change in a much deeper sense. Experience changes, in fundamental ways, how we think and who we are. That means that a rational decision-maker must consider how experience will change them. This also means that, since we can to some extent select what kind of experiences to have, we have some control over how we change.
But though learning is at the heart of cognition and though we are exceptional learners, the basic architecture of cognition creates both challenges for and limitations on the decision-maker. The nature of connectionist learning accounts for many of the ways in which people tend to be irrational about beliefs and decisions. Though many problems in transformative and general decision-making caused by the architecture of our brains manifest themselves as irrationality, that isn’t always the case; in fact, the biggest challenge springing from a connectionist architecture has nothing to do with rationality. Yet it’s a problem that plagues us at every level of human society and in every social endeavor we attempt.
The difficulty of explaining our thinking to others is notorious. It’s hard to say what you think. Hard to capture thoughts and emotions and the interplay of the two in your mind. Harder still to communicate them intelligibly. And, of course, it is equally hard to take the words we hear and use them to imagine the state of another’s mind.
The failure of words to express thoughts and describe minds drives a deep sense of otherness that haunts both personal experience and human social norms. Language’s inability to describe thought lives at the very heart of human experience as an unrelenting sense of aloneness. As much as biology, it may drive our need for committed relationships. Perhaps it underlies the rare sweetness of experience that we extract from friendship, from successful teamwork, and from family. It lives, too, in our easy hatred of outsiders, our fear of isolation and in the lurking belief that lives deep within even our closest and most tested bonds that the other is unknowable.
The ability to capture mental states in words is such a rare and precious gift that it, more than fine style, makes a novelist truly great. Writing style is a veneer on the surface of a great novel – the depth and foundation of which is an accurate depiction of human thought. And if that accurate depiction is the preserve of rare genius, what does that tell us about how incredibly hard this seemingly mundane everyday task must be?
What makes the translation of thought into words so very hard is, of course, the way we think.
The state of a connection system cannot be translated into an algorithmic procedure or a string (even a very long string) of sentences. Connection systems were selected for in biological systems and created in digital applications precisely because long chains of sentences (even millions of lines) describing procedures can’t solve basic problems essential to embodied action. Complex connection networks are literally unexplainable in ordinary human language. The only explanation we can give for why a connection system produced the answer it did is to point to the entire network and say it ended up with these nodes having these weights based on how it started and the data it trained on.
We might as well shrug our shoulders and grunt.
But that’s the way connection systems are. Their initial state is infinitely tunable, and the current state of systems in our brains has been changed by countless experiences. The mechanism for how experience impacts state is complex and obscure. And the current state cannot be described using any type of short-hand notation or set of rules. Since a process-oriented translation of the results of a deep-learning system isn’t possible, you can’t verbalize it in a way that can make its operation clear to anyone else. There is no way (in words) to explain how it worked or why it gave a specific answer. In data science lingo, connection systems are described as “black box” systems. They take input and they produce output, but the process in-between cannot be documented or described.
And, of course, the brain is a much deeper black box than a computer neural network. In the computer lab it’s possible (at least in theory) to monitor a deep learning system, see its initial weights and watch, step-by-step, how it trains.[1] All this is completely inscrutable in the brain. We have inputs and we get outputs. And the process of transformation is beyond description. This is every bit as true for our own internal consideration as it is for our social communication.
It’s demonstrably hard to translate thought into structured explanations. And those explanations are NEVER a comprehensive description of the way the thinking works or a fully accurate representation of our cognitive state. Explanations for why we think something are almost universally unsatisfactory; slipshod attempts to backfill process and logic onto a system that works quite differently. There should be no expectation that anyone’s explanation of their own thought captures the true nature of the internal system. This isn’t a failure of honesty, it’s a product of the limitations of language and the opaque nature of connection systems.
Words fail us.
It’s an unfortunate fact that trying to explain why we think what we do is often harder and more work than the original thinking was. When athletes trot out the “staying focused on the game” or “we were just doing our job” tropes, we chalk it up to a surfeit of media training and a lack of advanced education. But having imbibed since youth at the fountain of sporting cliches, the explanations are self-assumed to describe what they are thinking. We ALL frequently settle on explanations for why we think what we do without putting much cognitive effort into the process.
This terrible difficulty in communicating thought is widely recognized. Educators know it takes mastery of a subject to explain it well. Almost all of us have struggled mightily to translate what’s crystalline in our mind into words having the same vigor and clarity. Ditto for every expert who has tried to explain why you cut wood at a certain point, how you knew a burger was cooked to medium rare, what it is about a particular shade that makes a room pop, and why guessing curveball on that count was successful. There is a reason that so much apprenticeship is observation, attempt, and criticism not the imparting of rules – and the reason is that, as Mendelssohn pointed out, words are terribly imprecise for most forms of thinking.
With modern cognitive models, none of this is surprising. Connection systems are impossible to describe verbally. They don’t reduce to a simple set of beliefs or axioms and the way they function isn’t describable in algorithmic terms. Given how we think, the wonder isn’t how hard it is for us to communicate thought, the real wonder is that we can do it at all!
In fact, our ability to do this at all is a tribute to the amazing power of language[2]. For all its shortcomings, no cognitive tool is more important or powerful. The rich constructs of language allow us to map an immense amount of the world of objects and actions in a simple, incredibly powerful, short-hand notation. This shorthand allows us to build skyscrapers of thought piece by piece in ways that could never be done in any other fashion. And for all its deficiencies in explaining thought, language allows us to share, shape and criticize thought across space and time to build structures that grow higher and higher.
Human cognition is unique in the extent to which it uses tools to save, refine, and share work. Humans are described as tool users and nowhere is this truer than in the exercise of thought. It is not in clubs or charcoal or the wheel that the human genius for tools is best exhibited. It is in the tools of cognition – language, numbers, writing – that our tool genius is most fully expressed.
Yet picture a world where we could routinely use language to describe our thinking fully and accurately. Imagine how much easier it would be to communicate how we feel. To learn what it feels like to be someone else. To create bonds of trust and sympathy between people. If a creature could have powerful thought and transparency of expression, life would be immeasurably different and probably much better. That is not our world and never will be.
Philosophers talk about mind-reading – a term used not in the parapsychological sense but in the very literal sense of understanding what someone else is thinking. Evolution has given us powerful skills for mind-reading. We are expert readers of facial expression and bodily behavior. We have, also, the ability to use ourselves as a kind of mirror to understand what might be passing in someone else’s mind.
Yet so much even of our own mind is inscrutable. Conscious thought is a small part of what goes on in our brains and the thoughts that bubble into consciousness are neither causally explicated nor fully described by the language we have at hand. So, we are limited in our ability to understand ourselves and, for all our face reading and behavioral mirroring, crippled in our ability to understand anyone else.
The impact of language’s inability to accurately convey thought is widespread and damning. So often we feel alone. Cut-off even from those who live closest to us – even those who undeniably love and cherish us. What teenager hasn’t woefully reflected on the painful inability of their parents to understand the least little thing about them? And what parent, staring into the troubled eyes of their son or daughter hasn’t felt hopelessly cast out and excluded from a person they value more than anything in all the world?
Parents, friends, lovers – even the people who know us best are painfully unable to understand our experience and thoughts and language gives us no way to enlighten them.
Beyond its toll in misery and frustration, the challenge of communication makes the transformational decisions that determine the shape of a life much, much harder. Other people – other minds – are the only plausible evidence we have for seeing how life decisions might work out. There is no non-personal, non-historical way to reason about how some experience might shape us or what kind of life it might entail. It’s natural that we would search for– and crave – truthful descriptions of who people are, what they experience, and how those experiences changed them. It’s the kind of truth we do not much get. Sometimes this lack of truth is a culture problem. But even in the best cultural imaginable, most people are incapable of producing language that can adequately describe any of those three things much less how they are related. It is the work of the novelist or actor to express these fundamental facts truthfully – and it is HARD. So hard that only a small percentage of the best work by the most talented of people seems to us to capture even a part of what it is like to live and only a factionally small percentage of that work gives us a sense of some way to live better.
Words fail us not because we are slow, stupid, dishonest, or uncommunicative. Words fail us because of what we are and how we think. And because words fail us, much of what we would like to know to make better transformational choice is forever outside our understanding.
[1] Of course, watching a modern computer connection system train step by step would take many human lifetimes since the processor is so much faster than our brain.
[2] Which is, of course, also learned using connection systems. The human brain has a dedicated area specifically reserved for learning a first language and it’s the availability of this perfectly available and ideally structured system that makes learning one’s native language so effortless.