Transcendental Realism

Transcendental Realism
Dustin McWherter
IACR Pre-Conference Workshop
Cardiff University, 18 July 2016

Transcendental realism (TR) and philosophy of science
TR is the philosophy of natural science, or what might today be called 'general philosophy of science' or just 'philosophy of science', that critical realism (CR) advocates
It was the first component of CR to be developed and was introduced in Roy Bhaskar's A Realist Theory of Science (RTS), first published in 1975
Most people who take an interest in CR do so because of their interest in social science, but this makes a philosophy of natural science all the more important:
The way we answer some of the most fundamental questions in the philosophy of social science—i.e., what the proper objects of the social sciences are, what their methodologies should be, and thus how they can genuinely be scientific—will often depend on the way we view natural science or science in general
This is because the scientificity of social science will be gauged in relation to that of natural science, which is usually taken to be unproblematically scientific, whatever other issues may arise
Consequently, if there are significant problems in our philosophy of natural science or science in general, then it's likely that those problems will be transferred to whatever conception of social science we might pursue

RTS and 20th century philosophy of science
This was one of Bhaskar's motivations in writing RTS: to challenge some of the basic assumptions that were hamstringing philosophy of science at that time and to propose a more adequate alternative
The agenda for 20th century philosophy of science was largely set by positivism, which was heavily influenced by David Hume's empiricism
And so some basic tenets of Humean empiricism formed the basic context in which problems were treated in the philosophy of science
The most influential aspect of Hume's empiricism in this regard is its critique of natural necessity and causal laws, which resulted in the notion that, when it comes to causal relations or laws, the best that science can do is to track the regularities, or constant conjunctions of events, that are given in experience
We can't know any mechanisms, structures, or powers behind those sequences of events that generate them or make them necessary, and we don't directly experience anything that shows the connection between them to be necessary
So if there is any necessity in our knowledge of causal laws it has to come from us: either in a psychological habit, as in Hume, or in the way we conceptualize what we experience, which Kant initiated
In either case, a constant conjunction of events is at least a necessary condition for a causal law, and for some it's a sufficient condition as well
So you can see how all of this results from an empiricist epistemology: if we can only know what we directly experience, and we don't directly experience causal necessity, then the best we can do is focus on the regular sequences of events we do experience

The case for ontology
So TR opposes this approach to causal relations and laws, but before I say more about that I want to draw your attention to another basic tenet of Hume's empiricism that 20th century philosophy of science absorbed: a prohibition on ontology
Hume famously proclaimed that any work containing metaphysics should be 'committed to the flames', because he thought metaphysics or ontology always led us into confusion in taking us beyond what we directly experience
And so we can't say the world is this way or that. We can only say our experience is this way or that. We can never know what's going on 'behind the scenes'.
A similar idea occurs in Kant's distinction between appearances, which we can know through experience, and things in themselves, which can't be known at all
So what ends up happening is that ontology gets reduced to or collapses into epistemology, so that we can only talk about how we know the world and not about the way the world 'really is'
And that involves what Bhaskar calls the epistemic fallacy, which is the attempt to do away with ontology by converting any ontological question into an epistemological one, or to construe any ontological issue in epistemological—and thus anthropocentric—terms
But one of the main contentions of CR is that ontology can't be so easily quarantined from our theories of science or knowledge and that the attempt to do without ontology will only generate an implicit ontology
So for example, even if you say that we can only know sequences of events in experience and we can't know anything beyond them, you're at least committed to the view that the world as science knows it consists of such events and sequences of them
This notion that the world is constituted by experience is what CR calls 'empirical realism', which can have Humean or Kantian variants
At any rate, the point here is that even if you try to be neutral on the way the world is you end up with a fairly determinate commitment to the world being a certain way despite yourself
Therefore, CR seeks to show that ontology in general is both possible and necessary for an adequate understanding of science, and it also advocates a particular ontology to replace the implicit ontology of empirical realism and the conception of causal laws it contains

Transcendental arguments and immanent critique
One of the main arguments of RTS and for TR in general is what Bhaskar calls 'the transcendental analysis of experimental activity' (TAEA), which is a set of transcendental arguments that simultaneously provide an immanent critique of the Humean theory of causal laws and justify an ontology of generative mechanisms as the basis of causal laws
Before I discuss this argument in particular, I want to discuss transcendental arguments (TA) more generally and the role they play in CR
In my experience, the use of TA in CR makes some people wary, and that's often because they think TA involve some kind of infallibilism or foundationalism, or a rigid kind of apriority that's incompatible with naturalism, or that they lead to idealism because of their association with Kant
Like Bhaskar and others, I think all those assumptions are false, and that TA aren't just a legitimate way to reach conclusions in philosophy but also a highly recommendable way
An understanding of how CR's conception of TA compares with Kant's is key here
A TA can be broadly defined as an argument which takes some relatively uncontroversial feature of our knowledge or experience and then argues that something else—e.g., an epistemic principle or ontological structure—makes that phenomenon possible and thereby serves as its 'condition of possibility'
This can then be put in the common syllogistic form:
Only if Y, then X
X
Therefore Y
So a TA takes it for granted that X is the case and then argues that condition Y makes X possible, and therefore Y must be the case if X is
For example, Kant famously argues that cognitive experience is made possible by spatiotemporal forms of intuition and 12 conceptual categories of the understanding
Although Kant scholarship is quite varied and ongoing, there's good textual evidence for the view that Kant took his TAs to be infallible
And because he was mainly focused on the conditions of all experience rather than some region of experience, I think Kant tended to blur the kind of relative apriority that is specific to TA with the more traditional notion of apriority that involves a more rigid independence from all experience
At any rate, this infallibilist understanding of TA was seriously challenged by Stephan Körner, who showed that any TA's contention that Y is the condition of possibility of X can never decisively eliminate the chance that there could be alternatives to condition Y which we haven't thought of yet and which equally or better explain the possibility of X
And notice that Kant begins his TA with a premise that's seemingly impossible to doubt—i.e., that there is cognitive experience—but this doesn't save him from Körner's objection, because it's the inference to the conditions of possibility of experience, not the existence of experience itself, that the objection challenges
So even if Kant's TA can rule out the known alternatives to his conclusions, he can't guarantee that there's not some currently unknown alternative system of categories or even some non-subjective condition that equally or better explains the possibility of experience
(There have been Kantian responses to this objection, but I've yet to see one that really overcomes the basic thrust of the objection instead of finding fault with some of the details of its articulation)
The reason this objection holds is due to the very nature of the inference that produces the main claim of a TA, for when we try to reason from something to its condition of possibility, we're engaging in the kind of retroductive inference that we find in scientific reasoning that proceeds from some phenomenon to the mechanism that generates it
And that kind of retroductive inference is always fallible
Hence TA are compatible with scientific naturalism insofar as both philosophy and science so construed share similar methods of reasoning
So what Bhaskar does is to accept these fallibilistic consequences of Körner's objection and defend a modified conception of TA
If we accept the fallibility of TA, then Körner's objection is no longer an objection but simply a statement of the limits of TA
And even though we can't rule out currently unknown alternatives to the conclusions of TA, we can rule out currently known alternatives, and this is how TA function as immanent critiques of rival positions, which is very important for CR
You can see this in the way that TA start with a premise that is relatively uncontroversial and therefore widely accepted and then proceeds to a conclusion that isn't widely accepted
In so doing, it provides an immanent critique of rival positions by showing that those positions are incompatible with the uncontroversial premise accepted by advocates of those positions
A couple more points before moving on to the TAEA:
1) the kind of apriority involved in TA is simply relative to the phenomenon being explained, not absolutely a priori
So the conclusions of Bhaskar's TAEA aren't intended to be a priori with respect to all experience but only with regard to particular instances of experimentation and the application of its results
2) I'm unaware of any good reason to think that TA are inseparable from idealism, and I think the burden of proof is on anyone who thinks they are, since there's no obvious connection between a relatively abstract method and a substantive doctrine
There have been attempts to shows this inseparability, some of which I've addressed in publications, but I don't think any of them succeed

The TAEA's presuppositions
In my reconstruction of the TAEA, four basic presuppositions or premises are taken for granted:
1) Experimentation is a significant and intelligible feature of science (this just says that they are significant, not how)
2) Scientists initiate the sequences of events that experienced in experiments (since experimental settings are constructed and experiments executed by humans)
3) Constant conjunctions or regular sequences of events prevail (or can prevail) in experiments (repeatability of experiments; closed system)
4) Such constant conjunctions in some way give us epistemic access to causal laws, however the latter may be subsequently conceived
So if you reject any of these claims, you're not bound to the conclusions of the TAEA

Nature is, to a significant extent, an open system
An 'open system' should at this point just be understood as any domain wherein constant conjunctions of events aren't readily available (why this is the case will be clarified later)
So the contention here is that for the most part constant conjunctions of events aren't readily available outside experimental settings
If they were, we wouldn't need to construct and carry out experiments
If we could just observe constant conjunctions of events without experiments we wouldn't need to initiate them within experiments, and so the fact that we do initiate them within experiments entails that we don't normally observe them without them
Thus the extent to which nature is an open system is commensurate with the extent to which experiments are non-redundant
And so nature being an open system is a condition of the possibility (or intelligibility) of experimental activity

Causal laws must be distinct from constant conjunctions of events
This is a negative claim insofar as it takes a stance on what causal laws are not, and it's supported by a few arguments
The first is a reductio ad absurdum which says that, if causal laws were constant conjunctions of events, the experimental scientist would actually be generating a causal law when he or she initiates the sequence of events that occurs in an experiment
This would make scientists responsible for the existence of causal laws rather than simply the discovery of them
This would attribute quite an awesome power to scientists, but the real problem here is that it would be a power whose exercise would be arbitrary and unconstrained enough to make it epistemically worthless
For if scientists generated causal laws instead of discovering them, they'd have no epistemic reason to initiate sequence of events in experiments, and so the intelligibility and epistemic significance of experiments would be impossible
To avoid these consequences, Bhaskar argues that one must instead identify constant conjunctions of events as the empirical grounds of causal laws rather than causal laws themselves
So here already we have a point against the epistemic fallacy of empirical realism, for the argument here is that causal laws can't be reduced to the means by which we know them in experience
Another reason causal laws must be distinguished from constant conjunctions of events is that, if we were to identify them, things like the disruption of events in experiments or human error in the construction and execution of experiments would entail the extreme fragility or even the nonexistence of many causal laws
For example, anyone can interfere with and ruin the most carefully constructed experiment, thereby changing the sequence of events that occurs, but we'd never think such a person has the power to change the laws of nature
We could only resist that conclusion, though, if we distinguish causal laws from sequences of events, for the interfering person changes the sequence of events that would otherwise occur
Also, if we recall the previous conclusion that constant conjunctions of events rarely occur outside experimental settings, then we can see that if causal laws were constant conjunctions of events, there would be very few causal laws that exist outside experimental settings, which would mean that science has discovered very few causal laws that actually exist in nature
Consequently, the distinction between causal laws and constant conjunctions of events is a condition of the possibility of experimental activity, for too many absurd consequences result from its elision

Causal laws are the transcendentally real tendencies of generative mechanisms
They're 'real' in the sense that they exist independently from the conceptual, perceptual, and practical/material conditions we use to identify them in the social activity of science
They're 'transcendentally real' in the sense that their reality is a condition of the possibility of scientific knowledge via experimentation and application
Hence TR
A generative mechanism is a causal power something has in virtue of its structure—a power to make something happen and thus to generate events
Importantly, a generative mechanism is something that can be possessed without being exercised
And the concept of a tendency is a modification of the concept of a generative mechanism intended to highlight the fact that it can also be exercised without being realized (due to countervailing powers)
And so with the concepts of tendency and generative mechanism you have the idea of something that is responsible for the production of events and as such is irreducible to them
So if causal laws are tendencies, then causal laws are trans-actual—i.e., they are something more than their actualizations in sequences of events
This makes causal laws real possibilities (as opposed to merely logical possibilities or actual events)—possibilities of activity
1: Experimental results must be applicable to nature
If sequences of events in experiments give us something that we wouldn't already have, and do so for an epistemic purpose, then those sequences of events must be epistemically significant with respect to nature
If this weren't the case experiments would be too insular and arbitrary to be worthwhile
2: Generative mechanisms make extra-experimental application intelligible
If experimental results are applicable, then causal laws must pertain to open and closed systems alike and therefore be compatible with both irregular and regular sequences of events
There must be something about them that links what we find in experimental settings to nature so that knowledge gained in the former can be applied to the latter
According to Bhaskar, a generative mechanism fills this role well
It's independent from yet compatible with sequences of events insofar as it's a power to produce such events
And it's compatible with both regular and irregular sequences of events, and thus open and closed systems alike
This is because, as a distinctive causal power to produce a distinctive effect, it can operate with uniform consequences if its activity is unimpeded, and it can operate with variable consequences if its activity is impeded by other generative mechanisms
This would also explain why experiments need to be repeatable: we need to be sure that a generative mechanism really is operating unimpeded, and that others aren't interfering
So scientists need to intervene in the course of nature to isolate generative mechanisms for study, and it's precisely because generative mechanisms exist before, during, and after experiments that experimental results can be applied, for we must assume they're the same mechanisms when we isolate them, activate them, and observe them in conjunction with others
Furthermore, if generative mechanisms are operative in experiments, then we can explain how a scientist can initiate the sequence of events that occurs without being responsible for the existence of a causal law, because the scientist is simply activating the generative mechanism
3: Causal laws must be tendencies
Finally, if causal laws exist in both open and closed systems, then we can't define them in a way which says what they always do, because they may not always do anything in an open system
But at the same time we must say something about what they do, for otherwise the constant conjunctions of events we get in experiments would be neither informative nor applicable
So we must define them in terms of what they tend to do, because this tells us what their distinctive effect is without committing us to its actual occurrence
And so causal laws must be the tendencies of generative mechanisms, but how are they transcendentally real?
They're independent from the practical/material activity of experimentation since they must exist whether or not we intervene in the course of nature to study them
They're independent from the concepts and theories we have of them since they must be discovered a posteriori
They're independent from perception since they (as opposed to their effects) have no perceptible qualities

The TAEA thus points us toward:
The transitive and intransitive dimensions, which are the social process of theoretical and practical scientific activity (transitive) and the independent reality which it is the task of that process to investigate (intransitive)—the epistemological and the ontological
The interconnection of epistemic relativism, ontological realism, and judgmental rationalism
Epistemic relativism characterizes the TD, ontological realism is commitment to the ID, and judgmental rationalism is the relation between the TD and the ID, for it is the reference to an independent reality that allows us to rationally choose one theory over another as having greater explanatory power
The domains of the empirical, the actual, and the real, which are characteristically the domains of experience of events, events themselves, and generative mechanisms and their tendential activity
The real encompasses the actual, which encompasses the empirical
Contra empirical realism, the real can't be reduced to the actual or the empirical, and so actualism must be rejected
Philosophical and scientific ontology, which are the general ontology of generative mechanisms presupposed by science (philosophical) and the particular generative mechanisms discovered by science (scientific)
This gives us a kind of 'zigzag' relationship between methodology and ontology: the transcendental method gives us a philosophical ontology; the philosophical ontology has methodological consequences for science (e.g., prioritize explanation in terms of mechanisms over prediction of events); and the scientific method gives us a scientific ontology



'As far as certainty is concerned, I have myself pronounced the judgment that in this kind of inquiry it is in no way allowed to hold opinions, and that anything that even looks like an hypothesis is a forbidden commodity, which should not be put up for sale even at the lowest price but must be confiscated as soon as it is discovered. For every cognition that is supposed to be certain a priori proclaims that it wants to be held for absolutely necessary, and even more is this true of a determination of all pure cognitions a priori, which is to be the standard and thus even the example of all apodictic (philosophical) certainty.' (Kant, Critique of Pure Reason, Axv, original emphases). See ibid A2, B22, and A13/B27 for similar claims.
7