China brain

Is it time to revise functionalism?

Computational functionalism defines conscious states by the causal role they play within a system: which inputs produce them, how they interact with other internal states, and how they contribute to behavioural outputs. These roles are described in terms of a program, analogous to that of a Turing machine. Functionalism¹ is an attractive theory of mind because it abstracts away from biology. By doing so, it allows for the possibility of non-biological conscious systems, such as AI systems and mind uploads.

What was originally attractive about functionalism quickly became one of the main arguments against it. In a previous essay, I discussed how Hilary Putnam came up with functionalism and later came to reject it, on the basis that it led to counterintuitive conclusions. Putnam was not alone in this. Ned Block, Putnam’s student at Harvard, presented a similar worry through a now famous thought experiment.

In this essay, I present Block’s thought experiment. I then explain when and why counterintuitive consequences matter for assessing a theory, and why we lack good positive reasons to accept functionalism’s conclusion. Finally, I offer some thoughts on how functionalism might overcome this objection.

The Rise and Fall of Computational Functionalism

§ Functionalism’s expansive view

Functionalism allows consciousness to be multiple realisable, meaning that conscious states could be instantiated by different kinds of physical systems. As a result, it is more liberal than views that identify conscious states with particular biological structures. In principle, biological beings with very different physiology from humans could have conscious states, provided they instantiate the right functional organisation.

The same reasoning extends beyond biological organisms. If a complex non-biological system could implement the same functional roles as a brain, functionalism suggests it too could possess conscious experiences.

§ The setup

Block (1978) demonstrates this consequence. In his thought experiment, Block imagines a nation whose population is roughly comparable in size to the number of neurons in an average human brain.² The entire nation of China is organised to simulate the workings of a single brain. Each person functions as one neuron, communicating by two-way radio with others in a pattern corresponding to neural communication. Together, these interactions replicate the functional organisation of a human brain. The China-brain then possesses all the elements of a functional description of mind: sensory inputs, behavioural outputs, and internal mental states causally connected to other mental states.³

If the nation of China could be made to function in this way, then, according to functionalism, it would have a unified conscious experience. The population of China could collectively be in pain, even if no individual member of the population experienced pain.

§ The argument

Functionalism accepts that the nation could be organised in such a way, and would therefore be conscious. For Block, however, it is hard to believe that a system made up of billions of individuals, each merely following local instructions, has a single, unified conscious experience. Sure, each person in the system would have their own separate mental life, but there seems to be no further subject who has the collective experience.

The argument can be summarised as follows:

1. According to functionalism, a system is conscious if it implements the right functional organisation.

2. China-brain implements the right organisation.

3. Therefore, according to functionalism, China-brain must be conscious. [From 1, 2]

4. It is counterintuitive to think China-brain is conscious, and we have no independent reasons to think that it is.

5. Therefore, we should assign high credence to the claim that China-brain is not conscious. [From 4]

6. Functionalism predicts that something is conscious when it likely isn’t. [From 3, 5]

7. Therefore, functionalism is implausible. [From 6]

§ Theory choice

The fact that a conclusion is counterintuitive does not by itself suffice to reject it. Many successful theories once seemed deeply counterintuitive. The ideas that the Earth moves around the sun and that time passes differently depending on the speed at which an observer is moving, both conflicted with ordinary intuition.

What made these theories convincing was not that they felt intuitive, but that there was strong positive arguments for them. In each case, the theory explained existing observations, unified phenomena under a single framework, or generated novel testable predictions. Heliocentrism helped explain the apparent retrograde motion of the planets more elegantly than geocentric models. Special relativity predicted effects such as time dilation, which were later confirmed by experiments involving highly precise clocks.

When a theory makes testable predictions, those predictions provide a way to choose among competing theories. If a theory predicts phenomena that are later observed, this gives us reason to increase our confidence in it. Such evidence can shift the balance between competing theories by showing that one has stronger empirical support.

On the other hand, when theories are empirically equivalent, or at least when they cannot currently be distinguished by observation or experiment, the choice relies on the comparative strength of the explanations they offer. We assess explanations by theoretical virtues such as parsimony, explanatory scope, coherence with other well-supported theories, and the extent of their counterintuitive commitments. A theory can be counterintuitive and still worth accepting, but in the absence of independent empirical support, its counterintuitive consequences count more heavily against it.

§ The lack of positive arguments

Besides being counterintuitive, functionalism does not clearly satisfy other theoretical virtues.

Explanatory power—It has a simple theoretical structure, but achieves this simplicity by greatly expanding the class of conscious systems. As a result, it currently lacks explanatory depth. Its criterion for consciousness is so general that it counts many systems as conscious without explaining why any of them have subjective experience rather than merely intelligent behaviour.

Inter-theoretic coherence—It fits well with the spirit of cognitive science, which tends to describe the mind in functional and computational terms. However, it sits uneasily with the apparent biological dependence of known conscious systems.

§ Lessons

The China brain thought experiment demonstrates that, in its current state, computational functionalism is implausible because it lacks many theoretical virtues we rely on when assessing theories. Block’s thought experiment therefore leaves functionalists with one main option: to restrict the functional conditions for consciousness so that China-brain isn’t conscious.

To do this, functionalists must move away from a purely computational view and towards one on which the relevant causal role must be realised in some more specific kind of mechanism. Such a view has been advocated by Piccinini (2020). On his account, the functional roles of mental states is tied to the specific physical mechanisms in the brain that carry out those roles. The emphasis is not on an abstract algorithmic pattern alone, but on the concrete mechanism that performs the relevant function.

This revision makes the view less liberal than computational functionalism, because it does not automatically imply that a nation passing radio messages implements the kind of mechanism required for consciousness. It is therefore not vulnerable to the China brain objection in the same way.

Bibliography

Block, N. (1978). Troubles with functionalism. Minnesota Studies in the Philosophy of Science 9:261-325.

Piccinini, G. (2020). Neurocognitive mechanisms: Explaining biological cognition. Oxford University Press.

1. Unless otherwise stated, I use ‘functionalism’ to refer to ‘computational functionalism’.

2. For the purpose of this thought experiment, assume that China’s population is comparable to the average number of neurons in a human brain.

3. Other features can be added to satisfy additional functionalist requirements. For example, the China-brain could also be connected by radio to a body that provides sensory inputs and receives behavioural outputs. However, these additions can be ignored for the purpose of the argument.