Imaginative Intuitions and Fanciful Fantasies: On the role of intuition in thought experiments

(1)

(2)

Introduction

“Fantasy is an exercise bicycle for the mind. It might not take you anywhere, but it tones up the muscles that can.”

Terry Pratchett

Much of philosophy is dominated by elaborate discussions about brains in vats, painted donkeys, experience-machines, run-away trolleys, sensible knaves, body-swapping

devices, states of nature, ‘philosophical’ zombies, Chinese rooms, missing shades of blue, and the list goes on. These scenarios are usually called thought experiments (henceforth: TE’s). They are hypothetical, and sometimes inherently impossible scenario’s. Since the birth of philosophy, philosophers have been using them to think about epistemology, ethics, meta-ethics, personal identity, justice, physicalism, the mind, empiricism, etc.

Philosophers do not have a patent on this method. Theoretical physics is also rich with “what if”-scenario’s. Think only of Galileo’s falling bodies, Maxwell’s demon, Newton’s bucket in absolute space, Schrödinger’s cat, or Einstein, running alongside a light-particle in his mind. Both scientists and philosophers use narratives about things that haven’t happened to gain insight into the world and the concepts we use to describe it. 1

TE’s have been used to test both scientific theories and philosophical analyses. Yet it is unclear how this use of TE’s to test theories and analyses is supposed to work. How can hypothetical scenario’s, “what if”-cases, provide us with knowledge about these

things? Since we can only put information that we already have into our imaginings, how is it that TE’s seem to be able to provide us with genuinely new knowledge? Can they

legitimately do so, or are they just useless fantasy?

The question above is also called the problem of informativeness. An often heard answer to the question is that TE’s elicit intuitions and that those intuitions serve as evidence for or against a philosophical claim. Alternatively, philosophers sometimes claim that it is intuitive that a TE will have a certain outcome, or that a claim about a TE is intuitive. In those cases, no specific ‘intuition’ is being referred to, but rather the fact that some claim is appealing or self-evident. TE’s might be informative because they draw our attention to such self-evident truths by having us recombine information that we already have in

strange and interesting new ways. Not only philosophers appeal to intuition in these ways.

I leave aside TE’s in history here. How would the age of exploration have turned out if the Hongxi Emperor 1

had not burned down the Chinese fleet in 1424? Or TE’s in economics. What would happen if the EU reinstated internal tariffs? Entertaining these kinds of questions is a controversial practice (less so in economics), yet often done in order to emphasise the importance of certain historical events or of certain economic policies.

(4)

Einstein explicitly appealed to his intuition in his light-beam TE and he is not the only one to have emphasised the role of intuition in TE’s in science.2

The following example may help to clarify what the role of intuition is supposed to be: One of the earliest TE's in political philosophy can be found in book 1 of Plato’s Republic. Socrates engages in a dialogue about Justice with the elderly Cephalus. Asked to define Justice, Cephalus defines it as ‘giving to each their due’. In response, Socrates points out to Cephalus that this definition implies that Justice would require returning a borrowed sword to a person that has gone mad. Agreeing that this is a very unhappy consequence of his definition, Cephalus rejects it as inadequate.

Why does Cephalus reject it though? Socrates offers no real argument here. Even though prima facie Cephalus’ definition of Justice seems fine enough, the idea of giving a sword to a raving lunatic seems so straightforwardly (evidently) wrong, that Cephalus is convinced that his definition must be too. But there are no premises that bring Cephalus to this conclusion. After considering the scenario, the conclusion is simply arrived at directly. No one in his right mind should give weaponry to someone not in his right mind! It seems so obvious that, if this is what justice requires, then it cannot be Justice with a capital ‘J’. Where does this feeling of obviousness come from though? And why is it sufficient to completely overturn Cephalus’ definition of Justice?

Recent years have seen a fountaining of literature on the (lack of) reliability of intuition and, by extension, the (lack of) reliability of TE’s. Yet the way in which intuition and TE’s are supposed to be linked is a topic that has gotten less attention. The purpose of this thesis is to investigate this link between TE’s and intuition. The questions that I want to answer are (1) what is the function of intuition in TE’s? and (2) given this function, what is it that we can learn from TE’s?

In order to answer these questions, I will first discuss what TE’s are and how they are used (chapter 1). Then, after considering a dominant alternative view, I will adopt a view of intuition that understands it in terms of tacit information processing (chapter 2). This will put me in a position to investigate the link between intuition and reasoning with imagined, hypothetical situations in chapters 3 and 4. In chapter 3 I deliberate over different responses to the problem of informativeness before suggesting how -in the light of the so-called ‘mental models’ view- tacit information processing can give an answer to the problem of informativeness for at least a subset of TE’s. In chapter 4 that analysis is extended to TE’s that are of most interest to analytic philosophers that employ the method of conceptual analysis. Finally, in chapter 5, I will examine various ways in which we can expect our intuitive system to misfire in reaction to a TE. Also, I will discuss a problematic practice of philosophers: explaining our intuitions by referring to their cause. In my

conclusion I will return to the central questions of the thesis.

See p.49-50 of his Autobiographical Notes. Einstein writes that “From the very beginning it appeared to me 2

intuitively clear that, judged from the standpoint of such an observer, everything would have to happen

(5)

Chapter 1 Fantasies and other experiments in thought

In this chapter I will discuss the phenomenon of TE’s itself. As the aim of this thesis is to evaluate the role of intuition in TE’s, I first ought to take a closer look at TE’s themselves. What are they? What are they used for? How are they used? Are there any important qualitative differences between TE’s that could influence the role of intuition in them? (Short answer: yes).

First I will discuss several examples of influential TE’s in order to show something of their incredible versatility. As TE’s are applied to different domains, in different ways, for different reasons, and to different effects, it might be wondered if anything general can be said of them. I argue that this is possible; TE’s most characteristic features are that they 1) involve (reasoning with) a hypothetical scenario and 2) are used to test theories. The 3

dialectical function of TE’s is to either challenge, or generate theories by means of an imagined situation.

This being the case, it is natural to wonder how TE’s could actually challenge or generate theories. How could a hypothetical situation provide evidence for or against a theory? One often-heard answer is that our intuitions about TE’s have something to do with this. Yet before delving into the question of what intuitions are (in chapter 2) and into the epistemic question (i.e. the matter of what kind of knowledge TE’s can provide) in chapters 2 to 5, I will provide a two-part taxonomy of TE’s in §1.3. This taxonomy is very much in debt to Tamar Gendlers tripartite distinction of TE’s, introduced in her book Thought Experiment. I differentiate between predictive and evaluative TE’s. 4

The primary reason for doing so is that it will turn out that our intuitions have a different role in predictive TE’s than in evaluative TE’s. It is thus of importance to the central question of this thesis, i.e. what is the role of intuition in TE’s? These different roles will turn out to be relevant to the epistemic question (to be discussed in chapters 3 and 4). A secondary reason for doing so is that the distinction is interesting in its own right, as it may help to clarify how TE’s achieve, or fail to achieve, what they (or their author) set out to achieve.

Predictive TE’s involve playing out of certain assumptions in order to predict what would

happen in the TE. The question we ask about these scenario’s are: “what would happen?” These TE’s are in the business of making empirical predictions, even if in some cases these predictions could never become empirical because of their inherent impossibility, e.g. because they suppose frictionless planes, or veils of ignorance. Many -though not all- TE’s in physics are predictive.

I use the word ‘theory’ in the broadest sense possible here, since there are obvious differences between 3

theories in physics and, for example, analyses of the concept of KNOWLEDGE in terms of necessary and sufficient conditions, all of which have been (con)tested with TE’s. More on that in §1.2.

T.S. Gendler (2000), p.25. 4

(6)

Evaluative TE’s are not concerned with what would happen in the TE. Rather, they are

concerned with the question of how to evaluate the described facts of the TE. The question we ask about these scenario’s is: “How, given that X happens, do we describe or evaluate that?” For instance, in John Locke’s classic TE in which a prince and a cobbler swap all their memories, the facts of the scenario are not in dispute: Locke simply states that there is a soul-swap. Whether or not that is possible, or how that is supposed to happen, is irrelevant to the case. Of philosophical interest is how we evaluate, or describe, the soul-swap: do we say that the prince and cobbler have swapped their personal identities? Does that also mean that ones personal identity is identical to ones memories? Many -though not all- TE’s in philosophy are evaluative.

We do not have the same kind of epistemic access to predictive TE’s as to evaluative TE’s. Both types of TE’s are used to test theories, yet they do so in different ways. Predictive TE’s make empirical predictions; They predict that some specific outcome ‘X’ would

happen under the described circumstances. Thus, the outcome of a predictive TE depends on whether or not X would, in fact, happen if the circumstances described by the TE were to hold.

The outcome of an evaluative TE crucially depends on how the author of the TE, or his/her audience, evaluates or chooses to describe the TE. May we turn the trolley? Do we say that the cobbler is still the cobbler when he has the prince’s soul? The main

epistemological issue with evaluative TE’s is how to judge the validity of these evaluations. Evaluative TE’s are often said to rely largely on our intuitions about the case. We may be said to have the intuition that a cobbler -ensouled by a prince and thus with all his

memories and ‘princely thoughts’- really is that prince (and that personal identity must therefore be independent of the identity of substances). As such it may seem natural to solely focus on evaluative TE’s in this project. Yet intuition is also claimed to play a significant role in predictive TE’s, albeit in a different fashion. It is a central element of Ernst Mach’s analysis of the use of TE’s in the sciences. Tamar Gendler also argues that intuitions are an important part of predictive TE’s. Can these accounts match up, or are they really about something else? I will argue for an integrative view of intuition that explains its role in both types of TE (in chapters 3 and 4).

§1.1 Thought experiments? Some examples.

Do we know what TE’s are? Despite a recent surge of interest in the topic of TE’s, there is little consensus on this. The stories that we recognise as TE’s are incredibly

heterogeneous. Can we honestly say that there are significant similarities between, say, Galileo’s falling bodies, Thomson’s runaway trolleys, and Hobbes’ state of nature? Is there anything general shared between them? Brown tells us that “If we are ever lucky enough to come up with a sharp definition of thought experiment, it is likely to be at the end of a long investigation.” (Brown, 2nd edit. 2011) and also: “We recognise them when we see them (Brown, 1991, 1)”.

(7)

So without further ado, let’s take a look at some TE’s, one from the history of science and the others from philosophy, to see if there are any similarities to be found. Galileo’s balls:

Galileo’s argument against Aristotle is one of the best studied examples of a TE in science (Sorensen (1992a), Gendler (2000), Brown (1991), Häggqvist (2009a), Norton (1996), Kuhn (1964), Cargile (1987)). It has become the paradigm case of a persuasive TE in physics. In his Discorsi, Galileo brings to mind Aristotle’s claim that heavier bodies fall faster than lighter ones. To challenge this claim, he asks us to imagine what would happen if a cannonball were tied to a lighter musket ball and they were both dropped

simultaneously from the tower of Pisa. As the musket ball is lighter than the cannonball, it should fall slower than the cannonball and therefore retard its falling-speed. However, this would entail that the musket ball and the cannonball combined fall slower than the

cannonball alone, which is patently absurd, since the two bodies together are heavier than the cannonball alone and, thus, should fall faster. The TE results in a paradox for Aristotle: if he is right, then the connected bodies should both fall slower and faster than the

cannonball. To resolve the paradox, Galileo asserted that heavy and light bodies must fall equally fast.

Galileo starts his TE with a theory: Aristotle’s claim that heavier bodies fall faster than lighter ones. In the TE, the theory leads to a paradox. The TE ends with Galileo supplying a new theory of falling bodies (namely that weight does not affect falling speed), which resolves the paradox. The TE contains a prediction: asked the question “How will the bodies fall?” Galileo answers that, all things being equal, they will fall equally fast. Thomson’s runaway trolley:

Although Philippa Foot introduced the first trolley-case, it was J.J. Thomson’s reckless use of trolleys in her (1976) paper ‘Killing, Letting Die, and the Trolley Problem’ that really kicked off the industry of ‘trolleyology’. Thomson’s (1976) paper was originally meant as a discussion about the question whether the moral rule that ‘killing is worse than letting die’ (which emerged out of debates about the moral permissibility of euthanasia and abortion) is a good guideline for making moral decisions. Thomson invents many versions of the same hypothetical situation in order to draw out the salient aspects of the problem that she is investigating. Her (1976) paper alone contains over 24 TE’s (!) about trolleys, magical health pebble’s, atomic bombs, and prodigious surgeons.

Variations of the trolley-problem are nearly endless, but the setup is always the same: a runaway trolley is hurtling towards people who are stuck on the track. The second instantiation of the trolley-problem in Thomson’s paper runs as follows:

“Frank is a passenger on a trolley, whose driver has just shouted that the trolley’s brakes have failed, and who then died of shock. On the track ahead are five people; the banks are so steep that they will not be able to get off the track in time. The track has a spur leading off to the right, and Frank can turn the trolley

(8)

onto it. Unfortunately there is one person on the right-hand track. Edward can turn the trolley, killing the one; or he can refrain from turning the trolley, letting five men die.” (Thomson, 1976).

Thomson assumes that, like herself, readers will find it morally permissible to turn the trolley. This, however, means that in some situations ‘we’ would find it morally permissible to choose to kill, rather than let die, when presented with the choice. Therefore the theory that ‘killing is worse than letting die’ doesn’t seem to provide good guidance in this case. This TE hinges on the evaluation that it is morally permissible to turn the trolley. It is that verdict about the TE that serves to cast the principle that ‘killing is worse than letting die’ into doubt.

Throughout the paper, Thomson presents many hypothetical cases in which she assumes that it is fairly obvious what the ‘right’ choice is, but in which it is not so obvious which moral principle would explain that rightness. Using this method, Thomson comes up with several morally salient reasons (such as having a claim and acting on the oncoming danger, rather than on a person in order to prevent a threat) that, in her cases, trump the claim that ‘killing is worse than letting die’.

Thomson starts out with a theory and, throughout her paper, continually adjusts it so as to fit all the hypothetical cases that she can come up with.

Gettier’s accidental truth:

Gettier’s seminal paper ‘Is Justified True Belief Knowledge?’ has become such standard fare in epistemology that it is commonly called ‘the Gettier paper’. He countered the then widely accepted analysis of ‘knowledge’ in terms of ‘justified, true belief’ (JTB) with but two possible scenarios in which a belief, though justified and true, turns out to be true merely by accident.

In the first case we meet a Jones and a Smith who are both applying for the same job. After his interview, Smith is sure that he will not get the job, as he was just told by the very president of the company that Jones would get the job instead. Smith is also sure that Jones has ten coins in his pocket (presumably because he saw Jones fidget around with those coins before putting them in his pocket). On the basis of all this evidence, Smith forms the belief that ‘the person with ten coins in his pocket will get the job’. Much to his surprise, however, Smith does get the job. Coincidentally, he also happened to have ten coins in his own pocket at the moment that he formed the belief that ‘the person with ten coins in his pocket will get the job’. This belief turned out to be true. Given the evidence he had, the belief was justified. Smith therefore had a justified, true belief, even though the truth of the justified belief was merely accidental.

Gettier asserts that in these cases we shouldn’t talk of ‘knowledge’, even though the cases satisfy what Ayer and Chisholm (and, perhaps, Plato) took to be the necessary and

sufficient conditions for knowledge. Many have agreed with Gettier and concluded that his cases falsified the analysis of knowledge as justified, true belief. It would be an

(9)

understatement to say that the Gettier-cases caused quite a stir amongst epistemologists; It changed the business entirely.

Gettier starts out with a definition, i.e. that justified, true belief equals knowledge. His cases satisfy the conditions for the theory, yet intuitively fail to satisfy the concept of ‘knowledge’. The case hinges on this evaluation of the case. For those who accept the ‘Gettier-intuition’, the theory of JTB as being necessary and sufficient for knowledge is falsified.

Both Galileo’s and Gettier’s TE’s supply us with a counterfactual scenario in which a certain theory is put to the test and found to be inadequate. Indeed, many TE’s work as ‘theory-busters’ and quite a lot of attention has gone to this function of TE’s. There are, 5

however, examples in which TE’s do not explicitly set out to falsify some theory. Examples of these types of TE's play a prominent role in contractarian theories of justice.

Rawls and the Original Position:

Rawls famously asked his readers to consider the decision-process of persons behind a veil of ignorance. Rawls calls this ‘original position’ “purely hypothetical”. Behind the veil of ignorance “no one knows his place in society, his class position or social status, nor does anyone know his fortune in the distribution of natural assets and abilities, his intelligence, strength and the like.” (Rawls 1971, section 3). The people behind the veil of ignorance are asked to deliberate, and agree on, the principles according to which societal institutions ought to be shaped. Which principles will they agree on? Rawls predicts that they will converge on the two fundamental principles of justice.

Rawls’s original position could be interpreted as a simulation in which people with certain characteristics and abilities are put together in order to deliberate about the society they would wish to live in. As people are a varied bunch, Rawls severely minimises the characteristics and abilities of the people in the original position in order to make the experiment predictable. They have no knowledge about their position in society, their wealth, health, skills, race, sex or intelligence and therefore will be unable to act

accordingly. This basically clears them of all self-interested bias and makes them perfectly conform elements, with no desires one way or the other. However, they do have a thin conception of the good (meaning that they will value what Rawls calls ‘primary goods’), as well as rationality, and the ‘two moral powers of moral personality’ (being the capacity to honour fair terms of cooperation and the capacity to decide upon a conception of the good).

Since all the people in the original position are nondescript by virtue of the veil of ignorance, the only really active elements in the TE are rationality and the two moral powers. What we see in the original position is these three elements, operating in an isolated, bias-free environment, upon a thin conception of the good. What Rawls is doing is asking us how we think that situation will play out. He asks us ‘what would happen when these bias-free people were to decide on the basic principles of a just society?’

See, for example, the model of TE’s developed by Soren Häggqvist (2009a), and that of Timothy 5

(10)

Rawls does not start out with a theory. Nor does the original position serve to falsify any theory in particular. His Theory of Justice is generated for an important part by the original position by playing out assumptions about certain elements (persons with three powers and a thin conception of the good) within a very specific setup (the original position). Implicit to the method is the assumption that this method is the right way for generating principles of justice. It wouldn’t be much of an experiment if it weren’t.

In any case, Rawls’s principles of justice are described as the result of this experiment. Insofar as we agree with Rawls about a) his theory of how the element of ‘rationality’ will behave in this setup, b) Rawls’s theory of how the moral powers will influence that behaviour in this setup, and, c) the logical consequences of (a) and (b), we will end up agreeing with him that people behind the veil of ignorance would agree on his famous principles of justice. Rather than formulating a theory at the outset and using a TE to falsify that theory, Rawls uses his TE to end up with a theory, i.e. his Theory of Justice.

These kinds of TE’s -which serve to directly generate a new theory- are what Brown calls ‘direct TE’s’. They are a subset of predictive TE’s.6

This small foray into the world of TE’s will have to serve as an incomplete illustration of their versatility. Versatility in terms of domain (science, ethics, epistemology, etc.), in terms of how they are used, and in terms of how they relate to their target theories. Galileo simultaneously falsifies and instantiates a theory about falling speed with his TE. Gettier's TE’s are meant as counterexamples to a theory of knowledge. Thomson used trolley-cases to falsify a theory about moral permissibility. Yet at the same time Thomson tried to find out which moral principles could explain our sentiments about her cases and those moral principles are next used to come up with a better theory of moral permissibility. Rawls instantiated his Theory of Justice wholly by contemplating the original position. The theory is seemingly generated directly by the scenario. These differences are significant and will be further explored in section §1.3. First, however, I wish to draw attention to certain similarities that all TE’s share.

§1.2 TE’s? A definition.

In §1.1 I have described several TE’s from different domains and with different functions. What are their similarities? One interesting element of the TE’s above is that they all relate to a specific theory, in the sense that they either seek to problematise or refute, or to generate one.

Galileo’s TE relates to Aristotle’s theory of gravity, Gettier’s TE to the theory that knowledge is justified, true belief, Rawl’s TE to a theory of justice, and Thomson’s TE’s to the theory that ‘killing is worse than letting die’. In fact, all TE’s relate to a specific theory or other. Harry Frankfurt’s dr. Black serves as a counterexample to the theory that ‘being able to do otherwise’ is a necessary condition for moral culpability. Searle’s ‘Chinese Room’

Brown (2011), The Laboratory of the Mind, p. 39-41. 6

(11)

challenges the theory that a sufficiently advanced AI would ‘understand’ language in the same way that humans do. Descartes dreamed up a malevolent demon to demonstrate that true knowledge cannot be based on sensory perception. Mary, the colour-blind

scientist, is a challenge to physicalism, just like philosophical zombies are. The TE known as ‘Maxwell’s Demon’ challenges the Second Law of Thermodynamics. Einstein’s 7

description of running alongside a light-beam challenges ether-theory and Maxwell’s equations. The list goes on.

I use the term ‘theory’ broadly here. It also includes analyses of concepts -such as the analysis of ‘knowledge' in terms of instances of justified, true belief, or the analysis of ‘personal identity’ in terms of bodily continuity- and philosophical positions, such as the ‘strong AI’ position, which holds that "the appropriately programmed computer really is a mind, in the sense that computers given the right programs can be literally said to

understand and have other cognitive states.” This is important because philosophers 8

often rely on TE’s as a method for testing conceptual analyses and philosophical positions. Frank Jackson calls this ‘the method of cases’. 9

This means that at a superficial level I do not distinguish analyses of concepts from scientific theories. Although this makes my notion of ‘theory’ rather promiscuous, there is no immediate reason to worry about this. Conceptual analyses usually spell out conditions for the application of a concept and hold that, in all possible cases, if those conditions are met, the concept should apply. This is the case when they are meant to apply descriptively, as well as normatively. Concordantly, scientific theories make predictions about how 10

physical systems should behave that are meant to apply in all possible cases in which the physical laws remain the same.

I will return to the issue of conceptual analysis in chapter 4. For now it suffices to say that under ‘theory’, I basically mean to include all general statements that have the form of “All objects obey the laws of gravity” and “All cases of JTB are cases of

knowledge”. What analyses of concepts and physical laws have in common is that they

To be precise: it is meant to demonstrate that the 2nd Law of Thermodynamics is a statistical law, rather 7

than a physical necessity. Searle (1980), p.418. 8

Frank Jackson (1998), p.61. 9

There are two senses in which we can understand the claim that, according to a conceptual analysis, a 10

concept should apply when certain conditions are met. The first is descriptive: in these cases the analysis of a concept and the concept itself are meant to co-refer: the analysis is correct when in all situations it will have the same referent as the concept (e.g. The analysis of KNOWLEDGE in terms of JTB is correct iff cases of JTB and cases of KNOWLEDGE always refer to the same sets of facts). The second is normative: normative conceptual analyses spell out rules for properly applying a concept; We may only apply a concept when certain specific conditions are met. Hacker is someone who extensively uses analyses in a normative sense, for example when he criticises neuroscientists for habitually ascribing ‘understanding’ to brains, rather than persons. “[P]sychological predicates apply paradigmatically to the human being (or animal) as a whole, and not to the body and its parts.” (in Philosophical foundations of Neuroscience, p.73). These differences between types of analysis are important. Many projects of conceptual analysis contain both descriptive and normative elements, in accordance with Carnap’s notion of explication. Note, however, that both descriptive and normative analyses of concepts have the form of a general claim: “in all cases, when conditions x,y, and z are met, the concept will/should apply”. Such general claims, be they normative or descriptive, are often targeted by a TE’s.

(12)

purport to apply in all cases. They both make claims that are supposed to be generally applicable. And this general applicability is exactly what TE’s often seek to challenge. As such, the function of TE’s that test general claims in philosophy is not fundamentally different from TE’s that test physical laws.11

With these caveats about my use of the word ‘theory’ out of the way, I argue that theory-relatedness is one of the central elements of TE’s. Even though we may not know exactly what TE’s are, we do know which features are most characteristic of them. Namely: (1) imagining a hypothetical scenario, (2) actively reasoning with this hypothetical scenario (3) in order to test some theory. Putting these three features together I would propose the following definition, which is only a small adjustment of the definition that T.S. Gendler proposed in her 2004 paper:

“To perform a thought experiment is to reason about an imaginary scenario in order to test a theory”12

Three elements are central in this definition. First, it involves reasoning about a scenario. This serves to separate thought experimentation from just entertaining some

counterfactual thought, such as when we enjoy a story. Second, the scenario is accessed via the imagination, rather than via observation. Third, the aim of a TE is to test a theory. This last feature separates reasoning with TE’s from reasoning with hypothetical scenario’s in order to find out whether, for example, some engineering solution is really possible (questions like: “could this building support a swimming-pool on the top-floor?” are of this type). Merely hypothesising is not sufficient for an imaginary scenario to be a TE.

‘Testing something’ is a rather vague term and fails, perhaps, to be very informative. It is meant to at least include the functions of confirming, disconfirming, corroborating, and generating a theory. This broadness doesn’t need to worry us overly much though, since the idea of the definition is to capture as large a class of TE’s as possible, while still being able to differentiate between reasoning with TE’s and other types of counterfactual

reasoning that are less interesting to philosophical method. I will call something a TE only if it conforms to the above definition.

Using this definition means that I will be ignoring other interesting uses of imaginary scenarios. Sometimes cases are used merely to provoke and interesting reaction. The object of interest in those cases is people’s reaction to the hypothetical situation, not how that reaction relates to a theory. It is, for example, highly interesting in itself that many

I am skeptical of the distinction between TE’s in science and TE’s in philosophy in general. I think that that 11

distinction is neither interesting, relevant nor informative. I shall not pursue the matter here however. Gendler’s definition is: “[T]o perform a thought experiment is to reason about an imaginary scenario with 12

the aim of confirming or disconfirming some hypothesis or theory” (2004, p.1154). She emphasises that TE’s

either confirm or disconfirm a theory. This makes her definition more stringent than mine, as it seems to exclude an important function that TE’s can have, namely using a hypothetical scenario in order to draw out the consequences of a theory. Schrödinger’s cat (and any other reductio ad absurdum by means of a TE) is an example of just such a TE.

(13)

people that choose to divert the trolley in order to save five lives, thereby killing one person, refuse to push a large man onto the trolley-tracks to stop the trolley and save five lives. Equally interesting is that some people who are willing to push the large man, would not be willing to kill a healthy person in order to harvest his organs, thereby saving five lives.

This puzzling variety of reactions has stimulated much philosophical (and

psychological) study and debate. Although these responses have, of course, been used in building theories about moral permissibility (not to mention meta-ethics), they have also been investigated as an interesting phenomenon in itself. Crucially, insofar as these cases are only used to elicit reactions, without using those reactions to make a normative point (e.g. to use them to argue against utilitarianism), they are not used to test a theory and therefore not a TE on the above definition. The difference is between taking responses to a hypothetical scenario (by philosophers or a larger public) as data points for/against a

theory, and between taking responses to a hypothetical scenario as an object of study itself (as psychologists do when they discuss people’s reactions to trolley-cases).

Another use of TE’s that I will ignore is their illustrative use. Hypothetical cases can be used to simply illustrate what a certain theory entails. A utilitarian might describe a situation in which government authorities decide to execute an innocent man in order to placate the masses, who are in a riotous mood. If the aim of that story is to simply illustrate what utilitarianism dictates in some situations, then the TE is merely an illustration of that theory and its commitments. It is not meant as a test.

Cases that were initially meant to merely stimulate debate, or to merely illustrate a theory, can be repurposed into TE’s proper and vice versa. Trolley cases only become TE’s as defined above when people’s reactions to those cases are used normatively, in order to put a moral theory to the test. The utilitarian case described above might be used by someone else in order to argue against utilitarianism (it was not originally meant to be used that way). The same narrative can serve a variety of purposes. Sometimes it may be hard to disentangle the different uses. Yet the purpose that is metaphilosophically most interesting, and that I will be referring to when I talk of ‘TE’s’, is that of using hypothetical cases to test theories.

§1.3 Predictive and evaluative TE’s.

I would now like to return to the distinction I made above between different ways in which TE’s can relate to theories. It is a two-part distinction, based on the three-part distinction T.S. Gendler makes in her book ‘Thought Experiment’. This distinction will be important 13

for understanding the different ways in which intuitions play a role in TE’s. It will also clarify some of the different ways in which TE’s can challenge (or come up with) theories. They either do so by playing out assumptions within a specific scenario (predictive TE’s), or they do so by asking us to how to describe or evaluate a certain scenario (evaluative TE’s).

Gendler, T.S. (2000), Thought Experiment, p.25. 13

(14)

Those are two qualitatively different types of reasoning and this is important in considering the role of intuitions in TE’s.

(1) Predictive: what would happen in situation ‘TE’?

Examples:

Galileo’s falling bodies, Simon Stevin’s inclined plane, Rawls’s original position, Mary the colour-scientist, Searle’s Chinese Room, Einstein and the light-beam, etc.

(2) Evaluative: given (1), how should we judge or describe situation ‘TE’?

Examples:

Trolley cases, Gettier cases, Fake-barn cases, Putnam’s Twin Earth, Newton’s bucket, Parfit’s teletransportation-device, Nozick’s experience machine, etc.

Predictive TE’s ask: “what would happen if {x, y, z} were the case?” and make a prediction, whereas evaluative TE’s ask: “given that {x, y, z} are the case, how should we describe or judge the situation?” The fault-line between the two types is whether or not actually 14

performing the TE could, theoretically, make a difference. When the central question about a TE is: “what would happen?”, it would theoretically be able to confirm or disconfirm the prediction that the author makes if we were to replicate the conditions described by the TE and then ‘run the experiment’. Do different bodies really fall in the way Galileo predicted? (in a vacuum: yes) Will people behind a veil of ignorance (presuming, for a moment, that we could simulate such a veil) really come up with the two principles of justice?

Trying to find out what would happen is not the same thing as trying to make sense of an imagined case (in the sense of deciding whether or not a concept should apply, or whether or not an imagined action is allowable according to our values), which is what we do in evaluative TE’s. We assess predictions differently than we do evaluations. This means that in order to assess a TE, it matters what type of TE it is. Does Searle claim that the Chinese Room system would, in fact, not understand Chinese? Or is he arguing that, given that the system would not understand Chinese in the same way that Searle understands English, we ought not to apply the concept of ‘understanding’ to the system? Both are reasonable interpretations of the TE, yet they are different sorts of claims that cannot be evaluated in the same way.15

This is because the epistemic access we have to these different types of questions differs. For predictions, of central importance is whether the assumptions used in the prediction are being played out right: does the TE make the right prediction given the

Gendler (Idem.) distinguishes between factive TE’s, that ask “what would happen?”, descriptive TE’s, 14

that ask “given what would happen, how should we describe what would happen?”, and valuational TE’s, that ask “given what would happen and how we describe it, how should we evaluative what would happen?” I have lumped descriptive and valuational TE’s together and renamed ‘factive’ TE’s to ‘predictive’ TE’s.

Although Gendler says that factive TE’s mostly correspond to science, she acknowledges that TE’s in the sciences can be descriptive and that philosophical TE’s can be factive.

Although note that evaluative TE’s depend on agreement on “what would happen” in the TE. Partly 15

(15)

assumptions it makes, or should it have a different outcome? Other questions pertinent to the assessment of a predictive TE: Are some of the assumptions bad ones and should they be replaced? Are any important facts being overlooked?

With evaluations, the epistemic question is how to judge their validity. Are

evaluations of TE’s well reasoned for? Are they widely shared? Are they salient enough? Are there any contending evaluations that may work better? Those are questions we can ask in order to assess the validity of evaluative TE’s.

It follows that to in order to resist the conclusion of a TE, different strategies may be required depending on whether it is a predictive or an evaluative TE. Against a predictive TE one can argue that the operative assumptions have not been played out correctly; It makes an incorrect prediction. Usually this strategy is not available against evaluative TE’s. In evaluative TE’s the question is whether the verdict we have about the TE actually counts for anything. Strategies against evaluative TE’s will be discussed in chapter 2 and 5.

§1.3.1 Predictive TE’s

Predictive TE’s can be adequately described with Laymon’s definition of TE’s from his (1991) paper:

A thought experiment is an ordered pair < φ , ϴ > where φ is a set of persons (audience and/or presenter) and ϴ is a set of statements {T, P1, P2,….., Pn, Q} where:

(1) T is a description that is not in fact true (because it is idealized) of any experiment in this world;

(2) Members of φ believe that P1, P2,….., Pn are scientific laws or principles;

(3) Members of φ believe that ∃x(Tx) & P1, P2,….., Pn ⇒ Q.16,

The point of predictive TE’s is to describe a set of circumstances and argue that, if those circumstances would be the case, then some ‘Q’ would be the result. Importantly, in predictive TE’s the resulting Q is meant to do the work of testing the theory.

Examples:

- Rawls: His TE asks us which principles the people behind a veil of ignorance would actually decide on. Rawls’ claim is that people in this situation would, in fact, agree on 17

Rawls’s Principles of Justice. This simultaneously justifies them. With Rawls, Q is the

Laymon (1991), p.168 16

Even though Laymon designed his definition only to explain TE’s in science, we can easily transpose it to 17

apply to philosophical TE’s too. To do so, P1, P2,….., Pn would have to include definitions of certain concepts and/or philosophical positions. Applied to Rawls’ original position, we might say that P1, P2 and P3 stand for: a thin conception of the good, rationality, and the two moral powers of moral personality. Within description T (people behind the veil of ignorance), these then lead to Q (the principles of justice) according to φ (Rawls and the people convinced by him).

(16)

fact that people behind the veil of ignorance converge on the Principles of Justice, which is what generates his theory.

- Galilei: Here Q is the paradox that the tied-together lighter and heavier bodies must fall both faster and slower than the heavy body would alone and that result conflicts with Aristotelian physics.

- Jackson: Q is the prediction that Mary the colour-scientist “will learn something about the world and our visual experience of it”. Since according to physicalism, her colour-18

knowledge ought to be regarded as complete (meaning that she had nothing left to learn about colours), the prediction that she would learn something new contradicts

physicalism. Thus, if we believe that Q will be the case, then physicalism will have to go. - Searle: Q is the prediction that Searle would not understand Chinese, despite being able

to pass the Turing-test by answering questions about a story written in Chinese. This contradicts ‘Strong AI’, which holds that any computing system that passes the Turing-test will understand the language in which it can pass that Turing-test. Thus, if we believe that Q will be the case, then strong AI will have to go.

Note that conceptual analyses and predictive TE’s are not mutually exclusive. For example: Rawls’ original position is a tool for analysing the concept of Justice: it is analysed in terms of, amongst other things, equality (hence the veil of ignorance) and rationality. In order to draw out the consequences of what this analysis would practically entail, Rawls uses a predictive TE. His TE can thus be seen as being central to Rawls’ analysis of the concept of Justice.

The goal in predictive TE’s is to convince the skeptic that Q would indeed be the case in the TE. This means that one way of resisting the conclusion of a predictive TE is to contest that Q would happen, since should it turn out that it would not, then the TE would fail. Thus, Harsanyi (1975) famously contested Rawls’ claim that people behind the veil of ignorance would use maximin reasoning: his assumptions about rationality were wrong. 19

It has likewise been objected to Hobbes that his state of nature is predicated on a false description of human nature. It has been objected against ‘Mary’ that being deprived of colour-sensations since childhood might actually prevent her from seeing colours when she is finally released (Atkins 2001). Examples abound. Such objections are not available (or, rather, not very convincing) against evaluative TE’s.

Some philosophers may balk at my grouping of ‘Mary’ and the ‘Chinese Room’ with predictive TE’s. The reason for doing so is simple: in the original texts, the authors explicitly make a prediction about what would happen; They make a prediction about which facts would hold in the TE. Jackson (1982) writes: “What will happen when Mary is

Jackson (1982) p.128 18

Harsanyi’s criticism highlights how conceptual analyses and predictions may intersect: using a different 19

analysis of rationality entails making a different prediction about the outcome of the Original Position. The claim is not merely that Rawls’ analysis of Justice is wrong. The claim is that Rawls’ TE would play out differently.

(17)

released from her black and white room or is given a colour television monitor? Will she learn anything or not?” (emphasis in the original). Jackson asks whether Mary will actually learn anything upon her release.

Similarly with Searle. He denies that he would, as a matter of fact, understand Chinese in the Chinese Room TE: “In the Chinese case I have everything that artificial intelligence can put into me by way of a program, and I understand nothing.” To make it 20

clear that this is not a conceptual issue, he later writes: “The sense in which an automatic door "understands instructions" from its photoelectric cell is not at all the sense in which I understand English.” Searle makes a prediction: he predicts that artificial intelligence will 21

not produce ‘understanding of a language’ in the sense in which he understands English. It may be objected that the real philosophical problem with those TE’s is conceptual. The problem lies with the question of what it means when we say that Mary ‘learns

something new’, or what it means to ‘understand Chinese’. If this is the central issue of those TE’s, then they are evaluative, rather than predictive, TE’s. I concede the point; discussions about these TE’s can be, and have been, made into conceptual debates. Just as the same TE can have different uses, so the same TE can sometimes be used both for predictive and evaluative purposes. However, Jackson and Searle both presented their respective TE’s as predictions about what would happen in the hypothetical situations they describe. The claim that is meant to do the philosophical work that they want their TE’s to do in their respective articles is that Mary would in fact learn something new and that Searle would in fact not understand Chinese according to the definitions that they favour. These supposed facts are what’s supposed to falsify physicalism and Strong AI

respectively.

§1.3.2 Evaluative TE's

In contrast with predictive TE’s in evaluative TE’s, performing the experiment would not resolve the issue of interest. Even if we could really swap memories between cobblers and princes, that would not settle any issues about personal identity. Actually setting a trolley loose on innocent people would answer no moral questions about those scenario’s (although it may bring up some new ones).

What is more, whether or not the narrative of an evaluative TE is likely, plausible, or even possible is besides the point in evaluative TE’s. It doesn’t matter that Thomson’s trolley cases are extremely unlikely occurrences. It doesn’t matter that painted donkey’s and countrysides dotted with fake barns are incredibly implausible scenario’s. It doesn’t matter whether Parfit’s TE about the teletransportation-device is possible or not (not to mention his TE in which people split like amoeba). The issue we need to resolve in those 22

cases is how to describe or evaluate the described situation.

Searle (1982) p.418. 20

Idem. p.420 21

Although Kathleen Wilkes (1988) does object to these types of TE’s on this ground. 22

(18)

Evaluative TE’s, even though not unique to philosophy, are found in every domain of philosophy. They invite us to make a normative judgement about a hypothetical scenario (“No, these are not cases of knowledge.”, “Yes, you may turn the trolley”). Usually, in evaluative TE’s the audience does not need to be convinced that Q is the case. It is assumed that it will accept that. In evaluative TE’s we are asked to judge Q; Their work begins at Q.

Evaluative TE’s invite the skeptic to (feel compelled to) describe or evaluate Q in a certain way. Gettier may serve as an example here: The resulting ‘Q’ of Gettier cases is an odd case of justified, true belief (JTB). The question then revolves around how to call this JTB; Knowledge or no? Philosophically interesting is not that these cases can arise (they can), but rather how to deal with them.

As Gendler puts it: “In conceptual and valuational cases, we already know what would happen; we are worried about how we should describe or evaluative it. So the puzzle here is: what could possibly be guiding us in making such a judgement?” What 23

indeed? The answer to that question is of central importance to philosophical method; to the method of cases.

An odd aspect about verdicts about (evaluative) TE’s is that nearly always there seems to be an argumentative step missing. These verdicts are often assumed to be ‘common sense’ or intuitive without further arguments for their validity. For example, it seems self-evidently wrong to call a Gettier-case a case of knowledge. This feeling of self-evidence is usually called ‘an intuition’. Similarly, to many people it just seems wrong not to flip the switch in the trolley-case. These judgements, or ‘intuitions’, are often thought to serve as prima facie evidence for or against a philosophical position.

But this is puzzling. Are evaluative TE’s simply meant to elicit responses that feel self-evident? Could such a response ever be sufficient proof for or against a theory? Furthermore, the answer simply serves to push back Gendler’s question; for what guides our intuitions about a case? If we are to assess the value of our verdicts about evaluative TE’s, then we better find an answer to this question.

This is the central question of the next chapter: what guides these verdicts? How are they produced? I will look at several answers to this question, before proposing an account of intuition that understands intuitions as implicit cognitive processes that influence reasoning and judgement. This proposal would neatly explain both the role of intuition in evaluative TE’s, as well as in predictive TE’s. Moreover, it will give us a tool for evaluating the validity of our intuitive judgements, which is of central methodological importance to the method of cases. 

T.S. Gendler (2000), p.27. 23

(19)

Chapter 2

Intuition

In this chapter I first discuss different views on intuition as they are viewed by

philosophers. Since intuitions about TE’s are often (said to be) used as evidence for/ against theories or analyses, it matters what they are and how they are produced. After arguing against the idea that intuitions provide a priori, prima facie evidence for the content of what is being intuited in §2.2, I will present the most current view on ‘intuition’ in the cognitive sciences. That is the view that I will take up in order to explain the role of intuition in TE’s.

The debate about intuitions in philosophy revolves for an important part around their evidential status; Do they have any and, if so, why? Generally speaking, rationalists about intuitions maintain that intuitions have some a priori evidential value (Bealer, DuBonjour, Sosa, Ludwig, and also Descartes). George Bealer is a strong advocate of this position. His (1998) paper has served as a sort of stepping stone for other rationalists about intuition, which is why I give Bealer’s position considerable attention in §2.2. I will argue that there are good reasons to be skeptical about the a priori nature of intuition. The empiricist explanation of the phenomenon of intuition is therefore more promising.

Empiricists about intuitions seek to deflate their status and often identify ‘intuitions’ with ‘snap judgements’. Snap judgements are an everyday occurrence: most of our judgments are made without an awareness of the inferences that support them. Snap judgements are made spontaneously on the basis of previous experience which,

unfortunately, can function as a bias. As intuitions about TE’s are simply snap judgements about TE’s according to empiricists, they are likely to express biases and are therefore not deserving of the high regard that philosophers of a rationalist bent hold them in.

Although I will side with the empiricists, I will do so in a roundabout way. Literature on intuition in the cognitive sciences bases intuitions in antecedent tacit information-processing. The link between such processing and those intriguing mental states philosophers call ‘intuitions’ is not straightforward. Nevertheless, such a link would 24

explain the same phenomenon that rationalists seek to explain without having to resort to some special type of mental state. I intend to forge the link between antecedent tacit information-processing and TE’s in chapters 3 and 4. In chapter 3 I will argue that this type of processing can provide an answer to the problem of informativeness in many TE’s. In chapter 4 I intend to explain how antecedent tacit information-processing can account for ‘intuitions about cases’ as philosophers (even rationalists) understand them.

Osbeck (1999) argues that the philosophical concept of ‘intuition’ and the psychological concept of 24

(20)

§2.1 Conceptual issues surrounding ‘intuition’

It is often claimed that TE’s succeed (or not) because of our intuitions about them: By reading the Gettier-example we get the intuition that Smith has the justified true belief that ‘the person with ten coins in his pocket will get the job’ without knowing this. Reflecting on Galileo’s TE, we intuitively know that mass must be irrelevant to falling speed. Reading about the original position, we share Rawls’ intuition that people behind the veil of

ignorance will converge on the principles of justice. Contemplating the first trolley-case, we have the intuition that ‘Frank may turn the trolley’.

This way of talking about intuitions -roughly as salient cognitive states that can arise spontaneously in response to TE’s- is commonplace. It is, as I understand it, what

philosophers usually mean when they discuss ‘intuitions about possible cases’. These intuitions are highly interesting for anyone thinking about philosophical method since they have done quite a lot of heavy lifting in their respective fields.25

Yet it is far from clear what intuitions are, how they are produced, or what their evidential status is. Moreover, it is far from clear whether things that we call ‘intuitive’ (e.g. I find it intuitive that the switch for the overhead light is next to the entrance) are to be

identified with ‘intuitions about possible cases’ (e.g. the intuition that philosophical zombies are possible).26 27,

So what are intuitions? Are they, as David Lewis held, “simply opinions”? (1983a: x). Are they inclinations to form a belief or opinion? Besides these positions, there are generally speaking two schools of thought about what intuitions are. The rationalist school of thought holds that there is something special about them, and that this ‘specialness’ gives them some epistemic standing. George Bealer, for example, holds thats they are “sui generis propositional attitudes in which a proposition seems true to the intuiter.” In other words: 28

these attitudes cannot be reduced to normal judgements. They are something else. Although fallible, these seemings count as a type of prima facie evidence for the truth of the proposition. Reasons for their specialness vary. Descartes appealed to distinct phenomenology (clear and distinct ideas) as a way for recognising basic truths, and this way of understanding intuitions has its more nuanced adherents in modern times. 29

Just think about the ‘Gettier-intuition’, intuitions about Frankfurt-style cases, Twin-Earth intuitions, Locke’s 25

‘cobbler and the prince’ mindswap-intuition and the immense effect they have had on their respective fields. Examples taken from Bengston (2015), p.4 and p.6.

26

Herman van Cappelen argues in his (2012) book Philosophy without Intuitions that ‘intuition-talk’ is 27

hopelessly heterodox (see p.30-33 for a multitude of different examples). Even philosophers who seemingly rely on the notion of intuition are not clear on what they really mean.

George Bealer (1998: 207). This definition would exclude the notion of ‘intuitive’ from the overhead light 28

example.

See, for example, George Bealer (1998), M. Huemer (2005) and J.R. Brown (2011). 29

(21)

George Bealer (see §2.2 for a discussion of his views) and more recently Bengson (2015) are influential advocates of the phenomenological account.

According to some, intuitions’ specialness is justified because of their source: when intuitions are the expression of our conceptual competence (Bealer 1998, BonJour 1998, Sosa 2007, Ludwig 2010), or of ‘pure ratiocination’ (Goldman 2007, p.19-20), then their evidential status is warranted.

Empiricists hold that intuitions are spontaneous snap judgements, made in response to a situation, be it real or hypothetical. Importantly, though, these snap judgements are not particularly special and therefore do not enjoy any special epistemic status. Gopnik and Schwitzgebel (1998) define intuitive judgements as such:

We will call any judgement an intuitive judgement, or more briefly and intuition, just in case that judgement in not made on the basis of some kind of explicit reasoning process that a person can consciously observe.30

According to Gopnik and Schwitzgebel's account, intuitions do not differ from everyday (snap) judgements:

We make intuitive judgements about such things as the grammaticality of sentences, the morality of actions, the applicability of certain terms to certain situations, Bob’s likely reaction to an insult, the relative size of two distant objects, and so forth.31

The ‘snap judgement’ view on intuition has it that ‘intuitions about TE’s’ are spontaneous reactions that are not derived from conscious reasoning and are beyond our control: we cannot help but have these reactions. It would be overselling it to claim that philosophers simply rely on these snap judgements as brute evidential facts, though. We do not blindly 32

apply our first response to a TE. Rather, we reflect on that response, commit to some serious armchairing about it, and only then do we submit our thoughts about our intuitions to the scrutiny of our peers. On that, empiricists and rationalists about intuitions agree. Where they disagree is the idea that merely having an intuition about a case amounts to having evidence of some sort.

None the less, both schools of thought hold that intuitions intuitions arise spontaneously. Both schools of thought hold that intuitive judgement is arrived at without the ‘interference’ of explicit reasoning: they are baseless in the sense that they are not consciously formed on the base of any other states. Yet rationalists also hold that intuitive judgements are a category unto themselves which deserve a status as prima facie evidence for whatever it

Gopnik and Schwitzgebel (1998, p.77) 30

Idem.

31

Van Cappelen (2012, p.187) heavily criticises this characterisation of the use of TE's. 32

(22)

is that has been intuited, whereas empiricists do not distinguish them from everyday judgements. The main difference is thus in how they are valued.

One of the most thorough (rationalist) treatments of the subject is by Bengson (2015), who compares intuitions with presentational perceptual states. Bengson argues that in 33

addition to being contentful, intuitions have several properties that are relatively uncontroversial:

(1) They are baseless: They are not consciously formed on the base of any other states.

(2) They are gradable: Their quality may vary. Like perceptual states, the quality of intuitions is better when the intuiter is not distracted, has enough time to process information, is properly rested, etc.

(3) They are fundamentally non-voluntary: Unlike beliefs or opinions (and like perceptual states), intuitions immediately present themselves. We simply have them, even though we can choose to (or even be inclined to) disbelieve them. (4) They are compelling: they tend to dispose or incline assent to their content.

(5) These states do not merely dispose or incline assent; they also seem to rationalise such assent, in the (psychological) sense that they tend to make formation of corresponding beliefs seem rational or fitting from the first-person perspective.34

Both empiricists and rationalists about intuitions embrace (1) - (4). It isn’t very clear whether (5) is a property that is present when we make a judgement about, say, your standard-issue Frankfurt, or Gettier, or trolley case.

Yet (5) is there because the go-to examples of rationalists are such intuitions as: the intuition that for any predicate φ there is a set whose members satisfy φ, the intuition that p and not-p cannot both be true, and the intuition that Smith doesn’t know that ‘the person with ten coins in his pocket will get the job’ (in the Gettier-case). These are usually

compared to the Müller-Lyer illusion (See Bealer 1998, and Bengson 2015): despite

knowing that both lines are equally long, we do see a difference in length. The difference is visually present. The difference is that the Müller-Lyer illusion is a perceptual presentation, whereas intuitions are intellectual presentations. 

Bengson’s description of presentational states: “[Presentational states] do not simply represent the world 33

as being a certain way; in addition, they present the world as being that way. For instance, in having a visual experience in which it looks as if there is a red apple on the table, the world is not merely represented to one as being such that it is true that there is a red apple on the table. In addition, in having this experience, it is thereby presented to one that there is a red apple on the table. One has the impression that this is so (even if, as it turns out, this is not so).” p.10.

Bengson (2015) p.14-17. 34

(23)

FIGURE 1 The Müller-Lyer Illusion.

Importantly, the claim is not that such intuitions are infallible (The naive comprehension axiom is not true, because it can lead to a paradox; The Müller-Lyer illusion does not present the world as it is, because the two lines are equally long). What is odd about

intuitions is that they persist, even in the face of contrary evidence. We don’t suddenly stop being fooled by the Müller-Lyer illusion when we find out that it is just an illusion. Despite knowing about of Russell’s paradox, naive set theory still seems true.

The idea is that intuitions about TE’s can have the same qualities as the intuitions mentioned above and that, since those mentioned above are in good standing, intuitions about TE’s (that have the same properties) are also in good standing. Although I think that it is doubtful whether intuitions about the naive comprehension axiom (or the Müller-Lyer illusion) are the best templates for intuitions about TE’s, the claim is that our intuitions about TE’s can share the same characteristics and therefore can have the same validity. As my aim is to explore the role of intuitions in TE’s, and as at least (1) - (4) are accepted by all sides of the debate, it seems like a fair demand that the notion of intuition, as developed in the cognitive sciences, must accommodate at least those four properties. First though, I will discuss some of the shortcomings of the rationalist position. By breeding some skepticism about the rationalist account, I hope to create the room for a more

naturalistic account of intuition, based on tacit information processing, which I will briefly canvas in §2.3.

§2.2 Rational seemings or seemingly irrational?

In this section I will argue that George Bealer’s rationalist conception of intuition fails, even on its own terms. The arguments against Bealer should apply to several other rationalist conceptions of intuition in which either intuitions’ distinct phenomenology and/or their etiology have an important function. The purpose of problematising the rationalist notion is not to knock it down completely. By providing arguments for taking a skeptical stance towards rationalism about intuition, I mean to create room for a different, more naturalist explanation of intuition.

The reason for the failure of the rationalist view is twofold:

-

First, rationalists often use phenomenological criteria to distinguish ‘good’ intuitions from mere snap judgements: we can identify them by how they feel to us. I argue that

(24)

firstly this way of picking out valid intuitions actually undermines their evidential value because disagreement about them -either about their existence in general, or in concrete cases- is a real possibility and, secondly, that there are good reasons to doubt our ability to actually pick out ‘good’ from ‘bad’ intuitions and that we therefore should not rely on intuition as evidence even if a priori justified, rational intuitions really do exist.

-

Second, several rationalists justify intuitions’ role as evidence by referring to their etiology: intuitions are good evidence because they are the product of our conceptual competence. Since intuitions are baseless (i.e. not consciously formed on the basis of any other states), we have no way of introspectively knowing that this is the case. It is a mere assumption; An assumption that we should not accept.

George Bealer, in his (1998), defends a rather influential rationalist view of intuition as prima facie evidence for the truth of what is intuited. “Intuitions qualify as evidence, and the correct explanation of this fact is that intuitions have a strong (albeit indirect and fallible) tie to the truth when subjects are in suitably good cognitive conditions” The tie to the truth 35

comes from concept possession: “[I]t is constitutive of determinate concept possession that in suitably good cognitive conditions intuitions regarding the behavior of the concept have a strong tie to the truth”36

Bealer has a highly specific and restrictive notion of what counts as an intuition proper. That is also the basis for his complaint that psychologists have not studied the right, philosophical notion of intuition and that their research has therefore no bearing on his account of intuitions as evidence. Indeed, Bealer claims that there is a fundamental 37

difference between ‘rational intuitions’ on the one hand, and physical intuitions and “beliefs, hunches, guesses, judgements, common sense, and memory” on the other. 38

Whereas physical intuitions are about our expectations about what might happen in the physical world, rational intuitions are about the application of non-physical concepts. They are ‘sui generis, irreducible, natural propositional attitudes that occur episodically’. Only 39

rational intuitions count as evidence of what is intuited.

Bealer (1998), p.203 35

Idem. It is not clear whether all intuitions come from concept possession according to Bealer, since he also 36

discusses intuitions about logic. Bealer (1998), p.213-214 37

Idem, p.213 38

Idem. Bealer actually provides a taxonomy of different subcategories of rational intuitions such as 39

possibility intuitions, consistency intuitions, and concept application intuitions. However, since these are all rational intuitions and thus all have a) the phenomenology of being an intellectual seeming, and b) evidential status, their differences are not relevant here.

Imaginative Intuitions and Fanciful Fantasies: On the role of intuition in thought experiments

Contents

Introduction

Chapter 1

Fantasies and other experiments in thought

Chapter 2

Intuition

-

-