The case for removing concepts from cognitive science and AI research
It can be difficult to convince someone that concepts don’t exist. Everyday experience appears to provide overwhelming evidence to the contrary. Concepts are not only intuitively perceived to be active in daily life, they are also a widespread feature of theories across AI and cognitive science, where they are assumed to be necessary for symbolic and logical thought¹. Most who read the title of this post would be tempted to brush off the argument as patently, demonstrably absurd. It’s akin to trying to convince a European 500 years ago that God doesn’t exist, when everything around them appears to be evidence of, and indeed presupposes God’s existence. Any contrary argument is likely to be taken as the result of sophistry or word-wrangling, or because some critical piece has been neglected.
Despite their seeming obviousness, it is worth noting that there is still no complete and unambiguous explanation for what concepts are, or how they work on thoughts —and indeed how to program them into AI. The human ability to learn and create concepts is multifaceted and complex. AI theories and implementations generally only touch on one or two of its features, while neglecting large numbers of counter-cases. This has lead some researchers, notably Lawrence Barsalou, to suspect that the way we think of concepts is flawed. Perhaps the whole notion of concepts — as a native mechanism for grouping experiences — is untenable.
This article is well-worth reading. Below, I have made a few notes based on my reading of it. To my eye, it demonstrates the existence of consciousness as a thing, the existence of a very real subjective world, the high probability that this subjective world is not entirely confined in your head, that consciousness is a primary of existence and not confined to our brains, and also, importantly for this website, why FIML works so well.
(The sections in quotes are from the article.)
Firstly, concepts: they exist within consciousness and are used to reason, analyze, communicate, organize, and so on. They are probably a features of consciousness itself, depending on how you define them. They need not be stable.
Secondly, FIML:
To begin with, there is no scientific experiment or empirical observation that can be used to prove that any given concept “exists”, and by extension that concepts exist at all.
No. FIML practice provides unlimited empirical observations that concepts exist. FIML is a scientific experiment and can easily be repeated as many times as you like.
To objectively prove that any given interpretation matches reality, you would somehow have to compare your subjective mental concepts against an objective view of the real situation. But the latter isn’t possible.
Yes, it is possible. FIML is precisely that—a means ‘to compare your subjective mental concepts against an objective view of the real situation’.
FIML accomplishes this by allowing two subjective consciousnesses to objectively compare their mutually ‘subjective mental concepts’ against each other. To claim that ‘an objective view of the real situation’ can only be achieved by some other means is absurd. The very best means to objectively compare subjective states is to have two honest informants compare them based on a shared micro unit of communication in the real-world in real-time. This is what FIML does.
The discreteness of concepts is a built-in requirement of language itself, one that does not necessarily reflect what an individual mind is doing.
This is very true but shows only that the generalized use of concepts is both rigid and murky. Individual use of concepts with full access to individual real-world, real-time subjective data, shows concept are part of consciousness itself. This consciousness is deeply entwined with language, but if micro-analyzed with FIML it can be seen to be not nothing but a very real something that fully exists in consciousness and can be discovered and analyzed with language following its identification in real-time real-world (RTRW) situations, when concepts are psychologically of greatest importance.
The author has shown how and why general semiotics/ concepts lead to psychological agony when they are all we have, all we are able to think and experience with. As soon as we learn how to isolate concepts/ semiotics in the RTRW moment through FIML, we can see how deeply real they are and how deeply significant they are to psychological, spiritual, existential health. The micro data discovered through FIML and objectively compared by two honest informants reveals what we humans are as nothing else can.
I agree with the part on introspection but do not find it exhaustive. FIML solves the problem that “the act of introspection itself require[s] that I connect the specific, unique [semiotic] image of the house to available, common identifiers’. FIML solves this problem by having partners compare their ‘specific unique’ subjective states/ semiotics based on their mutually agreed objective data. This comparison often reveals differences of interpretation, mistaken interpretations, contradictory interpretations but all of these can be resolved due to the mutually agreed upon objectivity of their data. This does show that language is often vague and sloppy but it also shows that the subjective consciousnesses of partners can be mutually understood and cleared up. At the end of a successful FIML exchange, both partners will agree on what happened and be well-satisfied with their mutually shared resolution. This can be proved by repeat FIML exchanges, which will clearly show both partners that their their understanding of language and their inner psychological states have greatly increased.
Since this essay pertains to computer programming, I can easily see why an over strong belief in rigid concepts is a dead end. But it is conceivable (OMG!) computers will be able to do something like FIML between themselves. As for humans, these problems are very real but can be solved. At the same time, these problems and their solutions all but prove that our consciousness is its ‘own state’, a primary feature of existence, maybe something like a quantum wave which is or partakes of a reality that subsumes much more than just humanness.
I agree with most of the conclusion:
And just like all other concepts, the need to structure my introspective activities in terms of concepts may disappear at any time if it becomes useless. It is always possible for me to think about my own experiences without believing that I have concepts. I can consider the usefulness of each specific thought connection, with utility being the mechanism pulling thoughts together and apart; all while remaining oblivious of concepts. Concepts are only a passing, convenient invention of introspection, adequately serving the purposes of self-understanding.
That statement is very Buddhist and describes the recognition that mental constructs are empty, impermanent, and not to be clung to. FIML shows us this as well. Both Buddhism and FIML also show us that there is an underlying or overarching reality of consciousness larger than us, just as our eyes show us that light is larger than us.
Just as cognitive science long ago eschewed consciousness as the driving force of all mental activity, recognizing the illusory nature of concepts is the next step in the development of the field.
This is stated too starkly. Consciousness need not be the driving force in mental activity.
Someone may claim my argument on consciousness is tautological. I can agree with that in the sense that the argument against what I say is also tautological. The key point here is when we make claims about human consciousness and human reality, we have to base them on something. At the end of the day, this is our lived experience. My lived experience is rich with spiritual material. Someone else’s may be rich with other kinds of material that eschew spirituality. Their view contains and describes me as my view contains and describes them.
I hope the author of this article will try FIML and from that experience look on consciousness differently. ABN
American Buddhist Net 
Hi, as the author of said post I appreciate your breakdown and analysis. I should say the parallels to Buddhism are near the mark, and I find that branch of philosophy to be ahead of its time, especially with respect to our understanding of existence and non-existence. Indeed modern cognitive science can be said to be starkly anti-Buddhist, dogmatically focusing on clear definitions and conceptual units, which is one of its blind spots.
As for FIML and subjective consciousness, I recognize that two people can achieve alignment between their concepts through communication, however, they are still both doing it from inside their own mind. This is the point – although the understanding feels like it has transcended the limits of its subjectivity, it is actually a perception of intersubjectivity which is still confined within itself. Even its understanding and belief in that alignment can easily be an illusion – what if you think you’ve reached agreement, but are mistaken? I may be in a dream and think I’ve reached an agreement with a person in my dreams too.
This is an unpleasant realization that many would want to avoid, but it is nonetheless undeniable from a pragmatic perspective. An analogy might be to ‘Google Earth’ – it looks like it contains the physical “world” inside it, but it is still a software application, and will never be anything but software. Its connection to the greater world is in the form (shape) of that software (how the bytes are arranged).
One last note, in the quote where I said: “Just as cognitive science long ago eschewed consciousness as the driving force of all mental activity”, the word “eschewed” means discarded, as in very few psychologists believe consciousness is the driving force anymore. I certainly don’t.
mycardboarddreams:
Thanks for your comment. I enjoyed your essay and commented on it because it is stimulating and interesting. A few more points below:
FIML is ‘objective” in the sense that the micro communicative unit observed by one partner is agreed upon by the other partner. I fully recognize this level of ‘objectivity’ is not perfect but it is orders of magnitude better than what people normally always use in interpersonal exchanges. When FIML is done dozens, hundreds of times, partners reach a level of linguistic/ semiotic understanding that is orders of magnitude better than what people normally reach. This mutual understanding is like a microscope that allows frequent observation of very fine detail. It’s a beautiful thing to do. I know of no other way to get there.
A lot of your argument is rather delightfully Buddhist, which is a major reason I commented on it. I did say, ‘Consciousness need not be the driving force in mental activity’. Buddhism recognizes delusive consciousness, wise consciousness, wholesome and unwholesome consciousness, deluded consciousness. Buddhist practice encourages wholesome consciousness through right views, thoughts, etc. In Buddhism, consciousness is a sense. Trained consciousness, like trained vision or hearing, works differently and better (if well trained) than untrained consciousness. Buddhist training trains consciousness, awareness, how we think, what we say, how we reason, etc. Is that a ‘driving force’? Let’s agree that training can happen and as such is a big deal.
Your sense of what concepts are, far as I can tell, is exactly what Buddhists mean by ’emptiness’. In Buddhism, concepts are empty. Buddhism goes further by claiming that everything in the relative human realm is empty.
My sense is that you are leaning too hard on the idea of concepts. What you say may apply in computer science, but is not a strong point philosophically except as a subgroup of empty things in Buddhist philosophy. In computer science, maybe concepts could be construed as bundles or connected aggregates (Buddhist term), which though empty may serve a philosophical ‘lesser purpose’ of allowing some kinds of calculations that bring good results.
At some level, there is here and elsewhere an ongoing difference of viewpoints—consciousness vs mechanistic or robotic, or whatever words you like. I think both sides are ultimately ‘tautological’ within the confines of human linguistics. My second to last paragraph in my original post goes into this.