[Editor's Note: Thanks to philosophy grad student and musician Al Baker for this guest post.]
The first time I heard the term "experimental philosophy," part way through my master's degree, it sounded like such an obvious oxymoron that I couldn't help but think it was a terrible idea. I shared, and continue to share, many of the worries that Seth pointed to in his recent post on the subject. Although I can certainly see the appeal in having some kind of respectable yardstick against which to measure any philosopher's intuitions (I share Dan Dennett's suspicion of arguments overly dependent on the word 'surely'), I think that many have probably underestimated the difficulty in obtaining empirical evidence actually capable of serving that purpose.
Some of the reasons why I think this go right to the heart of the theory and practice of the scientific method. Even the most tightly controlled experimental results admit of multiple interpretations, and the nature of the subject matter of much x-phi means that the highest standards of scientific enquiry simply cannot be met. Suppose we are running an experiment to test the assumption that people, in general, are moral objectivists. It turns out that, when given a certain scenario, people respond in the way that the designers of the experiment intended to indicate that they are moral objectivists. So what? Does this mean that most people are moral objectivists, or only think they are? What kind of moral objectivist are they? One of the implausible kinds, perhaps? Are people intuitively intuitionists for instance? Does this make intuitionism more plausible or just show that most people are in the grip of an implausible theory? It is no shame on x-phi that no experiment will be able to give us a watertight answer to whether the assumption they set out to test is true. What is more likely, as with the legion x-phi insights on various "trolley problems," is that the results from x-phi will force philosophers into potentially interesting theoretical acrobatics in order to accommodate the results into theories that they already prefer for other reasons.
I can confidently predict that no x-phi result will ever, by itself, settle a significant philosophical debate, but you know what? I think that's fine. I think it's fine from the point of view of armchair philosophy, because the theoretical acrobatics I just mentioned are a vitally important part of philosophical practice, and the more things that force us into engage in them the better. I also think it's fine for x-phi. This is partly because I think conducting intuition-hunting experiments are valuable to philosophy for the reasons I just mentioned, but also because I really don't think experimental philosophers are trying to do anything more. X-phi sceptics frequently get all hot under the collar about the supposed claims of x-phi that it can usurp armchair philosophy. That x-phi-ers somehow think that the right survey, with the right population, under the right conditions, will tell us once and for all whether or not we should pull the trolley-switching lever, or whether we survive being teleported, or whatever the hell else. Not only is this clearly rubbish, but it is highly doubtful that any philosopher, experimental or otherwise, thinks any such thing.
Those who say that x-phi has got too big for its boots fall into two camps. Those who think that x-phi has pretensions to do away with armchair philosophy, or instill itself as the new dominant philosophical method, and those who think that it cannot even do the intuition-hunting that so much of it claims to. The former camp are, to my mind, scared of a straw man, though I don't think there are as many of them around as the blogosphere (on both sides of the x-phi debate) would have us believe. No-one but the most foamy-mouthed of hardcore intuitionists would so much as hint at the claim that "if most people think that P, then P is true." The worry that x-phi cannot even reliably source intuitions is more problematic for those philosophers who embrace it, but I don't think this is a criticism that holds against all x-phi necessarily, rather it is a reminder that philosophers might perhaps not be the best people at designing experiments. Maybe we'll get better at determining which questions can be answered using x-phi and which can't, at how to isolate which variables and how to formulate a null hypothesis.
Or maybe these are problems distinctive of the kind of x-phi we are most used to reading about: the kind that asks people what they think of a morally tricky situation. Should we kill one to save five? Did the CEO intentionally hurt the environment? etc. etc. If what we're testing for is intuitions on intractable philosophical issues, we shouldn't expect the answers to be much less difficult to interpret than the questions themselves. Interestingly, and thankfully, intuitions about difficult problems are far from the only things that x-phi can test that might be useful to philosophers generally.
What things besides intuitions can x-phi help us establish? One piece of experimental philosophy recently reported in the British Journal of Aesthetics, Meskin et. al's "Mere Exposure to Bad Art" (the article is behind a paywall, but the piece is nicely summarised here), attempts to establish not intuitions but preferences: specifically, preferences for artworks.
The problem the paper seeks to address is this: studies have shown that "mere exposure" to particular artworks leads to increased preference for them over others to which one has not been exposed. One implication of this has been taken to be that the canon of great western art is maintained not by virtue of any special artistic merit, but rather just because those canonical works are the ones to which we are most exposed. Is the main reason The Beatles will live on in the canon of western art but New People probably won't that we all get played The Beatles a hell of a lot more than we hear New People? If so, this poses a problem for philosophers of art, since so much of our work is predicated on the fact that there are real artistic and aesthetic qualities that works have, and by virtue of which they can be judged as better or worse works of their kind.
Meskin and co called this contention into question with their study by showing that the "mere exposure" effect was not the same in all cases. Students were "merely exposed" to some good art (Millais landscapes), and some bad art (Kinkade "cottage porn"), in tightly controlled, randomised experimental conditions. It was found that, far from the exposure having a positive effect on the viewers for Kinkade's work, after only one exposure there was a statistically significant drop in preferences for his daubings. So, the paper claims, we can't maintain the claim that "mere exposure" is the principal determining factor in our artistic judgments, or at least the principal determining factor in what is included in the canon of great western art.
What's interesting about this paper as an example of x-phi is that, in contrast to the intuition hunting projects we see discussed more frequently, the result here can't even be mistaken for an attempt to answer the philosophically interesting question ("what determines our artistic preferences?") with an experiment. Rather, this experiment is making room for philosophy to happen. This is similar to the role I suggested that ethics x-phi actually plays in philosophical discourse - it forces proponents of certain views into interesting acrobatics - but in this case the experiment allows for the acrobatics to happen at all, rather than dictating the size of the trampoline we have to bounce on (if I might torture the metaphor a little). This experiment shows not that a certain philosophical position is the most widely held, but rather that a philosophical discussion of the subject is a worthwhile pursuit.
I guess the final thought is this: X-Phi is great if it is making space for philosophy to happen, which is what most x-phi does in actuality (even if not in intent). In many ways it's a shame that the most prominent examples of x-phi have to do with testing intuitions on philosophical problems. When the things being tested for are precisely those things at issue there's bound to be some deep design and interpretation problems to be uncovered in analysing the results. However, when our x-phi is designed to test not "folk" intuitions about philosophical problems, but philosophically significant features of the "folk" that are not in themselves philosophical views (like their artistic preferences in certain conditions, for instance), the data gathered is guaranteed to be far less mysterious, and probably far more useful.
-Al Baker
Image Note: The image is a Thomas Kinkade parody found here, referred to there as "Kinkade Road" by Akadajet.
We spent a little time making fun of Kinkade back on ep. 28.
Al, is academic politics aside is there a meaningful difference these days between x-philo and academic psychology?
some interesting conversations that are working on a different approach happening at the intersections of Object-Oriented-Ontology and arts/manufacturing/design/engineering: http://objectsobjectsobjects.com/
Interesting question. I think there are a lot of reasons to see them as exploring similar ground. Obviously the methodologies of x-phi are largely cribbed from experimental psychology (which is why I would feel better if more of it had some experimental psychologists involved).
Having said that I still think there are big differences between them. Philosophers and psychologists tend to operate under significantly different theoretical frameworks. What counts as an explanation for a philosopher may well not be the same as what counts as an explanation for a psychologist. This is not to say that any given philosopher won’t be more or less psychological in their outlook, or of course that either standard is in any sense better, except at investigating its own subject in its own way.
As for object-oriented-ontology, I’m afraid I’ve always found it rather hard to take seriously (though I suspect it isn’t intended to be taken overly seriously). I remember being told by a friend of mine at a prestigious Canadian university very earnestly about her professor’s fascinating Derridean takes on the anthropology of cheese. I can’t help but be suspicious of something that sounds like it would be so at home in a Fry and Laurie sketch.
I get the gist of what your saying but can’t help but feel that the rising stock of x-phil makes this ” What counts as an explanation for a philosopher may well not be the same as what counts as an explanation for a psychologist” less and less true, time will tell I suppose but I put my money on the psychologists being the last ‘man’ standing as the humanities are dismantled by the new economies/politics.
Not sure that any of the metaphysics of OOO (or Speculative Realism writ large)will make much of a notable impact in philosophy but the interactions with people who make things have some merit.
for those PELifers looking for some easy access to the Xphilo/pscyho hybrid check out:
http://bloggingheads.tv/programs/current/mind-report.
Well, what counts as an explanation in psychology for why Millais is in the canon but Kinkade is not will probably not meet all the desiderata for a philosophical account of the same question.
More to the point, since most x-phi *does* seem to be of the kind concerned with ethics that I pointed to as troublesome, I’m not really sure how x-phi could supplant armchair philosophy as a way of answering, for instance, whether willing the ends entails willing the means, or whether this changes depending on the moral status of the ends in question. All it can do, all it claims to do, is establish what most people intuit about this. It doesn’t seem to have the tools available to itself to say whether these people are correct or not.
Sure, if we’re looking at a new-economic-climate Battle Royale of the humanities perhaps philosophy will lose out to psychology, but only according to criteria that I wouldn’t give much time to. I’m frequently frustrated by how psychology experiments that are so nearly relevant to my research not quite answer the question I wanted because the team conducting the experiment were operating under different (and I believe flawed) assumptions. I think there’s much value to be had by employing psychological experiment to give evidence to support philosophical theorising, but I think there’s possibly much more to be had by philosophical examination of the assumptions governing the design and analysis of psychology experiments.
we will see, what we would like to nail down and what we can actually get a grip on are often not in sync, as Chalmers and others have noted “progress” in philosophy writ large is contentious at best.
The better bet (in terms of progress, who knows about viability) seems to be the various marriages between phenomenologist/enactivists and their colleagues in anthro/psychology, Alva Noe’s new book on art is on the reading list so maybe there is something there to build on, and of course the later versions of ANT (Actor-Network-Theory) are always worth a look:
http://www.heterogeneities.net/papers.htm
I don’t think it’s merely novelty that speaks to artists (just as they once flocked to the work of Wittgenstein) but yes there is probably a sociological aspect to this as for all of these matters…
Oh, and on your OOO point, I think you’re right there’s definitely *some* merit in the approach, but I think it’s probably more novelty value than anything. It’s not like examining the relationship between objects and the people who make and use them is a project alien to philosophy.
Well, unless you’re prepared to get into a long and probably tedious discussion about what legitimately counts as ‘progress’ in philosophy I think we’re going to have to recognise that we’re singing from different hymn sheets on that score and leave it there.
Al, based on your understanding of X-phi, could you weigh in on whether you find the Knobe effect as presented by Seth to be typical of X-phi, and whether you think the Knobe effect is a valid experiment. Thanks, Wayne.
Is it typical? I honestly couldn’t say. I’d be interested in seeing a meta-analysis on what kinds of questions were addressed by x-phi. My main point in the piece was that there is an interesting difference between x-phi that seeks to test for philosophically loaded intuitions and x-phi that seeks to test for other philosophically interesting features of people in general. I think the Knobe experiment is probably typical of the former kind of x-phi, whether that’s typical of x-phi in general I would need to do more research than I have to know.
I’m not entirely sure on what you mean by ‘valid experiment’. Perhaps the study could have been better designed, but that doesn’t mean the question isn’t one that can be answered in that kind of way. I think those kinds of questions are certainly worth pursuing experimentally, I just think the results have to be extremely carefully handled. I don’t doubt that the Knobe effect exists, the experiment certainly seems to establish that, but precisely what that means in philosophical terms, I’m really not sure.
Al
If we just go with Seth’s presentation of the Knobe effect as the standard of x-phi, there are several problems.
The first level of concern can be expressed by the difference of two kinds of experiments:
1) Running an experiment of flipping a coin and recording heads or tails for each trial (say 50), taking the results and calculating a (statistical) percentage to get this experiment’s results of the probability of heads (or tails), say 28/50 heads, .56, or 56. % heads in this experiment. If this were a scientifically structured experiment, the results would be preceded by a review of pertinent literature in the field, both theoretical and experimental. The data would be presented, and conclusions would be drawn relating the current results to the field of knowledge previously published.
2) A different approach would be to ask 50 people what they thought would happen if I were to flip a coin, would they predict heads or tails on the next flip, record the results and come up with x percentage of heads. What is the difference in these “experiments?”
My concern about the Knobe effect (x-phi?) as presented by Seth is that it not only resembles the second type of “experiment” and therefor is of dubious value even to begin with, but that in calling it an experiment, it masquerades as an experiment, giving false meaning to a mere opinion poll.
Here was my summary on Seth’s blog of the Knobe effect: “Don’t know if I’m missing something here, but the primary variable seems to be intent as measured by secondary variables of knowledge of a) will harm, b) will help, and with that knowledge the president knows of each outcome of a)/b), but does not care about the outcome of harm or help, and therefore is knowledgeable, responsible and thus intentionally disregards either harming or helping (in favor of profit). I do not see any intuition at play here, just the president’s statement of knowledge of not caring about harm or help and thus I don’t even see this as being an experiment so much as a true/false no-brainer. ???????”
As you can see I take issue with how the variables are fundamentally even defined, let alone the meaning of the words used (i.e., intention). In a valid scientific experiment this kind of fuzziness is defined much more clearly.
The question of validity is whether an experiment applies to reality. My impression is that the Knobe effect is impossible to be valid because the terms are not defined, rendering any experimentation illogical and thus invalid. It is also one step removed from reality, as in the second “experiment.” raising infinite difficulties
I think you zeroed in on the best approach when you said “I think there’s much value to be had by employing psychological experiment to give evidence to support philosophical theorizing, but I think there’s possibly much more to be had by philosophical examination of the assumptions governing the design and analysis of psychology experiments.”
A perfect blend of this process in effect is Marc Hauser’s “Moral Minds”:
A Summary: “He argues that humans have evolved a universal moral instinct, unconsciously propelling us to deliver judgments of right and wrong independent of gender, education, and religion. Experience tunes up our moral actions, guiding what we do as opposed to how we deliver our moral verdicts.
For hundreds of years, scholars have argued that moral judgments arise from rational and voluntary deliberations about what ought to be. The common belief today is that we reach moral decisions by consciously reasoning from principled explanations of what society determines is right or wrong. This perspective has generated the further belief that our moral psychology is founded entirely on experience and education, developing slowly and subject to considerable variation across cultures. In his groundbreaking book, Hauser shows that this dominant view is illusory. Combining his own cutting-edge research with findings in cognitive psychology, linguistics, neuroscience, evolutionary biology, economics, and anthropology, he examines the implications of his theory for issues of bioethics, religion, law, and our everyday lives.” Enjoy.
P.S., I disagree with his conclusions, but not with his methods, and if I were to dialogue with him, I would simply use similar methods in order to dialogue meaningfully on the issue.
Yikes–after reading some summaries of Marc Hauser’s Moral Minds, I found that
“Harvard Finds Scientist Guilty of Misconduct”
Harvard University said Friday that it had found a prominent researcher, Marc Hauser, “solely responsible” for eight instances of scientific misconduct.
http://www.nytimes.com/2010/08/21/education/21harvard.html
So I agree with his methods, except for cheating 🙁
His scientific objectivity in Moral Minds is not necessarily affected by his scientific misconduct, but I wouldn’t bet on it. Talk about ironic, and be careful talking to a moralist.
Hi Wayne,
OK, let me try and address these points and defend the Knobe experiment a little:
I agree that Knobe is an experiment of your second kind, but I am not sure why you say it is ‘therefore of dubious value to begin with’. Your comparison with the first kind of argument is, I think, a little uncharitable. If you want to discover how often a coin will come up heads, then yes you run the first kind of experiment. If you want to discover how often people *think* a coin will come up heads (an equally legitimate inquiry) you will need to run an experiment of the second kind.
So, a survey of people’s opinions on the ratio of heads to tails isn’t going to establish how often heads will actually come up, but equally a record of the actual ratio of heads to tails won’t establish what people think the ratio is. The difference between the two experiments is that they seek to establish (or falsify, if you prefer) different hypotheses.
I’ve actually just read Knobe’s paper: http://pantheon.yale.edu/~jk762/Side-Effect.pdf
It’s short and very informative. As far as experiments go, it isn’t very thoroughly reported, but I still think it can be taken as a reliable indicator of the existence of the Knobe effect, and the fact that there is such a thing that can be relatively well defined.
One interesting thing to point to is that Knobe himself says that “Clearly, ordinary language does not here constitute a court of final appeal.” He is explicit that all he seeks to establish is that people’s intuitions about intentionally bringing about side effects are influenced by their view of the moral status of the side effect itself.
The variable to be isolated, then, is the moral valence of the side effect in the story. Or should I say stories, because Knobe actually runs two different scenarios – one involving a CEO, the other involving a Lieutenant in the army. Now, I think one could reasonably raise questions about both of those scenarios, but at least this rules out the danger that the effect is due to some very specific feature of people’s attitude to CEO’s or the environment.
He doesn’t explicitly state a null hypothesis, or publish a lengthy statistical analysis (though I believe he has done so elsewhere) but he does point to the findings of a previous experiment he did as something he wants to challenge, which could probably serve as the null hypothesis here, that:
“people only consider an effect to have been brought about intentionally when the agent was specifically trying to bring about that effect”
Knobe’s experiment obviously challenges this because in neither scenario is the agent specifically trying to bring about the effect, but people intuitively ascribed intention to the agent apparently based on their moral judgement of the side effect itself.
Seems legit?
Al,
The sleight of hand in the design of this experiment is all a matter of true-false logic, written up in the guise of objective science. If I say I don’t care if I hurt people, and then hurt them, would you blame me? Of course. If I say I don’t care if I help people, and the help them, would you praise me? Of course not. That is how the questions were asked in the experiment.
That is the poor design, the illogic at the core of this experiment of mirrors and smoke in the name of science, with a claimed p<.001, highly significant statistical difference. See how that just must be right if it is statistical. Now we can claim that whatever theory we threw into the experiment to be validated (very, very, very real!).
Scientific experiments can go wrong on so many levels: poor research, poor apriori theorizing, poor design, poor definition of variables, poor control of variables, poor application of statistics to conditions, poor conclusions, poor aposteriori theorizing, etc. This one was doomed from the get go on logic alone.
You cant trust moralists or science. Validation is necessary, which is at the heart of good science.
But the very supposition that:
“If I say I don’t care if I hurt people, and then hurt them, would you blame me? Of course. If I say I don’t care if I help people, and the help them, would you praise me? Of course not”
is what is supposed to be tested by the experiment.
I suppose you think that proposition is true, from my quote above. I don’t get what you think Knobe is trying to prove beyond the statement you apparently agree with?
P.S., in a well regulated academic environment this type of stuff does not get published due to the checks and balances of experienced and qualified academicians, so it would makes sense to suspect studies of this type (x-phi in general?).
Al–sorry if you are picking up a negative attitude from me, just trying to express my concerns.
Let me be more specific about the experimental design:
1) “Subjects in the harm condition read the following vignette:
The vice-president of a company went to the chairman of the board and said, ‘We are thinking of starting a new program. It will help us increase profits, but it will also harm the environment.’
The chairman of the board answered, ‘I don’t care at all about harming the environment. I just want to make as much profit as I can. Let’s start the new program.’
They started the new program. Sure enough, the environment was harmed.
2) These subjects were then asked to determine how much blame the chairman deserved for what he did (on a scale from 0 to 6) and to say whether they thought the chairman intentionally harmed the environment.”
3) Subjects in the help condition received a vignette that was exactly the same, except that the word ‘harm’ was replaced by ‘help’:
The vice-president of a company went to the chairman of the board and said, ‘We are thinking of starting a new program. It will help us increase profits, and it will also help the environment.’
The chairman of the board answered, ‘I don’t care at all about helping the environment. I just want to make as much profit as I can. Let’s start the new program.’
They started the new program. Sure enough, the environment was helped.
4) These subjects were then asked to determine how much praise the chairman deserved (on a scale from 0 to 6) and whether they thought the chairman intentionally helped the environment.”
If I say I don’t care if I help someone, I have told you that my intent is not to help them, so you can not ascribe praise.
If I say I don’t care if I hurt someone, I have told you that my intent is not to avoid hurting them, and thus you can ascribe blame. I don’t need an experiment to deduce this.
I never thought you were being negative, I just wasn’t sure what specifically your point was.
I think you’re taking the experiment to have a normative conclusion that it just doesn’t claim to have.
The conclusion of the experiment isn’t that it’s right to blame the CEO in one instance and wrong to praise him in the other. It is merely that most people seem to think that way. You say it’s ‘obvious’, then the most you can accuse Knobe of doing is of giving evidence for something that’s obviously true (or at least, that most people think something that you think is obviously true)
However, I’m not sure if what you said *is* obviously true. If you say you don’t care if you help someone, as in the vignettes, you don’t have an “intention not to help someone” as you put it, you just are just absent of an intention to help them. The absence of an intention to do a thing does not obviously entail an intention to not do the thing. Even if you know you could do the thing. This is precisely why the act/omission distinction is so morally problematic.
An example: I have milk, eggs and flour in my cupboard. I know I could make pancakes, cakes, waffles, muffins etc. etc. I don’t care about all the tasty treats I could be having though, so I choose to make bread, and intentionally do so, using the last of the flour. Do I thereby intend to not make pancakes and intend to not make muffins and intent to not make waffles? You’re giving me a lot of intentions I didn’t know I had. I thought all I intended to do was make bread?
The thought there is that a side effect of me making bread is that I am no longer able to make muffins etc. I intentionally made the bread, but I really think it’s a stretch to say that I intentionally did not make all the other things I know I could have made. Maybe you disagree, but surely you can come with me so far as to say that it’s not obvious?
If you first tell me you do not intend (care) to make pancakes, cakes, waffles, muffins, etc., then you do have an intention (care) to not make them.
I can tell you right now that I don’t intend to go and get a job pumping gas tomorrow. Are you telling me it’s obvious that tomorrow I’m going to be constantly, intentionally and deliberately avoiding getting a job pumping gas?
Ok, I must be incredibly blind about this, so I will take some time and get back to it from a fresh perspective later. Thanks for your patience, regards, Wayne
Al, in response to your previous statements:
“However, I’m not sure if what you said *is* obviously true. If you say you don’t care if you help someone, as in the vignettes, you don’t have an “intention not to help someone” as you put it, you just are just absent of an intention to help them. The absence of an intention to do a thing does not obviously entail an intention to not do the thing. Even if you know you could do the thing. This is precisely why the act/omission distinction is so morally problematic.”
–I elaborate on my own previous statements (in quotes):
“If I say I don’t care if I help someone, I have told you that my intent is not to help them, so you can not ascribe praise.” This is not just absence of an intention. The CEO clearly states, I do not care if I help people with my program, which most people would interpret as lacking any claim to helping people.
“If I say I don’t care if I hurt someone, I have told you that my intent is not to avoid hurting them, and thus you can ascribe blame.” I don’t need an experiment to deduce this.” The CEO clearly states, I don’t care If I hurt people or not. This does not mean he intends to hurt people, but if he does, I think most people would ascribe blame, not intent to the CEO for hurting people.
The results are therefore an artifact of this underlying driven implication described above (the Setup) which forces the answers to the questions, separate from any real conditions of morality, responsibility, intuition or any other higher order possibility. This entire line of abstract discussion is precisely due to the fuzziness of this type of “experiment.” I think that is the best I can do to clarify what I mean, so I think we may just agree to disagree.
I think you’re right that it’s best to leave it here, though you should note that if you can at least agree that this has not been an unreasonable disagreement, there is value in the experiment showing that most people seem to think the way you do on the matter.
The disagreement, ultimately, is over whether the statement “I do not care if I help people with my program” entails that the speaker has an intention to not help people with their program. Here’s some more similar statements that make me think we shouldn’t ascribe intention based on those kinds of things – but if you’re unmoved by them then yes we should just leave it here…
“I do not care if eating my breakfast now means I can’t eat it later” (do I have an intention to not eat my breakfast later?)
“I do not care that driving to the shops will contribute to climate change” (am I deliberately destroying the planet?)
“I do not care that smoking will make me ill” (when I smoke do I intend to make myself unhealthy?)
Basically, I think you need room to distinguish between causing an effect *intentionally* and causing one *recklessly*. In the above cases I think I am not intentionally bringing about the results (there seems to be a difference between my smoking because I’m addicted and not caring that it will give me heart disease and smoking because I want to get heart disease).
One final point – in the experiment Knobe was careful to ask different people the ‘harm’ question from the ‘help’ question. The reason for this is obvious. If you were asked to address both scenarios, say you said the CEO intended to harm the environment, and then you were given the ‘help’ scenario, you would feel some pressure to be consistent and say that the CEO also intended to help the environment in the second kind of case (as per the findings of the previous experiment that Knobe is here disputing the conclusion of). So again I just have to say I don’t agree with you that the answers given in the Knobe study were obvious, or somehow illicitly extracted from the participants by the wording of the question. I think if you were asked both questions at once you would be pressured to give a uniform answer (he did or did not intend, in both situations, to achieve the side effect), which makes the findings of the experiment interesting.
Al–look at the confusion embedded in this statement (quoted from Knobe’s article):
“These subjects were then asked to determine how much blame the chairman deserved for what he did (on a scale from 0 to 6) and to say whether they thought the chairman intentionally harmed the environment.”
Did the subjects rate blame, or intent?
I think you should challenge Knobe in your doctoral dissertation and become famous.
If we are generous in accepting Knobe’s contamination of intent with blame, we could interpret Knobe’s surveys as accessing not an articulated concept of intentional action, but a pragmatic concept of intentional language. “Pragmatic” would refer to judgments people may make due to social context that may not be part of the semantic content of a sentence or judgement.
For example, if Tom says ‘I don’t desire to see Bill today’ people may judge that Tom desires not to see Bill today. Of course, that does not follow, but may be inferred. Conversation can lead to such implications when one is not being fully informative. If Tom did not want his audience to believe that he wanted to avoid Bill, he should have said more. In normal conversation, this sentence is a way of implying that one wants to avoid Bill, other things being equal.
Similarly, when one says ‘you did that on purpose’ or ‘you did that intentionally,’ one may be conversationally implying blame, but blame is not part of the semantic content (or core concept) of doing something on purpose (intentionally). We could point out that these implications are cancellable. When we imply blame by saying ‘you did that intentionally’ we may cancel the implication by adding ‘but, of course, it is okay to do that action intentionally’. So this feature of intentional language is pragmatic and not part of the semantic core of the concept of intention or intentional action.
However, I think the experiment is flawed in design by contamination of these concepts. A properly designed opinion poll would be more valid.