How to create a robot that has subjective experiences, part 2

Phenomenal experiences are shaped and determined by your unconscious motives

From Narrow To General AI
10 min readJun 30, 2024

This is the continuation of a post on how to mechanize the experience of subjectivity. I recommend starting with part one, as this post builds on those foundations. Whereas the prior post provided a high-level argument for its feasibility, this post addresses more detailed features and functionality. Part 3 is here.

We know about consciousness more directly than we know about anything else, so “proof” is inappropriate.

It seems to me that we are surer of the existence of conscious experience than we are of anything else in the world. — Chalmers, The Conscious Mind

There are few things people have as complete faith in as their own beliefs about their consciousness. They defend it with passion, based on the unquestioned evidence of their own inner perception. Though they may doubt the inputs of their external senses — the eyes and ears can be fooled after all — they do not doubt the evidence of the inner eye that sees their own experiences. Chalmers, in his book The Conscious Mind, points above all to this private source of evidence, and finds it axiomatic both that consciousness exists, and that a person’s judgments about it should be taken as sacrosanct. Not that he had any other choice; there are no alternate sources of evidence.

Daniel Dennett, on the other hand, doubts the reliability of introspection:

What we are fooling ourselves about is the idea that the activity of “introspection” is ever a matter of just “looking and seeing.” I suspect that when we claim to be just using our powers of inner observation, we are always actually engaging in a sort of impromptu theorizing.

You are not authoritative about what is happening in you, but only about what seems to be happening in you. — Daniel Dennett, Consciousness Explained

Chalmers assumes that as soon as you are conscious you are ipso facto aware that you are conscious, and grasp its many varieties and qualities. In fact it takes several years before you to even start to see your own consciousness. Until then you are interacting with and understanding the sense-world you see through conscious experience; just as a toddler sees and understands the world through her eyes, but does not yet realize she sees through her eyes. It takes sophisticated acts of introspection and reasoning to grasp that you are conscious, and what it feels like to be so. The resulting thoughts and beliefs are the entirety of that awareness. Everything you know about consciousness, even knowledge of your own basic existence must actively be driven and acquired through judgments about it.

It is also possible that your knowledge of the content of your consciousness is frequently in error. The only way you know what it is “like” to see red, for example, or if it is like seeing blue, or like feeling angry, is by observing and making that judgment. If this either step were faulty — and we showed that it can be — you may arrive at an incorrect conclusion, and the nature of the phenomena you experience would drastically change.

In general, discussions of consciousness rarely critically analyze the machinery of introspection and judgment. The goal of this post is to dig into that very topic, and ask: what is the formal nature of judgments and beliefs about qualia, and how does the machinery arrive at its conclusions?

Consciousness and determinism

If we take both Dennett’s and Chalmers’ arguments at face value, there is every reason to believe that experiences and interpretations of conscious phenomena are roughly deterministic. At the lowest levels, red inputs remain red inputs, red inputs are never blue inputs, and the machinery of these inputs cannot be changed by choice. Beyond this layer, you are compelled to think about and interpret the phenomena of colour in specific ways — not because they are “facts” you can learn about, but because the processes by which you interpret your experiences are constrained. Most of the processing is happening beyond your ken, prior to you realizing it, and under the hood, so to speak. You usually can’t see the causes of your thoughts and judgments — but that doesn’t mean there aren’t any.

We rarely think of ourselves as being “forced” to experience phenomena or consciousness, but this almost certainly the case, regardless of whether you approach the topic as a reductivist or a non-reductivist. You are forced by your motives to draw conclusions about your conscious experiences. You cannot help but interpret red as having red qualia, for the same reason you cannot help seeing a colour as exactly that colour. There is no impartial external arbiter transcending all brain functions, which would spot if it were being directed by underlying brain functions, or which could do something to prevent it. Every such arbiter would itself be part of that deterministic system, subject to its constraints. So, in principle at least, the machinery of phenomenal experience could be replicated in a robot, because judgments about phenomenal consciousness can be driven by unconscious machinery.

The hidden causes of qualia judgments

A good case study of these underlying drivers of conscious experience is of the phenomena associated with pain. It would be fair to say the qualia of pain have an “unpleasant” flavour — regardless of what other interpretations you may associate with it. It is tempting to infer, as many have, that pain has an unpleasant feel because, evolutionarily, this was useful for getting the mind to avoid whatever damaging events caused it; in other words, the unpleasantness of the qualia is what gets you to avoid painful events:

People in pain try to get rid of it or to diminish it. Why? The answer surely is because pain feels unpleasant or bad, because it is experienced as such. — Tye, Another look at representationalism about pain

The common understanding of pain as an unpleasant qualia.

However, to say that pain has an “unpleasant feel” begs the question. What does it mean to interpret something as unpleasant? Who is doing the interpreting? The above argument creates an infinite regress since it requires a second viewer inside the “Cartesian Theatre” of the brain who determines that the input was unpleasant:

Any account of pain that left in the awfulness would be circular […] Similarly, a proper account of laughter must leave out the presumed intrinsic hilarity, the zest, the funniness, because their presence would merely postpone the attempt to answer the question. […] We can break the regress only by discovering some viewer whose perception avoids creating yet another picture in need of a viewer. — Dennett, Consciousness Explained

So how does your mind know that pain is “unpleasant”? For starters, any knowledge you have about the unpleasantness of pain must be in the form of an explicit thought, one which you could learn and later recall. Pain itself is invisible to consciousness. For you to know that pain is unpleasant, there must be some process creating such explicit knowledge in the form of a thought — e.g. “this feels unpleasant”, or “ouch”, or some related abstract imagery — then attaching the thought to the current circumstances. Pain, then, is the underlying force driving the pattern of thoughts¹ about itself this way.

We’ve discussed in another post how your mind attaches such thoughts because they are useful “goals” to bring to mind. Even if they are never acted out, the thoughts you have like plans, or explanations, or intents. They represent the cause of a solution — e.g. you think “ouch” or “that’s painful!” because were you to produce those words out loud (cause) others might help relieve it, or at least not to add more of it (effect). Observe your own thoughts when you are alone and stub your toe. They will likely be of expressing your displeasure — the kind of thing you would actually do, and would be socially appropriate for you to do when others are around.

Pain, as a motive, may also attach other thoughts of getting away from the experience, or even how to compel other people to aid you. In many cases, comparing a present pain to a memory of a prior pain may help clarify its intensity and importance to others. All your attempts to describe your painful experiences are couched in the context of escaping it.

The driving force of learning interpretations (thoughts) about an experience is its underlying tension. In this case it is physical, but in most cases it is an internalized tension — e.g. a dangerous thing, an uncertain setup, an indicator of a possible problem.

Pain, as a motive, shapes your cognition by enforcing what kinds of thoughts you have about it, and what you learn. It builds “knowledge” about pain only insofar as those thoughts are functionally useful. The resulting thoughts and judgments are, of course, only symptoms of pain, not pain itself. But then so is your knowledge of the qualia, or phenomenal aspect of pain. Your knowledge about every qualitative aspect of your experience is part of the panoply of such pragmatic interpretations; as such it is coloured by the underlying motivation that created it.

There is no objective phenomenal “thing”, either inside or outside your mind, that you can empirically observe, and from which you may learn about qualia. Knowledge of the qualia of pain is a fabrication your own mind creates out of its needs, as it latches onto any nearby stimuli (found in thought or in the senses) that might be useful — a helpful sound, a comforting image, even an objective description to communicate your distress.

Desire seems to exert a phenomenological “tug,” and memory often has a qualitative component, as with the experience of nostalgia or regret. — Chalmers

This is what makes phenomena seem real despite being subjective. The qualia of pain, in this sense, is only one part of a gradient of useful interpretations you think up and attach to your experiences. It is driven by its own special motives, both built-in and derived ones, whose specifics must be learned — they are not the default mode you are born with.

The cue for awareness

As should be clear from the above, your unconscious motives play a critical role in creating conscious experiences. You only see and think what is useful for you to see and think. Discovering the phenomenal aspect of a stimulus does not begin with the appearance of that stimulus, it begins before that, with the underlying need to understand and interpret the stimulus in a certain way.

You don’t always realise that experiences have qualia. You must first be cued to attend to them, which means you are only motivated to do so at certain moments, and for specific reasons. For example, whenever I mention subjective experiences like colour phenomena, many people start to involuntarily attend to the colours in their vicinity, which they weren’t doing a moment before. That shift in attention is what creates all your knowledge of the phenomena of colour. As the unseen, underlying motives search for what they need, they create your stream of conscious thoughts out of what they find.

A rough illustration of how the underlying motive of pain triggers the mind to record its resolution — the removal of your hand. Henceforth that thought will occur each time you see a hot stove.

Specifically, a tension is triggered, after which the mind searches for and attaches a useful interpretation to surrounding stimuli. At first what it finds is haphazard, since the mind can’t know beforehand which of those stimuli will end up being helpful. For example, pain often motivates awareness of nearby incidental bodily sensations, insofar as they could help resolve (i.e. remove) the underlying concern.

This is why it is misleading to say that the unpleasant qualia of pain “cause” you to try to avoid it. By the time you are aware of the unpleasant qualia of pain, the underlying, invisible mechanism of pain has already had its effect on your mind. It is what drove your awareness and created your interpretation of its unpleasantness in the first place. Your initial awareness of any phenomena can not be forestalled by choice, since by definition they precede your knowledge of them. They must first deliver the phenomenal message to you consciousness. Even if you then try to resist or ignore the pain, such acts constitute additional, separately motivated thoughts on top of the initial awareness.

The division between conscious and unconscious layers is only illustrative. There is no physical separation between them. Rather, anything you declaratively learn about or express is defined as “conscious”.

Phenomenal judgments about pain aren’t free or random; nor should they be. What would we think of a brain that allowed itself to believe that a painful experience had only a fine and pleasant quality to it? Surely that it was fundamentally malfunctioning. You have little choice about what you think of painful experiences; even if you later try to lie to others about it. This necessity, the one that drives how you think about and judge pain, is what creates the deterministic nature of phenomenal experience. Judgments about phenomena are the deterministic, inevitable machinery of your brain making contact with your introspective observations. To replicate the experience of pain in a robot, its underlying motives must drive it to have those thoughts, as they are the means to its aims.

The experience of pain may perhaps be an exception in the world of phenomena. Since pain acts as a stimulus and a motive, it carries both necessary parts of awareness with it. How could we apply this same model to the subjective experience of colour, which does not seem to be underpinned by such motivations, and thus has no apparent trigger for you to be aware of it? Why would a mind attend to the phenomenal experience of colour, and why would certain thoughts and judgments be the result? Moreover, why do colour qualia seem more immediate, more concrete than those of pain?

In the next part of this series we will apply the principles above to explain colour phenomena.

¹ As well as learning and actions.

--

--

From Narrow To General AI

The road from Narrow AI to AGI presents both technical and philosophical challenges. This blog explores novel approaches and addresses longstanding questions.