Zombies

This article contrasts conceptual zombies with those of Haitian folklore and outlines the importance of the zombie question to cognitive science and efforts to create artificial cognitive systems.

Two Kinds of Zombies

Zombies from the literature of cognitive science and philosophy of mind, unlike the voodoo victims of Haitian folklore, are traditionally understood to be hypothetical creatures altogether lacking conscious experience who nonetheless behave indistinguishably from the rest of us. Philosophers’ zombies walk and talk as if they’re conscious, they appear to wake up in the morning, and over breakfast they even speculate on the meaning of dreams they claim to have had. They don’t realise they’re zombies, of course — no feeling of peculiarity spoils the pristine emptiness of their barren phenomenological landscapes — and an enterprising theorist engaging one in conversation about the topic might well hear all manner of insightful commentary about what the concept of zombies reveals about minds.

The traditional picture of behavioural zombies does not require that such creatures also be physically indistinguishable from their conscious counterparts. Confusingly, some theorists use the word ‘zombie’ to refer to this stronger variety, what I will call a ‘super-zombie’ or ‘absent qualia zombie’: a non-conscious creature indistinguishable both in terms of behaviour and in terms of physical construction.

Two Kinds of Possibility

It is straightforward to demonstrate that a behavioural zombie is logically possible. I have done this, for instance, in Mind Out of Matter, but the construction given there does not establish that zombies are physically possible. (Of course, that does not mean they are not; it simply means it has not been demonstrated.) Whether or not super-zombies are possible, however, depends greatly on the details of one’s concept of consciousness. While many philosophers routinely claim that super-zombies are logically possible simply because they themselves cannot think of any contradiction in the concept, this should not be taken as an especially convincing line of argument.

Zombies lie at the heart of the four-way relationship which obtains between a creature’s environment, its behavioural responses to the environment, the internal mechanisms which support that behaviour, and the creature’s phenomenal consciousness. One way to clarify the issues is to consider the infinite range of mathematical descriptions with which we might try to capture the first two items in this relationship.

Consider any finite set of measurements of a conscious creature’s environment and its responses to that environment. Infinitely many mathematically distinct descriptions can correctly match the behavioural responses with the environmental conditions. If instantiated by real systems, each of these would describe systems with seemingly conscious behaviour (and perhaps genuinely conscious behaviour, depending on the answer to the zombie questions). Some of those infinitely many distinct descriptions, however, will be straightforwardly incompatible with known physics. For instance, some theoretical descriptions might yield correct behaviour for some specified environmental conditions simply by retrieving them from an enormous lookup table so large that no cognizer actually instantiated in the physical universe could include such a table. Other descriptions, however, will be consistent with the laws of physics. These are the descriptions of physically possible systems for yielding the right behaviour under the specified environmental conditions. Finally, a still smaller set will be consistent both with the laws of physics and with the actual facts about the real world; these are the descriptions of the actual systems which really do yield the behaviour. Then the questions from above about the possibility of zombies become:

  • Are there any descriptions at all which do not also describe creatures which are conscious? (=”are zombies logically possible?”)
  • Are there any descriptions consistent with the laws of physics which do not also describe creatures which are conscious? (=”are zombies physically possible?”)
  • Are there any descriptions consistent both with the laws of physics and with the physical facts which do not also describe creatures which are conscious? (=”are super-zombies possible?”)

Possibilities and Mathematical Mysticism

In 1950, Alan Turing famously described the ‘imitation game’ — more commonly known these days as the ‘Turing Test’. The ‘test’ amounts to a question and answer session for separating the thinkers from the pretenders. Conceal a computer in one room and a human in a second, and if the computer regularly fools a human interlocutor into believing it is a real thinking creature, the computer wins. More often than not, philosophers now refer to consciousness rather than just ‘thinking’, and the question becomes whether passing the Turing Test guarantees consciousness.

Anyone prepared to assert the logical possibility of zombies should of course reject the imitation game as a conceptually adequate test for consciousness. Zombies obviously can pass the Turing Test — they are, by definition, behaviourally indistinguishable from the rest of us. The believer in zombies shouldn’t be satisfied just with the instantiation of a relationship between environmental stimuli and observable behaviour which is good enough to pass the Turing Test.

The opposite view, which I have labelled ‘mathematical mysticism’ in Mind Out of Matter, comes in two flavours, strong and weak. Strong mathematical mysticism is the unargued assertion that out of all the infinitely many mathematical descriptions of seemingly conscious relationships between environment and behaviour, all describe creatures which, if implemented, would be conscious. (I.e., zombies are logically impossible.) Weak mathematical mysticism is the unargued assertion that out of all the mathematical descriptions of seemingly conscious relationships between environment and behaviour which are also consistent with the laws of physics, all describe creatures which, if implemented, would be conscious. (I.e., zombies are physically impossible.)

The crucial observation is that, at present, empirically based arguments for either of the above views are very thin on the ground: that is why I have labelled them ‘mysticism’, because they express a brute article of faith and little more. Dan Dennett is perhaps the most famous example of a mathematical mystic: he suggests (Dennett 1995b) that there just isn’t any other way to display apparently conscious behaviour, of the sort sufficient for passing the Turing Test, except by actually being conscious.

But I believe the real question should be: which of the infinitely many mathematical descriptions of seemingly conscious relationships between environment and behaviour are also descriptions of truly conscious creatures? What sorts of descriptions are the right ones — and not just mathematically satisfactory input/output look-alikes — for creatures which really are conscious? The corollary: which of those functions which describe physically possible renditions of the right input/output relationships also describe truly conscious creatures? Put differently, how do we flesh out the remaining parts of the four-way relationship described above, that between a creature’s environment, its behavioural responses to the environment, the internal mechanisms which support that behaviour, and the creature’s phenomenal consciousness?

It would appear that behaviour specified at a gross level is insufficient: what matters is how a system achieves its relationship with its environment. In other words, it is not enough to say that a system behaves externally as if it is conscious, therefore it is conscious; the claim of consciousness must be grounded in some picture of the internal structures and processes which give rise to that behaviour. Moreover, actually grounding claims of consciousness requires a theory of consciousness — an explanation of why it is that particular internal structures or processes should be conscious. (In other words, it requires a defense against the strong claim in favour of super-zombies, that such internal structures or processes might still obtain even in the absence of consciousness.)

The upshot for research into real robots and other systems which might one day display seemingly conscious behaviour (or even enjoy real phenomenal consciousness) is that one’s view of whether consciousness itself is essential, superfluous, or useful but dispensable for the performance of particular behaviours may constrain the class of architectures which one attempts to implement. As outlined briefly in the short introductory note on qualia, the easiest answer (‘consciousness is superfluous’) forces certain other pressing questions, such as why, if it is superfluous, consciousness should have evolved in the first place.

This article was originally published by on and was last reviewed or updated by Dr Greg Mulhauser on .

Mulhauser Consulting, Ltd.: Registered in England, no. 4455464. Mulhauser Consulting does not provide investment advice. No warranty or representation, either expressed or implied, is given with respect to the accuracy, completeness, or suitability for purpose of any view or statement expressed on this site.

Copyright © 1999-2020. All Rights Reserved.