Skip to main content
Loading Explore guides…
Case Method · Acedra
Home
Explore
Philosophy
Case Method
Sign in to save
Chat
If YouTube asks you to sign in, open the video on YouTube.
Watch on YouTube
Reload player
Transcript
376 segments
Copy
Hide video
0:19
- I'll be talking today
0:23
about the case method in philosophy. This is a method that is part of philosophy dating back at least to the time of Plato, featured both in Western and in Eastern philosophical traditions. We think about particular scenarios, and we try to elicit some kind of judgment about them. We think of a situation and we ask ourselves, "Would this be the moral thing to do?" We describe a piece of human behavior and we ask, "Is this person free at this moment? "Is this a free choice?"
0:57
We describe a judgment and we ask ourselves whether it's belief or knowledge. And it's a very interesting method, and I think philosophy has recently, especially with the help of experimental philosophy, become quite self-conscious about the character of this method, about the nature of our activity as we reflect on particular cases, and this talk today is a contribution to that. I'm gonna be actually trying to get even a little bit more self conscious about the case method
1:30
and look at some of its pitfalls and some of its potentials. In particular I'm gonna be looking at this idea of stipulation: what goes on when I'm telling you the story before I even ask you the question? What's happening when I'm doing that? And I'm gonna try to argue that what happens when I tell you the story, something that is naturally perceived perhaps a little differently from the perspective of the experimenter who's designing the story to be told
1:57
and the naive participant who's listening to the story and maybe trying to figure out how she or he is going to respond to it. There's actually some systematic differences in the way the story looks to the experimenter who's using it as a tool and to the naive participant who is enjoying it as a narrative of some kind. So I'll give some sort of general worries
2:26
about the shift in perspective from the sophisticated experimenter to the naive participant and I'll run through a case study of some stories and some pairs of stories that motivated a position in epestomology known as interest-relative invariantism. Interest-relative invariantism originally came out as a traditional, armchair-driven approach in philosophy, and it got certain challenges from experimental philosophy. And at the end I'll try to extract some lessons,
2:59
both for experimental philosophy and for traditional philosophy. It's actually my view that what goes on in experimental philosophy is pretty similar to what goes on in traditional philosophy.
3:13
Some of the lessons that I'm extracting specifically for x-phi apply to our traditional friends as well.
3:22
First I'll just say something very brief about what I think is going on in the case method, just supposed to be a very rough characterization, nothing very philosophically controversial. I think it helps to start with a specific case. This is a case from the Indian philosopher Dharmottara
3:39
from about the year 770. And it's in his commentary on Dharmakirti's theory of knowledge. He actually goes through a series of possible definitions or analyses of knowledge, what is knowledge. The one he's talking about when he brings up this story is the identification of knowledge as the mental state that supports successful action. Who is the person who knows? He's the person who gets it right as far as his action is concerned in the world. And here's the story:
4:11
you have a thirsty desert traveler who's looking down into the valley ahead of him and hallucinates that there's an oasis there, and he forms the judgment there's water in the valley. As luck would have it, there is water in the valley, it's just in a well which is hidden under the sand.
4:31
And Dharmottara asks, "Does our traveler know that there is water in the valley "as he stands up at the crest of the hill?" And of course it's explained, yeah, on the strength of his hallucination let's say he does go down into the valley and then he finds the water, so that was a mental state which ends up supporting successful action for him. He wanted water and he got some. But did he know?
4:55
And Dharmottara feels that he didn't, and those of us who are familiar with the Gettier case tradition may recognize this as a Gettier case. Dharmottara actually has a number of fabulous ones. So here's a question: What is it that we're learning when we think about hypothetical cases like this? Very much at the object level Dharmottara would say you can learn, for example, that knowledge cannot simply be identified as the state of mind that supports successful action.
5:24
That's an unsuccessful definition of knowledge or analysis of knowledge. More generally, I think, there are various things you can learn. And here are two types of things you can learn which are sometimes somewhat difficult to disentangle from each other. On the one hand you can learn things about your intuitive capacities themselves. So you can learn by looking at how people respond to these stories, something about how people respond to stories,
5:52
how people intuitively attribute states of knowledge and mere belief to others.
5:59
What kinds of conditions prompt an intuitive classification as a situation as being one of knowledge or not? Then more boldly, when things are going well you can learn something about the targets of these intuitive capacities of ours. We seem to have an intuitive capacity to distinguish between knowing that something's the case and thinking that something's the case. That distinction is made in all known natural languages, and it's made very frequently. These are among our most common verbs,
6:28
it's made systematically. There is a systematic difference between describing someone as knowing something and as just thinking it. And I like to think that this intuitive capacity is not just random noise. I like to think that it's actually part of our social intelligence, that we're capable of recognizing when people know things and when they just think them to be the case. I don't think it's infallible. I think, as I'm sure all my epistemology and especially my x-phi friends would say
6:57
there are times when our intuitive capacities let us down, and we think that people know things and they don't actually know them. And there could actually be some systematic ways in which the capacity lets us down. There could be certain blind spots of this capacity, but I think on the whole, we are entitled to start with a relatively non-skeptical attitude to our intuitive capacities. Of course, if you started with a completely skeptical attitude, for example moral capacities, and you thought,
7:24
"Human moral judgments have no likelihood whatsoever "of leading you closer to the truth about good and evil,"
7:33
I think you would
7:38
be in a pretty difficult
7:42
epistemological predicament pretty quickly, and I don't think there's a lot of independent reason to support that kind of skeptical attitude. This is what I think might be going on with hypothetical cases, and something about the fact that they're intuitive has to be taken into account. I think the key characteristic of intuitive as opposed to reflective processing is that access to conscious working memory is not required for intuitive processing.
8:13
The classic example, if I ask you to multiply seven by 11 you're gonna get an answer spontaneously, involuntarily without awareness of any sequence of steps, any reasoning producing that answer. If you look at this the right way, you might recognize that as a picture of Marilyn Monroe. You don't actually know, and in fact I don't think you have introspective access to what the features of that array of dots are that would make you recognize it as her.
8:42
You recognize the emotions in other people's faces, whether they're sad or angry or afraid. You might actually even have theories about how you're doing that. Those theories are very likely wrong. It's not about the corners of the mouth, it's about weird stuff involving geometry of the eyebrows and so on. And so you're not necessarily aware of, in the first place, the processing. So multiplying seven by 11 is very different from dividing the number seven by 11.
9:09
If you try to go through that long division you'll be aware of an explicit series of stages. It won't be a particular mystery why you've ended up at the result that you get. Similarly, explicit syllogistic reasoning, where you have to be aware of the stages,
9:25
is less mysterious in character than intuitive processing where you find yourself with a response, as Steve Sloan puts it, "You're aware of the product but not the process." This is an interesting feature of intuitive judgment, and I wanna say in general, just the fact that you're not aware of the process doesn't mean that there's anything stupid about it. It can be very intelligent and very computationally rich. And intuition and this kind of immediate judgment
9:51
actually can tell us a lot about the world. But it tells us about it in ways that are not necessarily transparent to us. Thinking about how cases manage to elicit intuitions
10:06
we can introduce some vocabulary. We tell a story as a philosopher, as an experimentalist,
10:14
that we expect the reader to understand as involving certain basic features. So I might tell a story about a mental state that supports successful action. And I want to, as I tell the story of the man in the desert who goes down and gets the water, I want it to be relatively uncontroversial that his action was successful.
10:34
Or I describe a scenario in which there's some kind of random motion caused by a muscle spasm. And then after I tell you these stories I ask you a question about what I'll call some feature of interest. So I'll ask you: is this knowledge? Is this a free action?
10:53
And from the experimenter's perspective you have a division between basic features, which are gonna be the input to the relevant intuitive capacity that you're probing, and as an experimenter, you want to be telling the reader about those features, and features of interest, which are gonna be the features that are intuitively calculated by the reader. And as an experimenter you want to be asking the reader about those features. You don't want your story, of course,
11:20
to explicitly stipulate that this is a free action or there's no engagement
11:28
of the relative intuitive capacity here. This looks really simple, it's gonna turn out to be actually more complicated than it appears. And it's in part because, to the extent that we're looking at intuitive judgments to give us information about the target, it's typically because we haven't decoded the relevant, if you want, programming, of the relevant intuitive capacities yet. So we know that a lot of different features go into our recognition
11:56
of states of knowledge and belief in other people. In live action you're looking at things like direction of gaze. You can tell that I know that there's water in this bottle from the way I'm looking, from the sounds that you're hearing, from the things that I'm saying and doing. All of those are cues or input
12:16
to the capacity to attribute knowledge.
12:21
But until we know exactly how the capacity works we don't know. What are the basic features? What inputs are and aren't relevant? We don't know that yet. It's a little bit of a mystery to us exactly, first what knowledge itself is, and secondly how it's intuitively detected in human beings. So these capacities are somewhat mysterious, it's not just introspectable to us what we look at when we're doing things like detecting emotion or detecting knowledge. Those are kinda mysterious capacities,
12:52
so it can be tricky to
12:58
delimit and stipulate exactly what the basic features are in the story. Sometimes people are responding to story but not for the reasons you expect them to be responding to it. I wanted to talk a little bit about stories and two perspectives that you can take on them: the perspective of the subject and the perspective of the experimenter. So I wanna forget for a moment that we are philosophers who are probably very familiar with this case, especially as it was just, oh wait, sorry.
13:25
First I'll say, I'm going to talk about a very famous case but first I'll just say something about the way that stories are read. And I don't know if you can see this, it's in the public domain, but this is a train schedule, and this column right here indicates departure times.
13:43
Can anybody answer the question about whether there's a train departing Union Station at 8:05? Yes or no? (speaking from off mic) No, that's right. And how did you know that? Well, you knew that because you took what's called a closed-world reading of the train schedule. What you actually saw was an absence of information. That question: is there a train departing at 8:05? is different from the question: is there a train departing at 6:55? You would verify that by seeing it explicitly stipulated
14:13
in the schedule.
14:16
The train schedule does not have to explicitly stipulate that there are no trains departing at other times, you understand that to be the case because you take a closed-world reading of this material. You take all relevant truths to be explicitly stated, and sometimes people will even describe the closed-world assumption in very strong terms, like anything that's not explicitly on the list is false. Train schedules are read this way, but very very few other things are. Most of what we read,
14:43
we read in what's called an open-world format. So we bring our own background knowledge to bear, we fill in gaps. We don't infer the falsity of a claim from its absence, typically. I mean, read a novel or something like that, the protagonist takes an early morning train you do not infer that there are no other trains, you're reading open world on that, and in particular, your experimental participants can be expected to read things the way you read things that aren't train schedules,
15:12
the way you read most of the material you encounter in your life, which is in an open-world way. And here's the risk: as an experimenter, because you're very naturally focused on changes in explicit stipulation, especially when you're doing something like trying to set up a minimal pair of cases, there's just one little difference you're gonna tweak, it's very easy for you to focus on what's explicitly stipulated, and move a little bit in the direction of a closed-world reading of those texts.
15:44
So if I didn't tell you I was changing anything else, I haven't changed anything else, 'cause you're just supposed to focus on what's explicitly given in the the story. And that's maybe a little bit of a risk, we'll see some places where that risk might be a problem. Again, readers naturally tend to go open-world.
16:02
So when I say there's two perspectives on the case, I might schematize them something like this. From the experimenter's perspective, I've got a story to tell, which I take it has some kind of input of basic features, and so the basic features in a narrative, which elicits a judgment about knowing and thinking, the basic features might be descriptions of what the subject of the case is experiencing, saying doing and I'm gonna really focus on what it explicitly stipulated in my story,
16:38
and then I see that producing a judgment, and I'll make some kind of interpretation of the significance of that judgment, often looking pretty closely at the relationships between my explicit stipulations, and outcome of judgment. As a naive reader, when I read a story I'll appreciate the explicitly-stated basic features, plus also the intuitively-calculated ones,
17:03
which are the ones that you bring in when you're just trying to make sense of the story. We'll see something of that intuitive calculation in a minute. And again, I produce a judgment. But I'm not always conscious of the difference between what's been explicitly stated and what I'm intuitively calculating.
17:25
One of the things I'm going to try to argue for is the thesis that experimenters might be sometimes telling their participants more than they think they're telling them. Or less, in some situations. We're gonna very naturally focus on small changes in what's explicitly stipulated, but small changes in what's explicitly stipulated can generate larger changes in what is implicitly understood,
17:50
and that can be especially true in between subjects' designs. If you have juxtaposed cases where the reader gets to evaluate side-by-side,
18:02
and only one little change is made, they're gonna be more naturally cued to try to construe the stories as being the same except for that one little change. If you're just looking at a story on its own, some feature of it might take you in a direction quite different from the direction you would've gone if you'd looked at the other member of its so-called minimal pair on its own. So that's gonna be kind of a risk.
18:29
Another risk is that the intuitive capacity we're trying to investigate could be brought to bear just in making sense of the initial stipulations even before we've moved to the target judgment that the philosopher or experimenter is interested in. And lastly, you're not usually aware of the line between what's explicitly stated and what was generated intuitively. You can go back to the 1980s for this, Lynne Reder did some really excellent work where she gave people little stories,
19:01
they had to answer questions about the stories. These questions could be answered either on the basis of remembering verbatim some little phrase or sentence from the story, or reconstructing something that the story makes plausible.
19:17
And she found that people couldn't really tell which of these two strategies they were using. They're often not really conscious of whether they're getting something verbatim from the story or they're just expressing something that the story made plausible in some way. We're not usually conscious of that. You have a precise verbatim record of what was said,
19:39
but it's very fragile, and it degenerates almost immediately typically to preserve the gist and not the wording. And that's not a bad thing, it's just something to sometimes be aware of when you're interpreting your results. And Britt has a recent paper which is really great just on the fragility of this verbatim recall.
19:57
It's called something like What Did I Just Agree To? and it just goes into the strange effects
20:06
of almost immediate forgetting of passages
20:13
that you're supposed to be evaluating. You've forgotten them before you've even evaluated them half the time. Okay, I'm bringing up this example which Hamid also mentioned. Not because I think it's a bad example of anything, I think it's actually a classic example, terrific example of a Gettier case and a really useful one. I'm gonna ask you, even though you just heard about it earlier today, gonna ask you to just take a moment and read over the story just like you're a naive participant,
20:43
try to get the intuition about the case at the end.
20:59
Alright, so that's the story, and as this audience would know quite well, here's a standard takeaway from it: the standard takeaway is that Bob has a justified, true belief that Jill owns an American car, but he doesn't know that she does. He didn't know about that switcheroo with her car, he was just lucky that the car that she purchased happened to be also an American car, and the standard epistemologist thing to say about this story is that it shows that knowledge isn't equivalent
21:36
to justified, true belief. That's standard, right? We're good with that slide? Okay, so here's my question for you about that story: what kind of car does Jill now drive? Anyone? (speaking from off mic) Thank you. Here's the question I really care about: did the story stipulate that Jill now drives a Pontiac? Did the story stipulate that? (speaking from off mic) She bought one. (speaking from off mic) Her friend gave her one. Did the story stipulate that she now has a Pontiac?
22:06
That she owns a Pontiac? (speaking from off mic) She replaced it with a Pontiac. Actually that's not- - It was stolen. - No, the Buick was stolen. She got a Pontiac. Okay, sometimes it's interesting, because here's what the story says: the story doesn't directly declare that Jill is a Pontiac owner. It actually nests Jill's Pontiac ownership in a mental state attribution fog. Actually, an attribution of ignorance. Bob is also not aware that Jill has replaced it with a Pontiac. I wanna be clear,
22:42
the story actually does communicate to you exactly what you took from it: that Jill now owns a Pontiac, that's right, she does, but it did it in an interestingly, slightly indirect way.
22:54
It's gonna matter to us in a moment. Now, of course we want the story to convey that Jill's a Pontiac owner, right? Because it's just not gonna do that trick of showing this is JTB but not K, if it's really ambiguous in the story what is even true of Jill at the end. You want it to convey this, and we're gonna say yeah, by saying that Bob isn't aware that she replaced it with a Pontiac,
23:20
we are actually conveying that she did replace it with a Pontiac by saying that he's not aware of this. And why is that? It's because, in fact, well a very familiar point, know that, realize that, is aware that, those are all factive constructions, right? So they presuppose the truth of the embedded clause, and typically, if you have ordinary sentential negation presupposition is preserved under that. So if I say, "Mary didn't know that her ex-husband was stalking her,"
23:56
I'm communicating to you something about Mary's ignorance, and I'm simultaneously communicating to you something about Mary's ex-husband. Or likewise here, he wasn't aware that she'd replaced her stolen Buick with a Pontiac. So I'm communicating this to you. But here's some kind of interesting things: it's not strictly entailed.
24:20
This frame of words doesn't strictly entail the truth of the complement clause. Because it's possible to take this sentence here, which I said, if you just have this sort of unelaborated, Mary didn't know that her ex-husband was stalking her, I'd be telling you her ex-husband's stalking her. But I could on and say, because in fact he wasn't, now here I'm canceling presupposition, "Mary was experiencing paranoid delusions." I could elaborate it this way, and this is something that's going to show
24:49
you actually didn't directly get stipulated that Mary had a Pontiac.
24:56
I'm forgetting everybody's names. Jill's the one with the Pontiac. I didn't directly stipulate that, or Weinberg, Nichols, and Stich, whose very good vignette that originally was, they didn't directly stipulate that, but you figured it out. And you weren't aware that you had inferred it, I mean, typically we're not, we're not aware of the difference between what we're told verbatim and conversational implecature, even though I gave you every possible cue to like,
25:16
"Open your eyes, I'm gonna trick you." Alright, so that's kind of interesting.
25:23
And the interesting questions is: why would we use this kind of construction? If typically "Mary didn't know" would carry that implecature that her ex-husband is stalking her, typically if somebody's paranoid and they've got this crazy idea that their ex-husband is stalking them, you'd deliver that thought by saying something like, "For Mary thought that her ex-husband was stalking her." That's typically what I'd say. I typically wouldn't go for this more elaborate way of speaking,
25:53
but there are contexts in which the more elaborate way of speaking is precisely what you want. Look at this interchange between John and Paul. John says, "And then Mary wouldn't go out after dark, "because she knew her ex-husband was stalking her." And Paul says, "Mary didn't know that her ex was stalking her, he wasn't. "She was paranoid." Paul's use of the verb "to know" there is what's called echoic use. He's picking it up from John, he's picking up John's presupposition,
26:19
but this might be metalinguistic negation, he's saying that's not quite the right vocabulary item there. But if you don't have any special reason to expect metalinguistic negation, yeah, no one ever expects metalinguistic negation (laughter) then you're not gonna interpret the sentence that way, and we didn't. But we can create problems for this.
26:45
I'll just say a couple of quick things before I say what the problem is. Okay typically
26:54
presupposition is what's in the common ground between the two people who are speaking to each other. So if...
27:07
If Tom says to Bob, "Jane doesn't know that the meeting was canceled," typically that would be in a situation where it's common knowledge between Tom and Bob that the meeting's been canceled, and the point of the utterance is an utterance about Jane's ignorance. So that's already in the common ground. So this might kinda make you think, "Wait a minute. "How are we stipulating things "using these negative factives, "how can we be bringing new information to the table?" Like what?
27:34
It's supposed to be common ground between the experimenter and their naive participants that Jill has recently bought a Pontiac? Wasn't that part of the news that you were sharing by using the verb?
27:47
And in fact, if you look at how factives are actually used, they're used to bring in fresh information to the common ground all the time. I don't know if you guys know Jennifer Spenader's work, I hope I'm pronouncing her name correctly, she's in Sweden, she actually did some really exhaustive empirical research on how people use factives in vivo.
28:12
So she's working with the London-Lund Corpus for this particular paper
28:20
on factive presuppositions from 2003, and the London-Lund Corpus is a massive database of spoken English
28:29
from all kinds of sources, some of which seem dubiously ethical like recorded telephone conversations. I don't know quite what went on with that. (laughter) And she actually did a sort of exhausting task of going through these conversations looking at when people used factives, and trying to classify: does this information seem to be new? Is it being introduced to the common ground? Is it something that both participants already knew about? Or is it kinda hard to tell? And she found actually,
28:57
well her study I think, 58% of the time the information that's being brought in is something that's new coming onto the table for the first time. Maybe 20% of the time it's clear that both people knew it already, and the rest breaks down into sometimes it's not clear, might be old might be new. But she has all these great spontaneously-recorded conversations of people using factive I really like this one, I feel like I have to read this out. Speaker A said,
29:25
"When I worked on the railways these many years ago, "I was working in the claims department "at Pretona Station Warmington "as office boy for a short time, "and one noticed that the tremendous nnmber of claims "against the railway companies "were people whose fingers had been caught in doors "as the porters had slammed them." And Speaker B says, "Really? "Oh my goodness." And so it's pretty clear that this embedded statement here is not something that Speaker B knew already,
29:50
but there you are using a factive
29:55
to bring it in, to bring it to the table. So just because something's presupposed doesn't mean that it can't be brought in as new information. You accommodate with your conversational partners. And actually, I think this method of going to the corpus data and looking at how people are actually using these words is something that experimentalists could make much more use of. We talk a lot about usage,
30:19
I think looking at usage in the wild can be very informative.
30:26
Okay, one of the things I wanna say now is the line between input and output on a case, "Here are the things I'm telling you, "now what are the things you're gonna tell me back?" That line shifts depending on what I'm interested in as an experimenter. Unfortunately, the line also shifts for the reader depending on what they're being explicitly told and what they're being asked. So when we posted a Gettier case, we all took for granted that Jill drives a Pontiac. In fact, you even thought
30:55
that I'd stipulated that explicitly. But we could use the same sentence, actually, to test whether readers understand "knows that" to be factive. And some people have started trying to test this. Like, wait a minute. Is "knows that" really factive or not?
31:13
But there's some difficulty with this. Of course, you have to ask a different question at the end if you're asking basically whether the presupposed,
31:24
the embedded statement is accepted as true. Asking that question itself can change how the reader understands a text, however, and you can get some sort of strange effects from that, that people reinterpret the whole exchange
31:42
in a ways that kinda undermines the ordinary function of the factive. So as I said before, if you have an ordinary usage of a factive verb,
31:54
this is either common knowledge between the participants at the get-go, or if Bob didn't already know that the meeting's been rescheduled, he knows it now because he's gonna accommodate to understand that utterance. It's either existing common ground or being added to common ground, it's not controversial. But what if I had a factive statement like this? Susan knows that Tom was promoted. Followed immediately by the question: well was he? Was he really? Was he promoted?
32:20
What I'm doing just by asking that question is I'm threatening to take that out of common ground. And if I do that, I can get you to sort of reinterpret what I meant by that initial use of a factive verb. So the flat, literal reading, which is kind of what you hope your participants are going to do
32:41
when you start giving them experimental surveys on factive verbs, is to just respond "Yes, Tom was promoted. "You said that he knows that she knows that he's promoted. "Why are you even asking me that?" But that's the problem. "Why are you even asking me that?" That's a pragmatic problem in how this exchange is being understood. If you get clever on this and you take the experimenter, I mean, "Here I am, I'm in an experiment, "this person can't be asking me this stupid question,"
33:04
the question becomes non-trivial, and then you can shift to a projected reading of "knows." "Oh yeah, Susan knows alright," where you're really meaning to convey something like it seems to Susan that she knows. So it can actually be experimentally difficult,
33:25
difficult to calculate how people are understanding stories because sometimes the simplest things you want people to take away from the story, if you start questioning them about those elements, they're gonna stop taking them away. So that can be a little bit tricky. I'm not saying it's an insurmountable problem, there's easy ways to deal with it. Actually no, there's sometimes quite difficult ways to deal with it, but we've gotta be very conscious of the pragmatics of our question-asking
33:49
as we work with these scenarios. That's actually something that psychologists have been aware of for a very long time. I encourage you all to go to the dustiest corners of your library, or email me and I can email this to you, and dig up Bernard Muscio's 1916 British Journal of Psychology paper on the influence of the form of a question. I don't know if you guys have read this paper, it's magnificent. This is from the dawn of what was then known as the moving picture, and he showed a series
34:22
of roughly 30-second moving pictures to his experimental participants, and he'd just show little interactions: two guys in a room talking, an outdoor scene down by Marble Arch where there's some traffic going by, and then after they saw the clip, and they would even see it several times, he would ask different kinds of questions. He might ask: "Was there a gun on the table? "Did you see the gun on the table? "Was the gun on the table a revolver or a pistol?"
34:48
You can think of the kinds of questions you could ask. He was just fooling with, basically, the form of the question. These are shaded here, and the darker it gets, the more leading the question. So he scored all these questions on various axes of caution, suggestiveness, and I forget the third one. Maybe reliability or something like that. Basically looking at what kinds of questions can you ask people in order to get them to report that they saw the nurse on the bicycle in the story
35:19
when in fact there was no nurse, there wasn't even a bicycle. But if you ask the question just right, people will respond to it. And with these questions, people had the option of saying "I don't know, I didn't see it," or of giving a substantive answer of one kind or another. So the question you ask is always part of the story you are telling. And questions can presuppose things, questions can suggest things, in ways that, as experimentalists, we've got to be very conscious of.
35:51
And again on the pragmatics, from the perspective of the experimenter I don't mean to be doing anything funny with my question, but I might inadvertently be doing that.
36:02
Okay, so I'm suggesting there can be a shift in perspective, maybe something having to do with the pragmatics between the experimenter and the participant. I wanna actually work through a particular case study of some literature which I think has maybe gone down the wrong path here. This is the literature on stake sensitivity in knowledge description, and I'm actually gonna start with some psychological background which is not directly on point. The psychological background
36:33
has to do with adaptive decision making. As human beings, we do not invest the same time and energy in every single decision that we make. Thank god for that. If you're making a really high-stakes decision like which graduate school to go to, you've been admitted to several, you're weighing them against each other, you're going to draw up pros and cons and agonize about it, and think about it for a while. If it's just chicken versus tuna saald at lunch you might not spend that much,
36:59
you're gonna be hungry if you spend that much time. And there's just a absolute huge body of literature in psychology and behavioral econometrics on this,
37:13
and I'm not gonna talk about this a lot. But there's a lot of very very detailed work and some fabulous mathematical models.
37:24
It's pretty easy to check that there is this in fact going on. You can have situations where people have to, for example, purchase bits of evidence. Each bit of evidence costs, I mean, a lot of this work is done in business school so they actually do it with cash. Each bit of evidence that you pick, that you select before you answer the question costs you 50 cents. And you can pick as much or as little evidence as you want, and then you have to give your answers.
37:52
Maybe you're under time limits, maybe you're not, and then we'll just manipulate how big the cash reward is
37:59
for getting this particular question right. Extremely unsurprisingly, people purchase much more information for high-value questions. And there are all kinds of models about how you do that and why you do that, and of course in real life we do this too. If we're under time pressure we behave differently. But in general, our thinking is adaptive. That is, decision-making behavior is like all other behavior, right? It's something that we do, in part, with an eye on costs and benefits.
38:28
Costs and increased evidence collection and evidence weighing can be offset by gains in accuracy
38:38
for important questions. Now, it's not true that in general people who are high-stakes decision makers always collect more evidence and think harder. Sometimes they're not willing to, and sometimes they're not able to. For example, sometimes they might be extremely biased towards one answer rather than another.
38:57
They're in a situation involving wishful thinking, prejudice, something like that. Or they might be unable to because they're distracted or they're under time pressure. And again, there's a huge amount of empirical literature on that. This is all very not-surprising, right? Okay, I think there's actually a relationship between that literature and epistemological cases like these
39:18
which will be familiar to those of you who are in epistemology. There may be a shift between how we intuitively evaluate the mental states of people who are in very low-stakes, casual decision-making contexts and people who are in very high-stakes, life-or-death situations. This is a story which originally traces back to Keith DeRose but this is Jason Stanley's telling of it, about a couple of people who are at a bank. And in the low-stakes version of the story, it's a pretty casual matter
39:46
but they decide not to stop in at the bank on Friday.
39:53
Hannah's pretty confident that the bank will be open tomorrow, and here's her explicit evidence. "Since I was there just two weeks ago on Saturday morning, "so we can deposit our paychecks tomorrow morning." I have to say this is a piece of historical fiction for us now from before the days of direct deposit and also from, at least when I run this, I can't run this in Canada verbatim because all banks are open on Saturday. I don't know, how is it here? Is it hit-and-miss, 50/50?
40:20
Anyway, 50/50 would be good because that would make it a real interesting question. And the idea is you're supposed to think, "Yeah, she probably did this, "she probably does know." In this case, danger danger, they have an impending bill coming due, very little in their account, it's absolutely crucial that they deposit their cheque by Friday, they're outside the bank, and again this explicit evidence that's given for the decision is the same in both cases.
40:46
Hannah notes that she was at the bank a couple weeks ago, it was open, and then she reaches this judgment that she lacks knowledge that the bank will be open tomorrow and we're supposed to endorse that, we're supposed to say, "Yeah, that's right. "In this case she knows, in that case she doesn't know."
41:04
So that's the sort of standard philosopher's response to that, it's intuitive to see low-stakes as knowing and high-stakes as failing to know, despite the fact that they have the same explicitly-stipulated evidence about the bank. And that's supposed to be philosophically interesting or exciting. These guys, Jason Stanley,
41:25
perhaps at some point in his career John Hawthorne, although not anymore, Jeremy Fantl and Matthew McGrath, thought that these cases showed that stakes themselves are a factor in knowledge. Your possession or failure to possess knowledge depends in part on what is practically at stake for you in the decision. And that's supposed to be a surprise, because you would've thought knowledge depends strictly on the quality of your evidence and your thinking and whether the proposition
41:53
that you believe is actually true or not. You might think it just relies on that, "No," these guys say, "it also depends." How much money is in your bank account could make a difference to whether you do or don't have knowledge that the bank's gonna be open tomorrow. This is a radical and unusual position because it erodes the traditional distinction between practical and theoretical rationality. And it actually ends up having a lot of philosophical costs because of that.
42:22
And so in response to that,
42:27
a number of other, traditional philosophers who wanted to take what's called an intellectualist approach in which you deny that practical stakes matter directly to knowledge, the intellectualists agreed that they felt those intuitions. And I think this is something really important. It's really important when philosophers who are in opposed camps share common intuitions. They have some common basis of material that they can talk about. They felt the intuitions,
42:53
but they thought there's some kind of confusion, maybe we're just driven by conversational pragmatics here. So Williamson has the idea maybe in high-stakes situations it's not enough to know, you actually also have to know that you know. If this is a life-or-death matter, you don't just have to know that you've chosen the right graduate school, you have to know that you know that you've chosen. Probably more iterations than that.
43:16
And so then they'd say, "Well maybe we're just getting kind of confused "when we said this person doesn't know, "they didn't know that they know."
43:25
And these accounts, I think are interesting. They actually might have part of the truth in them but I think there's actually a simpler explanation of what's going on in these cases which I'll get to in a moment. So this is how the interest-relative invariantists like Jason Stanley, and the error theories are seeing the cases, they're seeing the cases as having these basic features: there's a moderate amount of evidence in both of them, and shifting from low to high stakes,
43:50
and the target feature, the feature of interest, is that intuitively, our low-stakes person knows and our high-stakes person doesn't. And this is either a mistake, which is what the errorists think, or it's a demonstration that stakes themselves matter to knowledge, which is what the interest-relative invariantists think. I think there's a simpler answer. I think if you look at this case the way it was originally stated, and you have in your back pocket the background
44:19
that human decision-making is generally adaptive,
44:25
we expect high-stakes people to think longer and harder before they make up their minds. Then High Stakes Hannah over here hasn't made up her mind yet, and actually in this formulation of the case, it's really obvious that she hasn't made up her mind yet. She's saying, "I guess you're right, "I don't know that the bank will be open tomorrow." In Keith DeRose's original telling of the story, the protagonist gets out of the car and goes to check, "I guess you're right honey,"
44:54
he gets out of the car and goes into the bank to double-check the hours. At the moment that he's getting out of the car, which is the moment where we're invited to assess his mental state, it doesn't look like he has a conviction that the bank is open the next day. It looks like he's in that state, that fluid state of weighing evidence and trying to make up his mind. He's not sure, he's going to find out.
45:17
He doesn't have belief one way or the other. So lack of belief, like you haven't made up your mind yet, that's a very traditional way to deny knowledge. You don't have to get all fancy pants and go for interest-relative invariances. Okay, probably because of that kind of criticism, people decided to start talking about versions of the case in which it was much more explicitly stipulated that our high-stakes character, despite her burning-high stakes and her skimpy evidence,
45:45
was making up her mind, "I'm confident."
45:50
But the problem here is that if you stipulate that somebody has burning-high stakes and skimpy evidence, and she's going ahead and just making up her mind anyway, you're essentially telling a story in which someone's coming across as being hasty or distracted, because adaptive thinkers don't usually make up their minds rapidly on moderate evidence. But if you're understanding the story that way, then again you've got a really traditional reason to deny knowledge in this case,
46:17
you don't have to go for interest-relative invariantism. So in the bank cases, the way I see it,
46:27
we're responding not just to the explicit stipulations, we're also responding to what we infer about the cases, thanks to our background knowledge of the adaptivity of human judgment. So in the original cases,
46:44
the way, for example, Jason Stanley just had them there, I think we'd ordinarily read those cases as involving high confidence in the first case, low confidence in the second, and that's something that you infer from what's going on, and that's a very traditional reason for you to shift from saying knowledge to absence of knowledge. In the amended cases, where you have stipulation of explicitly confidence in the case, then you have the same moderate evidence
47:15
and high confidence, it's easy just to focus on the switch in stakes between the cases. But if we can think about what you're inferring from the case, when you read it in your ordinary, open-book way, as well as what's explicitly stipulated, you're gonna be inferring that there's an ordinary level of human reliability in the first case, and some kind of compromised reliability in the second. And again, that's a very traditional reason for the shift. So changing stakes means
47:46
you're gonna be changing inferred traditional factors like confidence and reliability. I don't think you have to go for pragmatism or interest-relative invariantism, and these relationships are very well-known and they should be naturally reflected in our intuitive mental state attributions on any of our dominant theories of mental state attributions. And there's actually some relatively direct evidence for that going back to the '70s I can talk about it you want. Now, here's the shocker.
48:14
You'd expect, then, that since philosophers found it intuitive,
48:21
and I'm saying "Oh, there's all these reasons "you would read these cases this way," you would expect that folk intuitions, when you actually study them experimentally, would reflect changing stakes.
48:34
And here's the very surprising and puzzling thing: a whole series of very well-conducted empirical tests found only very weak effects. If you actually give people the verbatim Jason Stanley cases
48:49
you don't get much difference between them when you ask, "Does this high-stakes person know? "Does this low-stakes person know?" They actually tend to say yes to both, we don't tend to go from, "Oh well low-stakes knows but high-stakes doesn't."
49:03
And this was brought forward as the suggestion that armchair philosophy is in a lot of trouble, because a lot of armchair philosophy, both from people who believed in the intuitions and from people who were out to debunk them took those intuitions for granted, and assumed that we shared them with the folk. Something funny is going on. Let's try and see what it is.
49:23
When philosophers first read the bank cases, they were reading them as juxtaposed: "There's these two cases. "Here's the only thing that's different. "We're shifting from high to low stakes," trying to interpret the cases as being as similar as you can, being a good sport reading the philosophy article. And in fact, if you do juxtapose cases with laypeople, you actually do get something of an effect. What about between subjects' designs? And I'm gonna focus here specifically
49:50
on ones where, at the end of the story, the agent's level of confidence or belief is explicitly stipulated. Will they try to make sense of the story? This is a person who's in a really high-stakes situation and they're saying the bank is open. One way to make sense of this story is to think that the person actually does know a fair bit about whether the bank would be open, right? It's not in general the case that high-stakes thinkers know less about stuff than low-stakes thinkers do.
50:20
It's just if you held the evidence absolutely fixed between high and low-stakes thinkers, that you'd get that result, that you'd get the low-stakes people make up their mind, you give 'em a couple of cues and they're good. The high-stakes people are not satisfied with a couple of cues, they wanna go on searching. But in general, high-stakes thinkers, because they're in general at liberty to pursue more evidence and think harder about issues, are as likely or better to know about things
50:46
as low-stakes thinkers are. So here's the between-subjects design that Buckwalter pursued.
50:58
And it's kind of like that Jason Stanley case.
51:03
Again, the bolded text marks the change in explicit stipulations from low-stakes to high-stakes.
51:13
And he found most people saying yeah in both of these cases. Bruno knows that the bank will be open on Saturday. But the easiest way to explain this is when you're reading the story, you're trying to make sense of Bruno's entire package of behavior. And so he's a high-stakes guy and he says, "Look honey, I know it's open," and he's saying, "I was just here last Saturday, "and it's open 'til noon," one way you can make sense of what he's doing is to see him as knowing, right?
51:44
And actually, for what it's worth, in all these cases there was, as far as I can remember, there was actually a little bit of an effect, it just didn't achieve statistical significance. I believe, actually,
51:59
lemme see, Feltz and Zarpentine found statistical significance when they collapsed all their data together. But a pretty weak effect. And people tended to get results that were numerically in conformity with the standard philosophical intuition, but not in a way that was statistically significant. You can actually also get other effects, so Feltz and Zarpentine I think have another effect going on here which is there's a little bit of a confusion.
52:25
Actually, here again you see this is a low-stakes case, somebody's about to,
52:33
oh sorry, I put high and low on the wrong side. This is the low-stakes case over here, somebody's about to drive over a little ditch, this is a very high-stakes case, someone's gonna drive over a canyon,
52:49
and the scale's inverted so that's kinda confusing but...
52:57
people were asked whether they agreed or disagreed and you have 27% of people inclined to disagree
53:07
that it was knowledge in the low-stakes and a high of 36, but not statistically significantly higher percentage of participants disagreeing on the high-stakes case. One of the problems they have here is actually something kinda funny. I think there may be some confusion here with, one of the things I'll say here is it's really open-ended exactly how good the evidence is in these two cases. In both cases there's a phone call, but you're gonna try to read the story in a way that,
53:38
this guy's about to drive his truck across the bridge, gonna try to read the story in a way that acommodates that. So you're relatively likely to,
53:53
actually most people gave a kind of middling response on these. Doesn't seem outrageous in either case that he could have knowledge. There's a little bit of a shift, but not a statistically significant one. There's another problem down here though, which is that participants are explicitly told assume that the bridge is safe enough for him to cross. Please indicate how much you agree with the following statement: When John thinks to himself, "I know that my truck will make it across the bridge,"
54:17
what he thinks is true. There's an ambiguity in "what he thinks" and what that refers to. Is it the embedded "my truck will make it across the bridge"? Or is it the self-attribution of knowledge. I'm guessing a lot of participants read it as the self-attribution of knowledge there.
54:38
But just in ordinary speech, the "what he thinks" could pick out either. So if you imagine, I dunno, Tea Party Charlie saying, "I think that Barack Obama was born in Kenya,"
54:50
and I ask you, "When Tea Party Charlie says to himself "'I think that Barack Obama was born in Charlie,' "is what he thinks true?"
55:03
You might take that as a question of: well how good is Tea Party Charlie at introspecting his own thoughts? Or you might take it as a question about the birthplace of Barack Obama. And so that's another sort of hazard in the way the question's posed. If you're careful to hold the evidence fixed you actually can detect the effect. And so Angel Pinillos has done some work and I know that actually CUNY's own Wesley Buckwalter has done some follow-up work on this
55:36
which in some ways seems to be undermining of Angel Pinillos's conclusion, but I don't think it is, ultimately, and I can talk about that in question period if you're curious about it.
55:49
So in Pinillos's study, he actually looked directly at evidence collection and looked at how much evidence we anticipate people to collect before we're willing to call them knowers, and he had cases where somebody has to count the number of pennies in a jar, either for a trivial reward, or for a massive reward of money that's going to, I believe, give his mother a life-saving cancer operation or something like that. Maybe it wasn't something that big, but it was big,
56:15
and so people thought that Peter would have to count the coins more frequently if it was a very high-stakes contest that he was in,
56:25
and that just makes perfect sense, right, that we expect people's evidence acquisition behavior to be adaptive. So that's a reflection of that effect.
56:35
So you do see some kind of stake-sensitive effects if you're careful to hold the evidence fixed. And one way you can hold the evidence fixed is by giving people, for example, juxtaposed cases. So I'm saying that intuitive capacities respond to basic features whether those features are explicitly stated or inferred, and we can't, obviously,
57:01
we can easily, as experimenters, see which basic features we've explicitly stipulated. That's just very easy for us to do. It's harder for us to see or introspect which basic features are getting inferred. Okay, lessons for traditional and experimental philosophy. I see that I've actually gone past, I'm running late, so this is gonna be really quick. The Knobe effect presents you with an exquisitely clean minimal pair. (speaking from off mic) Oh, okay. So one of the very very beautiful things
57:34
about the Knobe effect is you're just switching one word. Actually you switch it a few times, but that looks like a perfect minimal pair. You're changing the moral valence
57:49
of a side effect, you're switching help to harm,
57:56
(speaking from off mic) (laughter) Oh well, very good that's important too. Thank you, thank you. I didn't even see that, I should've highlighted it.
58:07
So that looks like a really really clean little pair, and it's very easy for us to focus on that,
58:18
oh shoot, what was I gonna say? Sorry.
58:22
And I think there's been a wave of work lately which has been suggesting that we're probably shifting a lot more
58:33
than just the valence. In particular, Alfano, Robinson, and Beebe have been arguing that,
58:43
picking up on some suggestions by earlier writers like (mumbles) and Brozo, that the original idea behind the Knobe effect was that the mental states of the chairmen of the board are just absolutely the same between help and harm except for the moral valence of the side effect.
59:04
But some critics have argued that actually, changing the moral valence of the side effect forces you to change quite a bit of your representation of the mental states of the chairman of the board. And interestingly, it's been suggested that when we read this story
59:30
we might not actually even form the belief, sorry, might not even attribute to the chairman of the board the belief that what he's doing will actually help the environment. Here he is saying, "I don't care at all about helping the environment." And when we read this side of the story,
59:51
Alfano et al. have suggested that we actually do, when we read this, attribute to the chairman of the board the belief that what he's about to do is going to harm the environment. And if it's a condition on representing intentional action, that you believe that the thing that you're about to do should have that effect, if it'll bring it about, then that would be part of the explanation of what's going on here.
1:00:18
So the idea is, when you read the story, to make sense of it, you have to infer some quite different beliefs
1:00:27
for the chairman of the board in these two scenarios, and beliefs that are very directly relevant, and relevant in a very old-fashioned way to whether the action should be represented as intentional. Now, one of the things that's kind of interesting about that Alfano,
1:00:45
and I won't say this is decisive, I'm just sorta saying this is an interesting thing that James Beebe and Mark Alfano and company have been looking into. And they've just been asking people, basically, "Does this person believe?" and finding a really significant difference between those conditions.
1:01:02
It's interesting to me, in part, because the Knobe effect is so incredibly robust, and it's been produced with such a broad variety of materials,
1:01:14
some of which actually look like they get really close to stipulating the very thing that
1:01:22
Alfano, Beebe, Robinson think is shifting in these cases.
1:01:29
So those guys are saying there's a difference between whether you're actually attributing to the person the belief or not. This is from Knobe 2007, the stimuli here are from Knobe 2007 paper on clashing norms
1:01:45
and this is a norm which we recognize as morally dreadful.
1:01:52
So it's a case in which norm violation would be the good thing to do, and you can still produce a version of the Knobe effect
1:02:02
where the norm violation comes across as a...
1:02:09
oh sorry, this should be compliance rather than violation,
1:02:15
that's a mistake. The norm violation comes across as intentional and the compliance to the norm comes across as unintentional even though we've now crossed the moral valence of the stimuli. And incidentally I should have presented this a bit more clearly: the reason why Alfano and Beebe think you're attributing these mental states is that when you're about to violate a norm, you're essentially in a high-stakes situation. That's true back in this case as well. If you are in a situation
1:02:45
where you're about to violate the norm of taking care of the environment, you're about to harm the environment, you've got a sort of pre-standing reason not to do this, and so you'd be expected to think harder about it and actually make up your mind, whereas on this case, actually, it doesn't even matter to this person whether or not the environment is helped, and there's no background reason to avoid doing things that help the environment, so you don't even have to really
1:03:12
make up your mind about this. In this case we had somebody saying, "Look, I know I'll be fulfilling "the requirements of the law," this person is saying, "Look, I know I'll be violating "the requirements of the law." On the Alfano and Beebe interpretation, we actually read this compliance person as not knowing or not having made up his mind, even though he's saying "I know that." And I think this is an illustration, if the Alfano explanation is right,
1:03:36
I don't think it's killed by this pair of cases because it's an illustration of the fact that when you use "I know," you're often just using that conversationally. This isn't typically taken as a direction that a mental state attribution is to be made. For example, if someone says to you on the phone from the customer support line, "I know that customer service is very important to you," you don't read this as a statement that, "Oh that person knows that customer service is important."
1:04:08
That's just a conversational use, and so this is something that we discount. So this is getting very close to a case where you're explicitly stipulating something that people discount because of the content. And so I think that's an interesting sort of thing to watch out for. Sorry, my battery's really fading. I wanna say, this is not a problem specific to x-phi it's a problem for armchair philosophers as well. I wanna say when armchair philosophers like Keith DeRose
1:04:30
started fooling around with the bank cases they might not have known what kinds of changes they were invoking in people when they shifted from high to low stakes.
1:04:45
And I think this is because the operations of our intuitive capacities are relatively mysterious to us, whether we're traditional philosophers or whether we are experimentalists.
1:04:57
You can't just introspect what your intuitive capacities are doing any more than you can introspect how, for example, you're able to accomplish face recognition. You can have some sort of crude efforts to detect what you're inferring from what's being stipulated, but the problem is that you've gotta be careful with that because those crude efforts to check, "Are you inferring this?" can really backfire. If I ask you, "Are you inferring this from the case?" you can have the sense,
1:05:23
"Maybe I shouldn't be." There's various things you can do to try to help yourself out of these problems, and I think it's really good to be aware of the background psychological research on story comprehension, discourse representation, and on our target intuitive capacities. And I think, actually, another thing that's very good is to work on traditional philosophical theory formulation. So to try to approach the whole problem from the other side as well, and ask yourself questions like,
1:05:53
"Is the interest-relative invariantist theory of knowledge, "is that viable as a theory of knowledge?" and you can look at considerations of theoretical simplicity and unification to help guide you from that side as well. So I think there's a lot that philosophers can learn from x-phi, and I think there's a lot that experimentalists can learn, both from traditional philosophy and from the relevant branches of psychology. Thank you. (applause)
1:06:24
- [Carisa] Thanks, that was terrific. It's a great way to end the conference. We're gonna have a period of Q and A now. Questions for Jennifer?
1:06:34
- Yeah, I wondered about the bank cases and going back to traditional philosophy. So you looked at this as a difference in the confidence levels.
1:06:48
I wonder though, what you would say though in looking at Colin Radford stuff from the 1960s, right? So there's some- - Oh yeah, I love that stuff. - Yeah, so there's, I mean the case is similar in a way, and Radford's someone who wants to say, "Well, confidence has quite a lot to do "with whether or not you've met the belief condition." Then you have other people who think that confidence doesn't have much to do, so you can rather not
1:07:15
not have a high degree of confidence in something that you believe in, but you've still satisfied that belief condition. I wonder what you make of that. - Confidence is complicated, and there's a lot of different things that people will mean by it. I don't necessarily mean to suggest that you have to have a lot of really enthusiastic personal phenomenology.
1:07:44
I think there are a number of different treatments of confidence that might be compatible with the kind of approach I wanna take. Here's the key thing as far as I'm concerned: what I'll count as confident belief formation is really just the crystallization of evidence formation into a fixed result.
1:08:03
So when you move from the stage of weighing evidence, searching for "What am I gonna think here?" to the sense of having an answer, and you produce an answer, then you'll count as confident enough in my books. And so it may be actually even that Radford's reluctant student who's asked for the date of the Battle of Hastings or something like that, who's fumbling for it
1:08:27
but then he comes up with something, I'm gonna count him as off the hook when he comes up with it, even if he doesn't, even if he's not really robust in his response. But it's sort of an empirical question how confidence affects attributions of knowledge, and there is some work on this, there's empirical work on this. There's actually one piece of empirical work that I feel like I have to share with everybody because it's affected me in a very strange way.
1:08:57
This work is by a psychologist named Brennan, I believe.
1:09:03
I can send you the citation later if you want. You know when you're talking and you sometimes have filled pauses, like you're kind of "ummm" and sometimes you have silent pauses before you come up with an answer? - It happens. - Yes, it happens. So they had people evaluate
1:09:21
whether somebody knew an answer or not,
1:09:27
basically how smart do you look. If you are going to come up with the answer, do an unfilled rather than a filled pause. Silence, and then nail it. If you are not gonna come up with the answer, try umm-ing a bit. 'Cause you actually come across as knowing. There you come across as searching. But it's kind of like an empirical question how we register confidence and how we see the relationship between that and states of knowledge or belief. (speaking from off mic) (laughter)
1:10:00
But you may have a sort of overriding, non-intuitive background knowledge (laughter) And then you make those mental state attributions there. I think maybe if a random person had to evaluate Kripke on the subway it might not be so- (laughter) Don't put this on the web. (laughs)
1:10:21
- I know you were just introducing us about these particular experiments as kind of a key study from (mumbles) but I do see what about the particular experiments are actually really interesting themselves. You started with this experiment from Pinillos where he says that the number of times you have to check varies depending on the stakes, and then later on you talk about these studies by Alfano and colleagues where they show that whether something's morally good or morally bad
1:10:45
affects whether people think that you believe their ultimate proposition. And within experimental philosophy, those are usually regarded as two separate experiments, or two separate issues. But you were suggesting maybe we could come up with a unifying explanation of both of those findings. So I was wondering is there something we could do going forward that could test that hypothesis, the hypothesis that we could find a unifying explanation of those two findings
1:11:07
that are usually regarded as separate.
1:11:14
- Both of those findings have something to do with something that I take to be an existing item of knowledge, which is that human cognition is adaptive,
1:11:28
that higher-stakes decisions are handled differently from really low-stakes decisions. So I don't think experimental philosophers should set out to do that. That's already been done. Cognitive scientists have done that, we don't have to do that. There are some things we could do about the perception of the stakes involved in norm violation and compliance.
1:11:50
This seems to be kind of common sense to me that if you're asked,
1:11:57
"Is it a higher-stakes matter to do something "that might harm the environment, "or something that would help it?"
1:12:04
I think if you have some questions there you could verify.
1:12:12
I take it that's one of the premises in Alfano, Beebe, and Robinson's approach that we do tend to treat norm violation as risky behavior,
1:12:26
as behavior that requires more thinking. Actually, even if the norm that you're violating is one that's a morally bad norm, it's more hazardous to do it and then, so I take it we could do some work on that.
1:12:44
It's always possible that these effects, which seem to me to have some superficial similarities, each is being driven by a completely different effect.
1:12:55
If we can find more shifty phenomena
1:13:01
that work in a similar fashion, then that would be a good thing.
1:13:08
Of course, we could also disprove it,
1:13:13
we can maybe find that there's characteristics in the one case that don't apply to the other. And in fact, people have done work that's critical of Pinillos. Schaffer and Buckwalter have done a lot of work which I actually find really interesting, and I find this kind of work pretty compelling, and I think there are some problems in some of the experiments that Buckwalter and Schaffer have done.
1:13:41
For example, they have directly asked with high-stakers and low-stakers, both of whom were actually making the right choice in the end about, for example, whether the bank's open or whatever, they've asked people directly, here's what this guy thought, and here's what happened, and they ask people directly, "How satisfactory was his evidence "on a scale of one to seven?" People tend to say "Pretty good," on both sides, and I think somehow. And they're saying,
1:14:08
"Look, that goes to show were not imputing more evidence "to the high-staker than to the low-staker." My own feeling about this is that you have kind of a problem with the liquid scale there. Like yeah, your evidence was good enough to get you to the right place, so some of that's what's really going on, but maybe there could be more sensitive tests of how much you're imputing. I think it's fully possible when we do more work of the kind that Buckwalter and Schaffer are trying to do
1:14:35
that we'll find the adaptive cognition approach doesn't work. We can keep going with this, and try to figure out what are we implicitly attributing to these people. - Well their main piece of evidence in this case is that it works just as well if you change the no to some other word- - [Jennifer] Right, so they go through the whole deontic thing and I don't have a problem with that. I think what's going on is- - But do you think that would apply in the other case? Like if you said (mumbles)
1:14:59
- Well, they do tend to find the same results in a lot of different places. I think, actually, there's a problem specifically with changing it to "guess," that if you change to word to "guess" it doesn't make much sense, so people charitably read that as "this person didn't mean guess." How much do you reckon we're gonna have to do here, right? Because if you try to read it literally, it doesn't literally make sense. So people have to accommodate the question the best they can.
1:15:29
I think there's a lot of shallow processing in these questions too, so that's something I've done a little bit of work on. People don't always detect the substitution of near-synonyms.
1:15:41
I recommend to everybody the work of Tony Sanford at Glasgow, who's done a lot of work on semantic-change blindness where he gives people versions of a story then with just one little detail changed, and looking at the conditions under which you do and don't notice changes. It's amazing the changes you don't notice. And you do notice the ones that are in focus, the ones that are italicized, you're much more likely to notice them. Or you notice ones where there's a lot of semantic difference
1:16:05
between the items being swapped out. But if something's non-focal for you, which it would be for many of our naive participants, you can be relatively blind to "Did I just ask you about knowledge or belief? "Did I just ask you about thinking, hoping, guessing?" What is the right mental state to have here? I think the deontic stuff, they try to say, "Well really we're looking at "what people think this person should do." It's not even connected to something about knowing.
1:16:31
I would want to explain that by saying that knowledge is the key mental state which we take to govern action. We think that people ought to act on the basis of their knowledge, and it's not a good idea to act just on the basis of what you randomly happen to think to be the case. I try to explain it that way.
1:16:55
- Hi, thank you for your talk. The general theme kind of reminds me of response theory in literary theory which eventually popped out of the idea that the meaning of a book or a text is whatever readers interpret it as. But even if you don't buy that theory of meaning, a lot of what (mumbles) how readers respond to different texts and how you can achieve a certain response if you want it. So I was wondering if maybe in this way, experimental philosophers might have things to learn
1:17:20
from literary theory. - Absolutely, absolutely.
1:17:25
In my own experimental work I'm working with Raymond Mar who's a social psychologist at York University who's done a great deal of work on how we understand fiction and how empathy works when you read and what you learn from it. And I think there's a great deal to be gained by engaging
1:17:43
both with psychologists who work on fiction and with literary critics,
1:17:48
and just looking at literary texts. There's a lot of really nice examples of what I call echoic use of mental state verbs. When you become conscious of it as an epistemologist, then unfortunately every novel you're reading is gonna be like this. (laughter) You know, where an omniscient narrator will describe a character as knowing something, and you just, "No, gahd, wrong track." But that's kinda the versant
1:18:16
of your engagement with the text, that you yourself are put in a position where there is some distance between how the character is being explicitly described and how you know you have to understand that character, and that's part of the excitement of fiction. And I think it's definitely something we can learn from our colleagues in literature. - Again, awesome talk, very informative. I just have a two-parter, but the first is a clarification.
1:18:42
Did you say that when they increase the cognitive load on high-stakes decision-makers in the econometrics studies by making either this timed task,
1:18:53
by messing with their cognitive load, and part of that is increasing it, how did they perform, what's the change in performance when you give them time restrictions? - They perform much more like low-stakes thinkers. They'll go through fewer options, they'll feel better about the ones that they chose. It's the paradox of unfounded confidence. Actually, if you're a super high-stakes thinker and I give you all weekend to make up your mind, you're gonna make a better decision
1:19:18
and feel less confident about it than if you do it absolutely under fire.
1:19:25
And so they do all kinds of things. They did some very cruel experiments (laughter) involving a dot, you put people in a room with a dot matrix printer. Do you guys even know what that is? It makes that noise, yeah, and then you have to answer a lot of SAT-type questions and you'll feel more confident about your answers, you'll answer faster, and you'll answer worse. - I guess the question then becomes, I don't know if this would be interesting at all or even valuable,
1:19:52
but it seems like, given that you can manipulate people when they're either in a high-stakes or low-stakes environment to make decisions of the other sort, couldn't you kind of get these people in the DeRose bank cases or other cases like that and essentially, for our naive philosophy students answering these questions, put them in a high or low-stakes environment and see whether or not that would affect their willingness to ask questions about whether or not in the bank case they'd really-
1:20:20
- Oh, so are you thinking that we're gonna put our experimental participants that we're asking to make mental-state attributions into different kinds of environments? Absolutely that has an impact on your mental state attributions. If you're under time pressure and you're distracted various familiar biases really come out, and some things that are typically seen only in three-year-olds, you can bring them out in adults if you crank up the time pressure high enough.
1:20:47
- And that's with the- - There is existing work on that, yeah, absolutely. (speaks off mic) Yeah sure.
1:20:56
- Hi, this is about your characterization of intuitions. So two things: one, I was just wondering if that was meant to be a really principled definition or just sort of a gesture at the sort of things intuitions might be.
1:21:13
Yeah, it would be great if you could- - So when I've, yeah, I can answer that really quickly. I think I did a bad job of it, but I do mean this in a very, I think there is a psychologically real distinction between intuitive and reflective thinking, and I very much follow the work of Jonathan St Bernard Evans on this. There's huge clusters of features which are associated with S1, S2, intuitive, reflective, systematic automatic however you wanna call these different distinctions.
1:21:47
There's speed, effort, shared with other animals, blah blah blah blah blah. The question is: which one of these things is absolutely fundamental? And the argument that's been advanced in the literature recently, and I think there's sort of increasing consensus around this is that the fundamental thing has to do with the involvement of working memory, that you can really tell that a task is a task for reflective thinking
1:22:15
when it collapses under cognitive load.
1:22:20
'Cause we have this sort of restricted capacity for working memory. So you're adding up digits in your head, that's an explicit task, and you can't do that very well while simultaneously out loud saying the alphabet backwards or something like that. But you could do it while drumming your fingers on the table. This is an automatic behavior and it doesn't slow me down in adding up numbers in my head. - But if intuitions are judgments
1:22:46
for which we are not aware of the steps in the reasoning
1:22:52
or something like that, it seems like experimental philosophy isn't testing intuitions, right? Because in all those cases, at least for me, I'm aware of some of the steps in the reasoning. - Good, excellent, but you're not aware of all of them. And in fact, okay, that's an excellent point.
1:23:10
What I'm gonna say you're aware of is you're aware of a series of features which are feeding into your intuitive calculations. Just as you're aware, when someone raises the question, "Does she know that there's water in that "that she's about to pour out?" Well, you're aware of my direction of gaze, various things about the object, blahblahblahblahblah. You're aware of that. There's a key step in your reasoning,
1:23:36
sorry, there's a key step in your cognition which is not necessarily introspectively accessible to you. The best way I can kind of try to bring this out is as follows:
1:23:48
In epistemology, we do these categorizations all the time where we just have a story and we ask, "Does this person know and really think?" There's two ways that you can do this.
1:24:01
You can do this according to the dictates of an explicitly-given epistemological theory. So I could tell you what Goldman's 1967 causal theory of knowledge says. It says that in order for S to know that P it must be the case that S's belief that P is produced by the fact that P, or something like that. Give you an explicit definition, and then I'll give you as story. Could be Goldman '76, the Henry and the barn, and say, "According to Goldman's '67 theory, "would this be a case of knowledge?"
1:24:30
There your categorization judgment itself would be, I hope, explicit in that you would look at the story and you'd say "Yes, this is a case "where there's this kind of causal connection, "and according to Goldman's theory, "knowledge is something "which has that kind of causal connection," and you're explicitly clicking the categorization into place according to the stated rule of the definition that you're following. As long as you're doing that,
1:24:56
and that's a kind of reasoning that's very familiar to us, we do it in different kinds of contexts, I could ask you, "Is trespassing a Class B misdemeanor under Texas law?" and you would have, if you're a Texas lawyer, some kind of explicit understanding and you'd see that trespassing falls under that. But as long as you're doing that, you're not generating fresh and exciting knowledge about Class B misdemeanors under Texas law, or about knowledge itself, actually. All you're saying is,
1:25:22
"Well this fits what knowledge is "according to Goldman's theory." But as Goldman '76 himself proved, you can discover that you're dissatisfied with an existing theory, because you just have this intuitive feeling that this case, although you can see explicitly that it fits Theory blah and Theory X and Theory Y, you just look at this case and you feel like, "Yeah, but that's not knowledge. "That's not knowledge with a capital K, "whatever knowledge really is." And you can do that
1:25:46
without already having in your back pocket a story about what knowledge is. That's part of the beauty of Gettier's 1963 paper. Thank god, he didn't give us an explicit theory. "Knowledge is not JTB, knowledge is JTB+Q." He didn't do that, he just said, "Well here's a counter-example." He didn't have to have his own explicit theory to apply to the case. So he didn't have to do what I would call reflective thinking where the steps are in that sort of long division or syllogistic reasoning way,
1:26:14
all completely explicit in making the categorization judgment itself. You're absolutely right that your conscious of something as you read through the story and think, "Wait, but that's the identical twin, "and she didn't know that he went home," or whatever, and you're conscious of something there, but you're not performing the categorization itself through explicit reasoning the way you would in, for example, the legal case.
1:26:40
- A practical question: So if I wanted to run an experiment of Gettier's, is it enough if I now change the story to say John is walking and if he's looking at a clock, and the clock is not working,
1:26:56
but John isn't aware of that,
1:27:01
would I avoid the problems with the factive? - Oh, you know what? I actually didn't mean to suggest that...
1:27:13
that Weinberg, Nichols, and Stich are bad guys for having the factive there. No, I think actually this is a perfectly legitimate way to tell somebody that a Buick has been stolen.
1:27:27
He's also not aware that her Buick has been stolen, that part was also non-directly, it's totally okay to do this, but I was just trying to make the point that you guys infer it and you infer it without being aware that you're inferring it, and that's okay, but it's just an illustration of how you pick things up and color things in. And we can absolutely, we're told that this person is doing something good and this person is doing something bad. You infer all kinds of stuff
1:27:51
about their underlying mental states as a result of this. I don't know, if anyone else has children of an age where they got into these things called the Bionicles from Lego. Does anybody else know about Bionicles? Okay, one of the things about Bionicles, these are these little sort of robotty toys from Lego and you get them and you put them together, is that there's good ones and bad ones. The bad ones are called the Bohroks, I hope I got that right, the good ones are called the Mata Nui.
1:28:16
Have I got this right? If there's a Bionicle nerd, please correct me. And what makes them good or bad
1:28:24
is the color on the box, basically. (laughter) And it's so funny because you think no, but surely to be a good person or to be a bad person, or even a bad Bionicle, you have to have certain beliefs and desires, and there's certain very very interesting differences between mental states of good and bad people. Actually also, this is the thing that really bothers me, if you read the little comic books that come with Bionicles, you just have the good ones and the bad ones,
1:28:53
and they're actually behaving in more or less identical ways. (laughter) They both wanna take over the world, but these are the good people who are trying to get the crystal of power, and yeah anyway.
1:29:03
And as an experimenter it's really tempting to say, "Well this the red Bionicle over here "and that's the black Bionicle. "This one's good, this one's bad." But of course, even as you're trying to understand the Bionicles and play with them, you of course have to attribute all kinds of mental states to them to makes sense of that characterization.
1:29:21
- Can I just try and find what the moral was about the factives and presupposition. One moral seemed right to me,
1:29:32
that presupposition, or whatever you wanna call it, can't be about common ground. It's a way of putting forward new information in a conversation, a sort of common-ground idea doesn't seem to do justice to the very idea of what people are trying to get at. - You're usually trying to make something common ground in the sense that this is gonna be uncontroversial. Which is why it is so annoying when you're fighting with someone and they say,
1:30:03
"Look, I know you don't care about your colleagues, but," (laughter) Doesn't that burn you, right? Because we're all taking it for granted now that you don't care about your colleagues. But you could at least, you know, that's a burn, right? And I think it's a burn partly because of the way it's being pushed into
1:30:22
"Well this is uncontentious." So I think that there's something about what the main point of the utterance is. - Right (mumbles) So with this being aware that the, it's typically regarded as factive, as is no, regret.
1:30:42
Now I wanna put, I think it's in the next slide that you started talking about, is it presupposition I think? - [Jennifer] Oh okay, and there's all that grief about is it presupposition, is it entailment. - Right, and the idea of presupposition being preserved in this potential negation, and then this idea that, well but we've got these cases where, "Hey, it doesn't seem to be preserved." - [Jennifer] Oh okay, right. - And then there were always two stories about this,
1:31:08
and I think it's sort of evidence in favor of the non-popular story which is the one that, I guess it did, it was a dissertation in 1975, said, "Look what these cases seem to show "is in the cases of (mumbles) "it's as if it's really, "you've got the option of an obligature, "it's just a sort of scope thing with negation." In the case without negation, it just seems like an entailment.
1:31:32
If John knows that P, then P, right? If John forgets that P, then P. - No, because I think you can still have echoic use there, and so I mean- - Well, I think you certainly have to put the heavy stress on to get this. (mumbles) Some people are poor (mumbles) It's very hard to get without putting this very special stress or,
1:31:56
- Italicization. - Just a general reciprocity on it, you can't really get those readings. It does seem as if, at one point you did say it was an obligature, and it seemed to me that it is a better way of handling because you do seem to be able to have this very straightforward cancelability without putting heavy stress. It is true that in many cases you can say, "Well John didn't know that P "because it wasn't the case that P," but you just can't say, "Well no, John didn't know that P
1:32:22
"because it wasn't the case that P." It seems absolutely,
1:32:27
John didn't forget killing Batman because he didn't kill Batman. It seems absolutely fine.
1:32:35
- [Jennifer] But you wouldn't say it out of the blue unless your were in a conversation of context where someone was essentially insinuating that John knew. - It's insinuating that somebody, or that yeah. - Was it Batman? - Well whatever it was. (laughter) I'm struggling to find what the moral was you're pulling out of the discussion- - Well actually, okay, here's my secret moral,
1:33:00
that I've been doing experimental work lately
1:33:07
where I've looked at what people say about a bunch of really closely-matched cases. So for example, just Russell's stopped clock, Gettier case, and then matched variance of that example. So there's a variant matched for word length, matched in all kinds of ways, where somebody's looking at a clock and everything's normal. And do they know that it's three o'clock in the afternoon. And then cases where it's a justified false belief, let's say. They're looking at a clock that's broken
1:33:32
and showing the wrong time. Do they know that it's three? And the story explicitly stipulates that it's 3:30, and they pass a comprehension screen question saying that they know it's 3:30, and a shocking number of people will go ahead and, depending on how you frame the question, 15, 20% of people will go ahead and attribute knowledge in these cases. And I see this as performance error, and very much to my dismay some people have been picking this up and saying, "Oh no, this is a sign,
1:33:59
"no, it isn't really factive." And I think that's a disaster. And I think anybody who starts peddling that kind of theory really owes us.
1:34:07
They're gonna have to actually do a lot of repair work on discourse comprehension and so on. Generally we've got very very strong background reasons to believe that no is factive if you just look at how the flow of information works in conversations and so on. So I've been thinking about this because I'm paranoid, and because I'm paranoid, I'm worried that it's actually not easy to test
1:34:32
exactly how factive verbs work,
1:34:35
partly because of what goes on when you ask questions about them. I prefer testing these things in sort of indirect ways. There's ways of, for example, giving people little mysteries to solve where you have, I always do this first day of my epistemology class, give them little mystery stories larded with factive and non-factive verbs and I get them to deduce who the killer is or something like that. And they typically then have trouble explaining exactly how they did it.
1:35:00
But you can get people to deploy their knowledge so that you can detach a factive verb instance. - Is there any case in which you can get this to work very nicely? Especially with, say, "just found out that."
1:35:15
Why is Susan so happy? Well Susan just found out that Tom was promoted, and you know the addressee is dying to find out whether Tom was promoted. "Was Tom promoted?" It's a very natural way of now saying that was the information you were trying to get across. - Yeah, and I think actually there's some very interesting differences between thick and thin factives under negation as well. And this is all very complicated territory, and I don't know very much about Laurence Horn
1:35:40
on metalinguistic negation. I don't know very much about that. I would like to know more,
1:35:48
and I'd certainly not publish on this without doing a lot more research on it. But I'm interested in this kind of corpus work
1:35:58
on how people actually use factives, I think can be pretty instructive for us as philosophers and experimentalists. So I've been trying to do some work like that, just looking at the relative frequencies of different constructions. To what extent are people using first-person, third-person attributions of knowledge, for example. And what are the differences in the kinds of contexts in which people use first-person and they go third-person.
1:36:28
Are all first-person uses just conversational?
1:36:35
- [Carisa] More questions? - I just wanted to clarify something jumping off your answer to Zoey's point, and the stuff with the high-stakes. Are you saying, really just to clarify, are you saying something like the mechanism that generates intuitions,
1:36:52
A, that it's not going to be happening
1:36:58
or talking to or dealing with working memory faculties? And B, that maybe because it's not dealing with working memory in a strong sense, that it's not sensitive to cognitive load effects? That is to say, people will give the same intuitions given all sorts of manipulations of clear cognitive load. Is that what your answer was? Did your answer tell those two points about the intuitive mechanism? - Oh I see, so you're suggesting if mental state attribution is really intuitive,
1:37:32
then performance shouldn't suffer under- - Well I'm saying if it's intuitive and if it's separate from, or really there's a strong separation, a large separtion, between whatever mechanism is responsible for giving you these intuitions and whatever mechanism deals with syllogistic reasoning or theses kinds of things, or number manipulation, ie some robust working memory faculty. And if those two are really distinct, then impacts to increasing cognitive load
1:38:03
which disrupts working memory shouldn't disrupt the intuitive mechanism, right? - I think in principle there's something to that line of questioning, and the quick answer
1:38:15
is that the answer is not a quick answer, it's actually very complicated. I think there are a large number of different cognitive abilities involved in mental state attribution. And the key thing is something about our consciousness of the processing.
1:38:34
But it can certainly be the case, look hey, to construe these stories you have to have a sequence of explicit and introspectively available thoughts about them. That all has to be set up. And also even to track what's going on with somebody in your environment, depending on how complicated the scenario is, you may have to have a lot of explicit thought about what's going on, even if the judgment is presented automatically. It's actually very controversial in psychology right now,
1:39:04
the conditions under which mental state judgment is more automatic or more under control. - It seems like it's gonna be very hard to get the case where you don't have any working memory, even in the- - Yeah but my analysis
1:39:19
of the difference between these cases is I'm not saying we make
1:39:25
a low-stakes mental state attribution in S1, and a high-stakes mental state attribution in S2, I'm not saying that.
1:39:37
- So what are you saying? - Oh, so I'm saying when we look at this story we naturally,
1:39:45
as a result of automatic intuitive calculations we make, anticipate more evidence collection over here, and when evidence collection is held constant, lower confidence over here than over here. That's what we automatically anticipate. - Not the fact that guy's in higher stakes and that one's in lower stakes. - Yeah, the reception of that would be processed the same way in the two cases. - So just to, just another thought, that also means that intuitions,
1:40:18
we can't tell whether or not something is an intuition just by knowing that there was a question posed in a particular way and so we get a particular answer. We might have to know about that individual? It seems like- - Absolutely, absolutely. For example, one of the things I can do if I become very committed to a particular theory of knowledge, is you ask me a question: "Is this knowledge or not?" And I sit there and I think about my theory, and I say, "Yeah it is,"
1:40:42
and there I'm no longer using intuitions as a check on my theory, I'm using my theory as a way of getting me out of the hellaciousness of having to generate an intuitive response. And maybe, Craigie Hitchcock tells me people do this with causation, that when he goes to Igal Kvart and he asks him, "Okay, is this a causal relation here?" You can bet your booties that Igal's gonna give an answer which is gonna be basically generated by Igal's theories. So it's a worry, sure.
1:41:08
But it's also possible for epistemologists, like Alvin Goldman himself, to come up with responses that totally undercut theories that they themselves used to be really committed to. And it's possible for,
1:41:20
Williamson feels the intuitions that motivate contextualism. He hates contextualism, but he feels them. I feel them too. And that's embarrassing and awkward for me (laughter) and it puts burden of work on me to try and explain it away. But it's one sign that your devotion to a theory is not necessarily gonna contaminate all of your intuitions. You're capable of having intuitions that are strangely, and sometimes informatively, opposed to your theory. And maybe I'll come around one day
1:41:46
to accepting the truth of contextualism if I can't do a good job of explaining away the intuitions that I'm saddled with. - You'll also train yourself, and then you give intuitive responses - That can absolutely- - from the theory that would- - Yes yes, there is a risk of that. And I think it's good, for that reason, to keep mixing it up with fresh examples. I've been spending a lot of time lately reading medieval Indo-Tibetan epistemology, which is extremely rich in Gettier cases and so on,
1:42:12
and it's refreshing to me. You're tired of the fake barns? (laughter) Well try the case of the goat and the sword for a change and see if that works for you. And sometimes just our familiarity, "Yeah, fake barn, doesn't know," 'cause you've taught Epistemology 333 a few too many times, and then it's refreshing to work with different materials. And actually, sometimes you might find that there's something in the materials which is maybe a little bit defective.
1:42:38
That's actually a concern I have sometimes with replication, that we keep using the Pontiac and Buick case
1:42:47
and that's cool, but I think it's really good, also, to try and reach out for a huge diversity of other materials. If an effect is really robust, you should be able to find it when you're talking about...
1:43:02
cars and sports and purchasing jewelry and telling the time, you should be able to vary up those, that's not a point against replication, it's just a plea for more fresh material. And partly to keep us from getting too contaminated with our old intuitions. - [Carisa] David? - Yeah, I was wondering if you had any pragmatic suggestions for how people can tell what kinds of things they're implying with their descriptions of the case? As you said, because we don't know, or it's hard for us to tell
1:43:33
which things are explicitly stated by the words in which things were just conferring. It sounds like in some cases it could be quite difficult for us to make sure that we're not inferring- - I don't think we can necessarily make sure of everything in advance, partly because we don't fully yet understand how our intuitive capacities are working. What are they actually responding to? Face recognition turns out to involve all this weird stuff about the geometry of the jaw or something like that.
1:43:58
Who knew?
1:44:02
And so we can't necessarily figure it all out in advance. I think one very good thing to do is just start by looking at the work already in psychology
1:44:12
in the relevant domain. So look at the work in psychology on what happens in human judgment and decision-making as stakes shift, and you can find a lot of work on perceptions of shift there, so that can be fruitful.
1:44:25
But you may have to keep messing around. Actually, I would like to know, Josh you might know this.
1:44:35
When Beebe and Alfano and Robinson,
1:44:39
Christ, where's the Knobe effect?
1:44:44
When they came up with the idea that what's really driving the bus here is actually whether the chairman even believes, that was surprising, 'cause we'd read these and we had assumed, of course in both cases the chairman knows that the side effect is gonna be produced. And that's maybe like we're looking down at this from the experimenter's perspective and you want it to be the case that this information transfer
1:45:09
went just the way you as the experimenter would have it go.
1:45:16
I guess there was a study, was it Surian and Pellet's (speaking from off mic) Yeah, I wasn't gonna try and say that. They actually did something which was kinda relevant to them. But do you know how they came up with that idea to say, "Oh wait, maybe it's the belief that's really the thing?" (speaking from off mic) Yeah, how did they have that insight? 'Cause that's something like, it's a question, how did they figure, that's relevant to what David's asking.
1:45:36
How do you figure out that actually, secretly, you're not aware that you're doing this? You're inferring something different in these two cases. Especially when they're presented between subjects. You're inferring something different about what the mental state of the chairman is. - I guess the thing that came first was just that in the latter case, the harm case, people think he knows that he will harm the environment more than in the help case, and then all the theories
1:45:57
that were supposed to explain why that happens involve the idea that holding fixed that you believe it equally in these two cases, there's just something about the latter case that belief will count as knowledge. And then they think, "Wait, maybe there's-" - Do people even believe? - It's not just that you believe it equally, but you only know it in the latter case. Maybe people don't even really believe it in the former case. - And it's easy- - So perhaps we can actually insert explicitly,
1:46:20
"Well I don't know one way or the other "whether that's true or not," but I don't go into all that. - Then you should have an exacerbation of the effect I would presume. I have no idea. (speaking from off mic) - Well no, if you just explicitly insert it into the chairman's remark.
1:46:37
- But then in the latter he said, "Well I don't really know." Does he believe it? We could try right now. (laughter)
1:46:48
- Yeah, and I mean there's also the problem that when people talk about harming or helping the environment, their utterances are not always taken at face value. So you could see the person who says, "I don't know whether this will harm the environment," it's like, yeah he knows, he's just saying, "I dunno, maybe it won't."
1:47:08
There are a lot of problems here about, so you might wanna have him not say something, you might wanna stipulate something about him. We're gonna have to be careful with this. But it's a great idea. (speaking from off mic) - I really liked your insight that as experimentors or as philosophers there's sometimes a tendency to get into this tunnel vision, and it's because of the point of view that we have as kind of having the pan-out, taking a step back and seeing both perspectives
1:47:37
and that sometimes the subjects in these experiments don't necessarily get that view, and it reminded me of another talk, the empathic perspective-taking.
1:47:49
And the way I see it is that sometimes if you just only see one account, your subject in an experiment, you try and put yourself in one of the two situations, either high-stakes or low-stakes, and you still can't really put yourself in that situation. But when you see them contrasted, when you see that difference, then you realize, "Okay, the circumstances now, "I feel them more.
1:48:13
"I see what the contrast might be." Whereas if you were solely in one perspective, you wouldn't be able to. - I think that's a very perceptive comment. I think that's really good. One of the things we could look at is we can look at individual differences in empathy. So that was something I tried to do in my work on Gettier case recognition. Some people score higher in empathy than others, and that's correlated with their task performance on registering other people's mental states and so on.
1:48:40
If you're a higher-empathy individual, and you have one of these gray zone, borderline cases, you'll be more likely to see the target as justified in what they believe, because you're better at stepping into their shoes. But that's definitely something that you can investigate empirically. - [Carisa] Zoey? - Yeah, so in your response to Marilyn, it seemed like you were implying that if a subject has memorized how her theory will apply and just shoot it out,
1:49:09
then that's going to count as an intuition. - It could, it could. - It would, according to your definition, but here's a story where it seems really absurd: I go to my first philosophy class, my first ethics class, and someone gives me a trolley problem, and it's the fat man version, and I say my intuitive judgment is, "No, it's horrible to throw the fat man onto the tracks." Fast forward two years, I've (mumbles) and I've also read and talked about the trolley problem many times,
1:49:39
and now you give me a trolley problem and say,
1:49:44
"What's your intuitive judgment? "Should you throw the fat man over?" And my response just pops out, "Yes, of course I should throw the fat man onto the tracks." That really doesn't seem like an intuition to me, it just seems like it's a very fast application of the theory, and just because I'm not going through all the steps doesn't mean I should call it an intuition. - Do you think there's a special phenomenology associated with intuition? Is that what's missing for you? Or do you think-
1:50:12
- No, I think it seems like what you've offered might be a necessary condition for intuition, but it's not sufficient. It seems like there are other judgments where you're not aware of all the steps of processing- - I think awareness is sort of a fallible test, whether people can report that they're aware of something, that's a little bit of a fallible test for the involvement of working memory. Strictly technically if there was some working memory involved
1:50:41
you might not be, interestingly enough, aware that you had exercised it. One of the things we're not typically aware of is whether we are engaging working memory or not for solving a problem, interestingly. That would just be one more nusiancy thing to have to pay attention to.
1:50:59
I think I would need to hear much more from you about why it is that you feel that the very very well-rehearsed judgment wouldn't be an intuition.
1:51:11
I don't know if you have something you can say to- - Yeah, it seems like it really is relying on
1:51:20
inferences that you were explicitly aware of in the past, and don't need to make at this present moment because you've made them before, and you've memorized that that's the way the process should go. - Okay. But that counts as, even though its causal genesis is in past explicit inference I see no reason why it couldn't count as, at this point, this is intuitive for you. I think it won't have the same, should not have the same epistemic role
1:51:50
in theory construction as a pre-theoretical intuition would.
1:51:57
(speaking from off mic) - It was just an idea I had for trying to get out what implications people might take. It seems a lot of these deal with people via these characters and stories doing things, and they try to make sense of, they sort of guess things about these characters' personalities and attitudes in order to make sense of their actions. So one thing you could do is maybe before giving ads give these stories to a few people and give them open-ended questions like,
1:52:24
"Describe the chairman of the board," or "What do you think of the chairman of the board" or something. - Yeah, I think Josh you've done that kind of work, right? You've elicited people's judgments. Yeah?
1:52:37
- About the personality, like asking them- - You get them to make all kinds of trait attributions to the person. - Well to me (mumbles) it's just that trait attributions have no impact on it. It's just there more about this behavior, not the moral judgment about whether he's a good guy. - Well you don't necessarily ask them, "What do you think of the chairman of the board? "Is he a good person or not?" but you ask them open-ended, describe the chairman of the board, describe his attributes
1:53:04
and not necessarily give value judgments on him. - Yeah, that might not be a bad thing to do. Something that Tania Lombrozo did which I thought was very telling was she thought, there's a sense in which norm violations are much more informative for you than norm compliance. Norm compliance, who knows, who can tell? Norm violations, that really does, in a sense, tell you more about this guy. So she found people were willing to make much stronger predictions, in this case, of future behavior,
1:53:30
than they were in that case. So that is something along the lines of what you're suggesting. - I'm just curious whether there's a way to make the subjects read stories like the way they read train tables. - Yeah, I know right? (laughter)
1:53:49
- I'm just asking because, the Miyake written test, I don't know if you know that, but it really seems like they are meant to be read as train tables, they have the options as something, whether it's true or false according to- - Yeah, and don't get creative.
1:54:05
- I'm just wondering whether if you insert that into the comprehension test- - Yeah, okay. I have had these experiences trying to do empirical work where you just wanna grab your reader by the scruff of the neck and, "Read it this way!" And it's really hard to do.
1:54:22
I teach these really big classes in Toronto, and I have clickers for my students, I ask multiple-choice questions, they click the answers, and the funny thing is that I've had this experience of putting up, I don't know if you guys know Keith Lehrer's Haveit and Nogot case It's a case in which there's a person called Haveit who has a Ford and a person called Nogot, who does not have a Ford. And I had a question at the bottom of the screen: who has a Ford in this story? A Haveit. B Nogot.
1:54:52
And I will not get 100% (laughter) I get like 95% Keith DeRose says you should change his name to Haveaford.
1:55:02
So yeah, it's very hard to get people to read things exactly the way you want. This is actually something that I think developmental psychologists have done some really neat work on. They've got people to do the following: I have to figure out how I'm gonna operationalize this. This is testing children's understandings of factives. They ask the kids to,
1:55:22
"Pretend you are saying the following sentence:
1:55:27
"Keith knows that Bob bought a new skateboard. "Pretend you're saying that." And then they'll ask, "Did Bob buy," and then, "Pretend you're saying the following statement: "Keith thinks that Bob bought a new skateboard. "Did Bob buy a new skateboard?" So they try to get you to pretend to say something, and I know that sounds kinda crazy, but maybe that is something that could get you out of some of the pragmatics, is if you could make somebody pretend that they were a certain kind of character.
1:55:58
I know Hugo Mercier has done some really lovely work where he got people to pretend that they care about an issue. He tried to get people to pretend that they have a stake in the matter, and then people gave a more accurate reading of the story.
1:56:18
We certainly know quite a lot about conditions that improve and deteriorate the understanding of arguments of logical reasoning. So if you put people in certain kinds of social situations they understand arguments much better. I would like to know more about how you can get story comprehension to go down really well.
1:56:41
But part of the problem is when we are doing mental state attribution and we're figuring out, "What does this person know? "What do they think? "What do they want?" typically we are doing that very much in an open-world environment, right? We're not doing it where you're meeting some computer who has a data output of the objects that they are looking at and the muscle motions that they're making. So if you want something that's ecologically realistic in capturing our intuitive capacity
1:57:10
the way it ordinarily works,
1:57:14
you keep finding yourself back in that open-world position. But you can get some kind of funny effects from this.
1:57:23
In one of my studies I had a lot of filler questions. One of my filler questions, I wanted to have more questions where I had testimonial knowledge transmission and perceptual, I had a whole bunch of cases. I realized I didn't have that much with deductive reasoning, so I had a detective go through some reasoning about who committed a crime, and I wrote this little mini story so it was just crystal clear there were only three suspects, and this person's gotta be ruled out,
1:57:47
and that person's gotta be ruled out, lots of factive verbs. And then I had "so the detective knows that this person is the killer," and 95% of my participants gave the answer, "No he doesn't." And I was just like, "Why?" And I asked my son, who was 12 years old at the time and pilots all my work, and he always gives the right answer, but on that one actually, he'd gone with the majority. And I asked, "Why Alex? "Why are you saying this? "Isn't it obvious when you do the deduction?"
1:58:14
And he's like, "Mom, you read a story like this, "whenever it looks really overwhelmingly "like one person did it, "you can bet that's not the guy." (laughter) And so it's like he's suffering from the effect of all these Hardy Boys novels that he's been reading. I told my psychology collaborators about this and they laughed at me for expecting that anyone would do closed-world reasoning on a case. People don't do that. So you can have these genre effects. Unfortunately.
1:58:44
- I have a quick question. I'm not very familiar with the literature in epistemology in the Tibetan literature, but from what I've read, one of the things I find more provocative is that some people seem to think that, you've made some claims about our introspection and about how our intuitions are very opaque to us, especially in these cases of knowledge. But some others in the Tibetan tradition seem to think that people can actually get better at that, at their introspection,
1:59:11
and they can not only get better at knowing when people know things, but also at their own psychological processes behind their intuitions. So do you think there's any evidence to think that this might be true, and might it be relevant for this literature? - That's a really interesting question. I think there's certain kinds of task
1:59:34
where introspection could be something that could improve over time. But if the key feature of your processing is not going through working memory, introspection will always be silent about it. It's always gonna be below the radar. It's like these computations you make, like when I recognize you as Carisa, I'm apparently doing this on the basis of some freaky geometry stuff about your face, and I couldn't explain it if I tried. And no amount of introspection is going to make it possible
2:00:04
for me to bring that up to the surface. So I think introspection has some limitations about the kinds of cognitive processing that it can reveal to us. Introspection though, may reveal some interesting things.
2:00:20
Maybe some of the effects that we feel around things like conversational pragmatics you could get better at detecting over time if you trained yourself in various ways and read enough Greiss or something. I think maybe you could get better that way, that might be cause for hope.
2:00:40
- [Carisa] Any more questions? We have maybe time for one more.
2:00:46
- Does anyone just show video to these subjects instead of presenting a story? - Oh absolutely. It works the same way whether you tell a story, just like face recognition works the same way. If I bring Tom Cruise in the room or I show you a picture of him, or even just show you a really good cartoon, you're gonna recognize who that is. My collaborator Raymond Mar has done a lot of work, including FMRI work where he's compared how people respond to live-action movies
2:01:12
and how people respond to very very nicely-matched cartoon depictions of the same interactions. So same capacity. Sometimes live action brings it out a little bit sharper, you can get differences in performance. But it doesn't actually matter whether you give people a narrative stimulus or whether you give them a live-action thing to evaluate. - [Carisa] Okay, thank you so much Jennifer. - Thank you. (applause)
2:01:42
- [Carisa] Thank you also to all of our speakers.
Here to help you learn
Sign in to keep a private conversation with your course tutor.
Ask me anything about the material…
Sign in to chat