User Tools

Site Tools


The User Illusion

Cutting Consciousness Down to Size
Tor Norretranders

failed to fetch data: unkown error

This is a quirky but substantial book by a science writer. Considerable parts of the book are devoted to introduction of thermodynamics (Maxwell's Demon), information theory, Godel's incompleteness theorem and Turing's undecidability theorem. Although all very interesting, and a seemingly gentle introduction to these deep matters, I'm not convinced that these have a great role in explaining the human condition. More important are various musings on the nature of consciousness, especially on the vast size and importance of the Unconscious and the nature of human communication. He also introduces the idea of the The User Interface Analogy, though does not follow through as much as he might have.

Thermodynamics and Information Theory

He uses thermodynamics and information theory to muster support for these observations:

Intelligence Discards Information

1 out of 1,000,000 bits sensed makes it to consciousness, showing the depth of unconscious processing. This is an empirical observation that doesn't depend on information theory. Intelligence adds value by discarding information, by winnowing that 1 bit from a million. Although this understanding is a natural consequence of an information-processing view of perception, he discusses how information theory has evolved concepts such as computational depth which more closely correspond to the conventional meaning of information as codification of useful knowledge, and not just a bit stream (the more random the better.) He correctly points out that the filter model of this information reduction as initially understood by cognitive psychologists is not right, because there is compelling evidence from studies of priming and subliminal perception that our sense data is processed and understood to a much greater degree than we are consciously aware. For example, in conversation, we are unconsciously sending and receiving nonverbal communication.

The Tree of Talking

He develops the concept of the “tree of talking” where the sense data is cooked down by extracting meaning from it, then the conscious awareness of this meaning can be verbally communicated to another, which can then be elaborated into imagined sense images. He calls the information that is discarded in the cooking-down process exformation. This information is more or less restored by the listener, allowing a connection between the talker's concepts and the listener's concepts. Clearly exformation has a lot to do with the concept of “knowledge”. We apply our knowledge to glean the useful information from the sense impressions, and also use our knowledge to reconstruct the meaning of language.

Determinism and Free Will

He uses Godel and Turing to address the issue of determinism and free will. I agree with the general approach, which is that the biggest problem people have with the idea of human behavior being a consequence of causal physical processes is that it means that human behavior is “mechanical” and could be predicted by a sufficiently sophisticated observer. Physical Chaos alone is enough to insure that behavior can't be predicted in practice, and is therefore not “mechanical”. He does give a nod to this idea, but then tries to use Godel and Turing to show that prediction is theoretically impossible.

Godel Incompleteness

I don't feel that Godel has much to offer, since clearly the human mind is both inconsistent and incomplete, so knowing it must be one or the other adds nothing. Godel himself, and some who invoke him, felt that the very fact we can prove incompleteness shows that the human mind somehow surpasses the the power of a formal system. This seems wrong to me, and I suspect it is a form of Level Confusion. Though the author doesn't strongly endorse this view, I was unable to detect any other reason for introducing Godel.

Turing Undecidability

Turing's results on computational irreducibility show that even for an entirely deterministic process such as a computer, if the program is sufficiently complex, there is no way to tell what it is going to do other than to run it. This certainly demonstrates an aspect of the fundamental unpredictability of complex processes, but I think human unpredictability is more fundamental than that. If the human mind was really a deterministic Turing machine with well-defined I/O streams to a deterministic environment that could itself be simulated, then it would be in principle possible to simulate a human much faster than real time, and so predict their behavior. It may have taken Borges' Funes the Memorious a day to recall a day in perfect detail, but someone with better hardware could theoretically do it much faster.

I and Me

He has some interesting discussion of the distinction between and conflict between the conscious and unconscious self, which he calls your “I” and “me”. Perhaps because it is more neutral, I found this terminology preferable to the “rider and elephant” in The Happiness Hypothesis. He is in some ways more concrete because he spends considerable time discussing specifics of how a soccer player or musician must unleash their “me”, and generally only create error if they consciously think about what they are doing. He observes that this state of performing without conscious inhibition is pleasurable. Though he doesn't use the term, this is Flow, though flow can involve greater degrees of conscious thought, depending on the task (the elephant and rider working together.)

Me is God?

He rather unconvincingly advocates the “bicameral mind” theory that before around 1000 BC the human mind worked fundamentally differently, with no introspective self-awareness as we now understand it. Instead, the output of our judgment processes were heard as the word of god via auditory hallucinations. I don't think that this is literally true, but it is certainly the case that in a culture where god is expected to speak to people regularly, it is easy to see that stuff bubbling up from the unconscious mind will be interpreted as divine inspiration. Or to put the causation in the right direction, the tendency of the unconscious to work in mysterious ways created the need for gods as an explanation.

Libet consciousness experiments

He devotes a considerable amount of space to explaining the Benjamin Libet experiments on time and consciousness. Although I agree with Daniel Dennett in Freedom Evolves that Libet may be reading too much into the results by neglecting to consider all of the possibilities related to differing processing delays for different paths, these results did seem rather incredible at to me when I first ran into them, and are important to pointing out the degree to which the brain uses complex means to attain a subjectively simple desired result.

The User Illusion

He does eventually get around the fascinating idea in the title which I think is the most up-to-date metaphor for the nature of consciousness. See this analysis section.

Modern Life Information Poor?

He makes the suggestion that the problem with screen-and-text oriented communication and rectilinear urban landscape in modern life is not that there is “information overload”, but rather that it is information poor, at least in the sensory modes that have the highest bandwidth and best connection to unconscious processing. He suggests that the draw of being in nature is that we are presented vivid detail (such as the fractal patterns in a tree), and this is actually a great deal of information, but information that humans are “designed” to handle, and so we perceive it as gratifying. In contrast, artificial environments tend to have simple structures that are highly predictable.

Of course, this doesn't mean that “information overload” is wrong. Instead he mainly identifying another problem, though it does also suggest that the general trend in computer use toward richer sensory output with video and realistic 3D graphics may in some way help with information overload by making better use of our high-bandwidth senses and unconscious processing (such as 3D visualization of abstract data properties.)

books/user_illusion.txt · Last modified: 2015/04/12 15:50 by ram