Tuesday, July 17, 2012

"Propositional Attitudes": Belief vs. Desire Grudge Match

If contemporary analytic philosophers are to believed, "propositional attitudes" are a central feature of mentality, perhaps the central feature. What is a propositional attitude? It is a mental state like belief, desire, intention, hope, fear, etc., that can take a full proposition -- speaking loosely, a sentence -- as its complement. I can believe that grass is green. I can desire that we eat cookies. I can hope that aliens will rescue us soon. The mainstream view is that human action arises from the confluence of our beliefs, desires, intentions, and other attitudes.

Now there's a funny thing about how philosophers discuss propositional attitudes. When we list the canonical propositional attitudes we always seem to list belief first. Why is that? Is belief more propositional or more psychologically important an attitude than desire or intention? Not according to conventional wisdom. Also, desire always seems to be listed second. After that, it's a grab bag.

This is not just a quirk of conventional linguistic ordering. Philosophers studying propositional attitudes also typically analyze belief first and most centrally, desire secondarily if at all, and other attitudes only haphazardly. The literature in philosophy of language on "propositional attitude reports" (i.e., sentences like "Lois believes Superman is strong") is especially striking in almost always using only belief examples. The Stanford Encyclopedia entry on Propositional Attitude Reports might be typical. A search within the page yields 183 hits for "belief", 169 hits for "believe", and 2 hits for "desire". The two occurrences of "desire" are not examples of desire reports, but rather gestures toward an interest in propositional attitude reports generally, including desire.

This monomaniacal focus on belief despite an advertised general interest in propositional attitudes seems, on the face of it, likely to be distortive. Things might go differently, in some ways, for belief than for other propositional attitudes. It's risky to generalize from a narrow range of cases, especially if other people in your subfield are also generalizing from the same narrow range.

I thought it might be fun to quantify this "belief first" distortion. So I set up a grudge match between belief and desire in philosophical work on propositional attitudes. Belief enters, of course, as the heavy favorite.

Here's what I did. I downloaded all the entries in Philosopher's Index that included "propositional attitude*" in a keyword/title/abstract search. Within those entries, I looked for appearances of "belief*", "believe*", and "desire*". (I needed only one search term for "desire*" since same word serves as both verb and noun; the asterisk is a truncation symbol.) If both "belief*"/"believe*" and "desire*" appeared in the same entry, I recorded which term appeared first.

Here are the results. The total number of "propositional attitude" entries was 616. Of these, 382 contained neither "belief*"/"believe*" nor "desire*" in the keywords/title/abstract. (Most of these will presumably discuss belief and/or desire in the body, but the abstract is framed more, shall we say, abstractly.) Of the 234 entries that mentioned either belief or desire, 182 (78%) only mentioned belief, 13 (6%) only mentioned desire, and 39 (17%) mentioned both belief and desire. Among the 39 mentioning both belief and desire, belief appeared first in all 39 cases.

Final score: belief 221, desire 13. Belief wins, with blood on the moon! Tim Schroeder, time to get back to work!


carrie figdor said...

Not that your conclusion is wrong, but it may be overstated. belief* is used by some philosophers to indicate a belief-like state that is not belief, in (e.g.) discussions of deflationism. Also 'want' is also commonly used in examples of folk psychological explanation. that said, it isn't assumed that what goes for belief goes for desire and the rest; after all, belief is one attitude, desire another. Is the complaint that it's a mistake to always interpret content in terms of that-clauses (belief that it's raining vs belief in God, desiring that I have coffee vs. desiring coffee -- which seems kinda weird, actually). i have nothing invested in this, it was just a curious study.

Eric Schwitzgebel said...

Hi Carrie!

In mechanical content analysis of this sort, there's always noise, of course! My sense is that the issues you mention aren't going to have a big effect on the results.

"Want" is not a very common term in these entries, appearing in only 14 entries total, and skimming through them, about half of those use "want" not to refer to a propositional attitude as a target of philosophical discussion but rather in something like a signpost function. (For example, "I want to discuss a certain argument for the claim that....") In contrast, appearances of "belief*"/"believe*" and "desire*" are much more typically in the role of targets of philosophical discussion in entries that contain the phrase "propositional attitude*".

"Belief*" where the "*" is not a truncation symbol but an asterisk appended to the end of the word is harder to search for, since "*" can't be used as part of a normal search term in PhilIndex or Excel. But if the results of an MS Word term search are to be believed, the term occurs in not a single one of the 616 entries. A spot check suggests that it is at least rare.

On your final point: Yes, there is certainly a subliterature arguing that belief and desire are not or are not always "propositional attitudes". I'm not hostile to the idea -- my own work on belief can fly on either model -- but it's not the orthodox position.

Charles T. Wolverton said...

A simple explanation might be that in common speech we are more familiar with the formal structure "I believe that P" (where P is a straightforward assertion like "God made honty-tonk angels"). The formal structure of other PAs may be less familiar in common speech; eg, the formal PA "I desire that I have a cookie" usually would be expressed simply as "I'd like a cookie"; "I intend that tomorrow I am at your house" as "I'll be there"; "I hope that you forgive me" as "please forgive me", thereby obscuring the underlying PA.

A more subtle reason might be that the other PAs typically involve a change in current circumstance. Eg, "I desire a cookie" suggests that no cookie is immediately accessible, and while the ultimate objective is access to a cookie, the immediate desire is a change in circumstance - generally requiring an action, eg, uttering "please pass me a cookie". Similarly, the other PA examples might require walking to your house and uttering "please forgive me" respectively. Then implicit in trying to satisfy a desire would be a belief that some action will likely result in the requisite change in circumstance. If so, lurking behind the other PAs would be a belief, in which case one creating PA examples might be inclined to "eliminate the middleman" by replacing other PAs by beliefs, eg, that "if I ask politely, you'll pass a cookie", that "I can walk to your house", that "I can persuade you to forgive me".

So, you might want to check if some "belief" occurrences mask underlying PAs of other kinds.

Michael Schmitz said...

Great post, Eric. I believe the pattern you describe is not accidental at all, but manifests a very deep-seated tendency to favor theoretical, truth-apt, attitudes and speech acts over practical ones that I like to call the “theory bias.” The theory bias is massive and ubiquitous. And yet it can seem elusive, so that it’s nice to have the kind of quantitative evidence for it that you provide. In the history of analytic philosophy it is evident from the very beginning (though it goes back much further): it is present in Frege and glaringly obvious in the verificationism of the Vienna circle and the early Wittgenstein’s dictum that the world is everything that’s the case. Starting around the thirties of the past century there was a rebellion against the theory bias with Austin’s speech act theory and his attack against the “descriptive fallacy”, the later Wittgenstein’s insistence on the diversity of language games, Hare’s prescriptivism in ethics, and various attempts to account for imperative and other practical inferences by philosophers such as Hofstadter & McKinsey, Castaneda, Sellars, and many others. But in roughly the late 60s a backlash began with the rising popularity of truth-conditional semantics as a general semantics, cognitivism in metaethics and many similar tendencies. Ever since the theory bias has reigned supreme again. In contemporary philosophy it is manifest, for example, in attempts to reduce intention to belief and actional to perceptual experience. Perhaps most strikingly, it is usually just taken for granted that knowledge is a form of belief, though it is hardly obvious and in fact very implausible that practical knowledge of what to do, and of how, where, and when to do it, is a form of belief. Practical knowledge of what to do, it seems to me, is even more disregarded in comparison to theoretical knowledge of what is the case than, as you show, desire and intention are in comparison to belief. (Would be nice to get some data on that as well. The recent flood of literature on know-how, by the way, is no counterexample. On the contrary, because it just deals with the opposition between skill and discursive knowledge. The difference between practical and theoretical discursive, conceptually articulated, knowledge is still all but ignored.) Finally, the very notion of a propositional attitude as commonly conceived already embodies the hegemony of the theoretical, because each attitude, whether theoretical or practical, is supposed to contain, or be an attitude towards, something that as a truth value bearer is clearly on the theoretical side. This seems rather arbitrary and cannot make sense of the fact that we ascribe truth values to beliefs, conjectures, and assertions, but not to desires, intentions and orders. I think it is time for this construction to be revised and for the pendulum to swing back into the other direction so that the theory bias will be corrected. Your post gives me hope that this may happen sooner rather than later.

Eric Schwitzgebel said...

Very cool analysis, Michael! I assume you have or are working on a paper on this?

Charles: Your first explanation seemingly wouldn't *justify* the privileging of belief over other PAs, even if it explains it, right? Your second explanation might be partly justifying -- but see Michael's comment!

Michael Schmitz said...

Thanks, Eric, and yes, I am working on a paper on this, which should be done soon!

Eric Schwitzgebel said...

I'd be interested to see it when you have it in circulating shape, Michael.

Charles T. Wolverton said...

eric -

Could you elaborate on "but see Michael's comment!"? Altho it's on a plane at - probably well beyond - the limits of my knowledge, I don't find anything in it with which I disagree. Being a truth-skeptic ala Rorty, I assume that I don't suffer from"truth bias". I'm Sellarsian wrt knowledge and see the PAs as perhaps unifiable under the umbrella of behavioral dispositions, as opposed to being truth bearers. You are correct in observing that I was thinking of explanations rather than justifications, but only because my guess is that the phenomenon isn't justifiable.

Michael -

I too would be delighted to get a peek at your paper.


Scott Bakker said...

I find all machine reading exercises fascinating, but this is the first time I've encountered it in a philosophical context. It strikes me that your analysis could be elaborated into a whole new way to read philosophy more generally. Are others mining this technique as well?

Continental types have been harping on the problem of cognitivism for a long time as well.

On a different note, I was wondering if anyone knew of any work on the relationship between the way PA's muck up compositionality and the way contextualizing claims (via machine reading exercises like these, just for instance) tends to short-circuit the 'view from nowhere' more generally.

Eric Schwitzgebel said...

Charles, I was concerned about what seemed to me to be a belief focus in your comment: "lurking behind the other PAs would be a belief, in which case one creating PA examples might be inclined to "eliminate the middleman" by replacing other PAs by beliefs".

Eric Schwitzgebel said...

Scott, I agree that content analysis searches of this sort are underused in philosophy. One needs to be careful with them, but I do think they can add some rigor to otherwise off-the-hip claims of the sort, "philosophers typically say...".

Many historically-focused philosophers and fans of cultural variability will be sympathetic to your last point that contextualizing philosophers' claims more broadly can undercut our sense that we have a view from nowhere, as it were. But virtually no one is doing this quantitatively, with the exception of some recent cross-cultural work in "experimental philosophy".

Scott Bakker said...

If you ask me, "Philosophers Typically Say..." sounds like an awfully interesting title for a full-on article...

Charles T. Wolverton said...

I just ran across the following from Davidson's essay "Rational Animals" (p. 102 in the Sub/Intersub/Ob-jective collection) which seems a coherent statement of my intended point in the rambling second paragraph of comment 3 above:

I think I have shown that all the PAs require a background of beliefs, so I shall concentrate on conditions for belief [without which] there are no other PAs

If he's right, that seems to be a justification.

Eric Schwitzgebel said...

For the purposes of that article, that move might work. I don't think the argument generalizes beyond Davidson's specific anti-animal-PA goals, though.

Unknown said...

I would say the focus on belief to the detriment of other attitudes is because belief is the attitude which deals with truth. When we label a thought chain as something we have an attitude of belief towards we are effectively saying 'I believe it is true', i.e. that the proposition effectively says something about the real world. The analytical philosophers then immediately go off and talk about truth and the reasons you might have for holding something to be true - most famously (and unprofitably) in the special case of believing propositions to be true because of the definitions of the words and logical operators contained in the proposition itself.
This track has come to something of a dead end ever since Quine. But what has not been done and which I think is far more interesting is to investigate how we actually, practically think. The lack of philosophical debate on this is to me remarkable. The recent book Surfaces and Essences is a good step in this direction but it is amazing that we have taken so long to go even this far. When we form a proposition (be it true or not) e.g. 'the cat sat on the mat' - what exactly are we doing? how does it work? do we have separate concepts for each word? only for nouns and logical operators? how is it possible for us to translate this sentence into other languages? how do we then 'add' attitudes to it, such as desire, inquiry, intention and yes belief? is the assigning of attitudes to thoughts (propositions) a separate and subsequent process? how does it work? all interesting stuff - strange that not many people seem to be working on it....

Eric Schwitzgebel said...

Unknown Jan 10: If it's normative and hypothetical, that's logic -- but as you are suggesting, if it's our actual cognition, that requires a lot of empirical inquiry, and psychologists need to lead the way. There is of course a huge literature on concepts, and another related huge literature on cognitive structure in general (e.g., nativism vs. empiricism). I agree it's important!