(Target audience: Empiricists/Rationalists rejecting conceivability arguments (such as the zombie argument), no background in philosophy necessary. Goes into some philosophy of language stuff, but doesn't get technical.)

The zombie argument against physicalism about consciousness is scorned by empiricists. And – on the face of it – for good reasons: How could a priori reasoning show us anything substantive about the world? How could our cognition have evolved to reliably assess such arguments? And its premisses are assessed using intuitions – but intuitions seem epistemically dubious at best.

I think empiricists are wrong in rejecting conceivability arguments such as the zombie argument. This is my attempt to convince them that conceivability arguments 1) are compatible with and indeed necessary for empiricists philosophy, 2) don’t ever show us anything substantive about the world, and 3) are supported by evolutionary psychology and linguistics.

I will first look at a (hopefully uncontroversial) case study and then argue for 1) - 3).

The case study: Empiricists like to compare consciousness to the phenomenon of life, where initially there was reluctance to reduce the phenomenon to physical processes but eventually views shifted and now it’s the received view that we don’t have to posit a fundamental life force to explain the various observables associated with life. How did this shift happen? 2 things were required: First, scientific progress had to show how microscopic phenomena could account for a variety of macroscopic phenomena in the domain of biology. Second, we needed to realize that these macroscopic phenomena were related to life in the right way. Roughly: Microphysics explains *something* and the thing it explains must fall under the concept we’re interested in, namely “life”. The first step has to do with the world, the second with language.

How does the second step work? We never had a precise definition of “life”, and we still don’t. How then can we know whether the various macroscopic biological phenomena we’ve reductively explained are collectively equivalent to life? It turns out that although people are notoriously bad at defining their terms, they are very good at pointing out the extension of terms in various actual and hypothetical situations. This should not come as a surprise, since pointing at things and classifying them is roughly how children learn language. (Carnap 1955) has realized that philosophers may use this phenomenon to capture the meaning of terms even if they can’t find definitions: we can construct a function from situations to extension by confronting people with lots of descriptions of situations (in the form of texts, images, or videos) and asking whether a concept can be applied to a thing/property in that situation. These functions, which are basically lookup tables of situation/extension pairs, are called “intensions” by philosophers.

Suppose we had done lots of research and had the complete intension of “house” and “building” at our disposal. (This is not practically feasible since we would have to have a function from ALL possible situations to extensions!) Suppose further, we had to find out whether “if something is a house, it is also a building” is true. How could we figure this out? Interestingly, we don’t need to do any additional empirical work for this. We can simply check our intensions. If every situation in which something is a house is also a situation in which that thing is a building, then “if something is a house, it is also a building” is necessarily or always or automatically true!

How can this be useful to solve the second step in reductively explaining life without having a proper definition of “life”? We simply replace “building” and “house” and ask a structurally similar question with the concepts we’re interested in: “If something instantiates [put here a list of various macroscopic biological phenomena explained by microphysics], is it alive?” We could proceed empirically and compile the (incomplete) intension of all the concepts involved to solve the problem. But we can also just ask *ourselves* to figure out our *own* (perhaps incomplete, perhaps idiosyncratic) intensions associated with concepts and then answer the question based those. And here comes the punchline: This is ALL there is to conceivability arguments! One way of asking if all Xs are also Ys, is by trying to find an X that isn’t a Y. One way of finding out if all houses are buildings, is by trying to find (=conceive of) a possible house that isn’t a building.

Conceivability arguments exploit the fact that although the meaning of concepts isn’t consciously available to us in the form of definitions, we do have access to their (perhaps incomplete or idiosyncratic) intensions. Applied to the “life” case: We may not be able to define “life”, but we know that everything that has [complicated & possibly disjunctive list of biological features such as metabolism, replication, etc.] is alive. How so? We can’t conceive of a "life-zombie", a creature that exhibits all the macroscopic biological features that e.g. a dog has, but isn’t alive. This second step in the reductive explanatory chain is often so obvious that it is skipped.

Lessons for empiricists:

1) Since empiricists want to connect explananda (often formulated in natural language) to empirical explanations they rely on conceivability arguments as much as anybody else. (They sometimes don’t notice the conceivability-step because it is trivial.)

2) Because conceivability arguments are just a trick to access the meanings of concepts, they can’t ever show us anything about the world. They can’t tell us whether houses or buildings exist. Similarly, the zombie-argument can’t show us that dualism is true. It can only show us that consciousness – if it exists – can't be reduced to physics. That’s no argument against eliminativism and no argument for dualism.

3) Conceivability arguments may rely on “intuitions” in some sense. But that’s no problem for several reasons. First: We understand the cognitive process in question fairly well, know it has evolved to be sufficiently reliable to make language work, and we can even make the data more representative by aggregating folk intuitions about conceivability of certain situations. But more importantly: If people disagree about conceivability, they merely disagree about their concepts! (Ignoring some complications about limited cognition here.) That’s why the zombie arguments can’t be question begging (as empiricists often allege) in any interesting sense. Disagreement about the conclusion almost certainly is just a disagreement about the concepts in question.

How should empiricist treat the conclusion of the zombie-argument? The argument shows that many people have a concept of consciousness that prevents the phenomenon from being reduced to brain processes. Others don’t. That’s all. To find out whether dualism, Russellian monism, or eliminativism is true, we need to look at other considerations: Are there reliable first-person data? Could we possess concepts that refer to physically inert entities such as qualia? Can we make sense of the idea that first-person data are illusory? Etc.

Carnap, Rudolf (1955) “Meaning and Synonymy in Natural Languages.” In: Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition 6.3, pp. 33–47