 |
| Robbie Williams |
LB: What was your main motivation in choosing to investigate the nature of representation? How did you become interested in the topic?
RW: I’ve been fascinated since days in graduate school with underdetermination/indeterminacy arguments in the theory of representation. To illustrate with one I’m thinking about right now. Start from the following picture of the metaphysical grounds of belief and desire content: the content of an agent’s mental states is whatever it has to be, to rationalize the agent’s actions (the outputs) in the light of evidence (the inputs). An indeterminacy argument seeks to show that the constraints just mentioned can be met, not just by sensible-seeming interpretations, but by wild and crazy ones. The one I’m looking at right now, for example, constructs interpretations that have the agent thinking of themselves living in a universe that is void except in the local perceivable and manipulable environment of the agent (the agent’s “bubble”).
It’s fun to think up these crazy interpretations and show that they meet specified constraints. But what’s philosophically interesting is the lessons we draw. The most natural one—and my first instinct—is to draw the lesson that our account of the metaphysical grounds of content is incomplete, that we’re missing some `saving constraint’ that knocks out the crazy interpretations. And then the goal is to pin down what that missing ingredient could be. An alternative is to accept the arguments and revise our views about how determinate mental or linguistic content is. This strategy too raises interesting questions in the theory of representation—if we want to pinpoint what’s wrong with accepting indeterminacy of content, in my view we’ll need to say something about the theoretical role of representation, which pinpoints what we’d lose by embracing deviant interpretations. A final alternative—more popular historically than nowadays, perhaps—is to preserve the account of the grounds of representation but avoid indeterminacy, by rejecting presuppositions of the argument about the metaphysical character of the world that is being represented.
My graduate work focused on indeterminacy arguments and possible saving constraints—working in particular within the kind of `interpretationist’ metaphysical grounds for content David Lewis proposed (one element of which involves the rationalization story about mental content described above), which in certain parts of the philosophy literature has become almost orthodoxy, though rarely examined or defended in detail. The moral I drew from that work 10 years ago was largely negative—that living with indeterminacy wasn’t a real option because it’d prevent content from doing the work it needed to; and that the saving constraints that people (including Lewis himself) proposed didn’t work.
Maybe one moral at this point to have taken was that the Lewisian interpretationism was a busted flush, and to look instead at alternative traditions—salient among them being the causal-teleological accounts of Dretske and Mililkan, or (in the case of language rather than mental content) the ideas surrounding the causal theory of reference. But when I turned to look at those, the really compelling stuff that I could extract seemed to lack some of the virtues I prized in interpretationist accounts. Interpretationism, if it worked, would give a story about all content, not just special cases (reference to medium sized dry good and their observable properties). My hunch was that I wouldn’t find in these alternative traditions a satisfactory story about unsexy but genuine questions about what grounds the relation between the word `of’ and its semantic value, or about the grounds of content of highly theoretical beliefs remote from perception or action.
So this felt like unfinished business to me. For one thing, I’d said something about the way the metaphysics of representation wasn't—but the goal of course is to say something about how the metaphysics of representation is. Second—speaking personally—I struggled to find a presentation of the overall interpretationist picture that satisfied me as a real theory rather than a theory-sketch. The Lewisian story about linguistic content I thought I had a good grip on—but how the story about mental content went was much murkier. Thirdly, I was prepared to bet that some version of the interpretationist story was going to be part of the overall account of mental content—and since others in the literatures I was reading seem to be betting the same way, it seemed the division of philosophical labour required someone to go investigate whether that bet was going to pay off.
LB: In the first couple of years of your project, you were looking at whether the representational properties of belief have a genuine explanatory role. What sort of themes emerged as part of your research?
RW: One theme that emerged in the first couple of years was an argument about what is required to defend the spirit (as well as the letter) of a reductive realist theory of representation. I argue you need to do more than put forward a metaphysical theory that identifies where representation lies in the naturalistic world. You also need that metaphysical theory to be purely about representation. Let me expand on that.
We can illustrate the distinction here with a story about measurement. Suppose you have a community, the metricians, who report lengths simply by associating numbers with objects. So they would say “the length of the table is ” where we would say “the length of the table is 1 metre”. Since our measurement practice has an explicit role for a unit, we can easily report the same thing in other forms “the length of the table is 6 feet such-and-such inches”. The metricians' practice doesn’t give them this flexibility. Nevertheless, their un-paramaterized length concept is perfectly servicable and entrenched in their everyday life, engineering and science.
Some of the more reflective metricians start looking for a metaphysical grounding for the metricians’ length-concept. They ask for a story about where in the world to locate this relation that pairs that table, say, with the number 2. We know the sort of resources they could use—they could talk about the ratios between the length of the table and the length of one of their standard measuring sticks, and note that this is the same ratio as that between 2 and 1. They could show how the system of ratios can be represented by real numbers.
So the metricians have a successful up-and-running practice of length measurement and a story about the metaphysical basis of it. Does that mean that we should think that there’s an especially interesting relation between the table and 2, as opposed to between the table and 6.xyz? Surely not—after all, the imperialists in the neighbouring country could have a corresponding practice of reporting lengths, again without explicit reference to units, but which corresponds systematically to our measurements in feet and inches. There’s nothing wrong with the metricians and the imperialists ways of measuring lengths, or the metaphysics they offer. But their practices are parochial. When they use the relation that maps the table onto the number 2, they’re not talking purely about length—length is the intended subject-matter, but the relation they use to glom onto it has an admixture of an arbitrary choice of unit, in a way that isn’t explicitly acknowledged in their talk.
In contrast to the metricians and imperialists, the synthethisians don’t report lengths by appeal to numbers at all—not even via an explicit choice of unit as we do. They talk explicitly about the ratios and comparative distance relations between various objects. There’s nothing parochial about their talk—they capture the subject-matter directly (though it may from a practical point of view be a clunky way to proceed).
Might the same be true of representation? Might our representation talk (that I see this table or believe of it that it is solid) be similiarly parochial and embed an unacknowledged element of arbitrariness? This sort of questions has been raised before—Davidson used an analogy to the theory of measurement as a way into his concerns about indeterminacy, and Field in 1975 put his finger on exactly the issue as I understand it. But I think the message hasn’t been fully absorbed. The point, generalized, is that even if you having a successful theory of some aspect of the world, and even if you have a naturalistic and vindicatory metaphysics of the same, then there’s still a question about whether you are purely reporting on the facts about representationor whether there’s a perspectival element to the whole theory-and-metaphysics shabang. In short, there’s still a question about the objectivity of the theory of representation.
That’s got implications for the way we think about the philosophy and metaphysics of representation. By now, we know many candidate ways to defeat the letter of the Quine-Putnam-Davidson indeterminacy arguments, even if we may squabble over the details (shall we throw causation into the mix, or would an appeal to natural properties be better?). But suppose you’ve got your candidate—a naturalistic and vindicatory metaphysics of entrenched theorizing about representation. The real hard task is to say why appeal to causality or naturalness isn’t an arbitrary admixture to the theory, grounding only a parochial representation-relation—representation from a certain point of view. If you back down at this point, conceding that there’s no objectivity in the inclusion of causation rather than some permuted variant or naturalness vs its variant, then I think in spirit, if not in letter, you’re with the Quineans and Davidsonians, thinking there’s no real interesting sense in which we’re any more representationally-related to balls in front of our eyes than small furry creatures on another galaxy. It’s just as staggeringly implausible as it is in the original formulations. So the stakes for our conception of what we’ve achieved are very high.
There are in fact two really hard tasks here—first to articulate what being “purely about X” amounts to; and second to apply that understanding to the debate about representation, and see if which if any extant theories pass the test.
That is where debates about the explanatory role of representation becomes vital, in my view. I’ve been defending the line that this question of the objectivity of representation depends on teleology of the concept—the point and purpose which it serves for us. An arbitrary admixture to the metaphysics of X is one that can be varied while still achieving the purposes in question. And I’ve been arguing that considerations of charitable interpretation (as opposed to causality or naturalness) can be defended as non-arbitrary on these grounds.
LB: As you know, we are working on a project investigating the potential benefits of beliefs that are inaccurate in some important respect, so we find some aspects of your project very interesting and relevant. In particular, your project considers different ways in which content should be ascribed to beliefs in an interpretive framework, including informativeness, accuracy, and simplicity. Do you think accuracy should play a fundamental role?
RW: I’ve been fascinated since days in graduate school with underdetermination/indeterminacy arguments in the theory of representation. To illustrate with one I’m thinking about right now. Start from the following picture of the metaphysical grounds of belief and desire content: the content of an agent’s mental states is whatever it has to be, to rationalize the agent’s actions (the outputs) in the light of evidence (the inputs). An indeterminacy argument seeks to show that the constraints just mentioned can be met, not just by sensible-seeming interpretations, but by wild and crazy ones. The one I’m looking at right now, for example, constructs interpretations that have the agent thinking of themselves living in a universe that is void except in the local perceivable and manipulable environment of the agent (the agent’s “bubble”).
It’s fun to think up these crazy interpretations and show that they meet specified constraints. But what’s philosophically interesting is the lessons we draw. The most natural one—and my first instinct—is to draw the lesson that our account of the metaphysical grounds of content is incomplete, that we’re missing some `saving constraint’ that knocks out the crazy interpretations. And then the goal is to pin down what that missing ingredient could be. An alternative is to accept the arguments and revise our views about how determinate mental or linguistic content is. This strategy too raises interesting questions in the theory of representation—if we want to pinpoint what’s wrong with accepting indeterminacy of content, in my view we’ll need to say something about the theoretical role of representation, which pinpoints what we’d lose by embracing deviant interpretations. A final alternative—more popular historically than nowadays, perhaps—is to preserve the account of the grounds of representation but avoid indeterminacy, by rejecting presuppositions of the argument about the metaphysical character of the world that is being represented.
My graduate work focused on indeterminacy arguments and possible saving constraints—working in particular within the kind of `interpretationist’ metaphysical grounds for content David Lewis proposed (one element of which involves the rationalization story about mental content described above), which in certain parts of the philosophy literature has become almost orthodoxy, though rarely examined or defended in detail. The moral I drew from that work 10 years ago was largely negative—that living with indeterminacy wasn’t a real option because it’d prevent content from doing the work it needed to; and that the saving constraints that people (including Lewis himself) proposed didn’t work.
Maybe one moral at this point to have taken was that the Lewisian interpretationism was a busted flush, and to look instead at alternative traditions—salient among them being the causal-teleological accounts of Dretske and Mililkan, or (in the case of language rather than mental content) the ideas surrounding the causal theory of reference. But when I turned to look at those, the really compelling stuff that I could extract seemed to lack some of the virtues I prized in interpretationist accounts. Interpretationism, if it worked, would give a story about all content, not just special cases (reference to medium sized dry good and their observable properties). My hunch was that I wouldn’t find in these alternative traditions a satisfactory story about unsexy but genuine questions about what grounds the relation between the word `of’ and its semantic value, or about the grounds of content of highly theoretical beliefs remote from perception or action.
So this felt like unfinished business to me. For one thing, I’d said something about the way the metaphysics of representation wasn't—but the goal of course is to say something about how the metaphysics of representation is. Second—speaking personally—I struggled to find a presentation of the overall interpretationist picture that satisfied me as a real theory rather than a theory-sketch. The Lewisian story about linguistic content I thought I had a good grip on—but how the story about mental content went was much murkier. Thirdly, I was prepared to bet that some version of the interpretationist story was going to be part of the overall account of mental content—and since others in the literatures I was reading seem to be betting the same way, it seemed the division of philosophical labour required someone to go investigate whether that bet was going to pay off.
LB: In the first couple of years of your project, you were looking at whether the representational properties of belief have a genuine explanatory role. What sort of themes emerged as part of your research?
RW: One theme that emerged in the first couple of years was an argument about what is required to defend the spirit (as well as the letter) of a reductive realist theory of representation. I argue you need to do more than put forward a metaphysical theory that identifies where representation lies in the naturalistic world. You also need that metaphysical theory to be purely about representation. Let me expand on that.
We can illustrate the distinction here with a story about measurement. Suppose you have a community, the metricians, who report lengths simply by associating numbers with objects. So they would say “the length of the table is ” where we would say “the length of the table is 1 metre”. Since our measurement practice has an explicit role for a unit, we can easily report the same thing in other forms “the length of the table is 6 feet such-and-such inches”. The metricians' practice doesn’t give them this flexibility. Nevertheless, their un-paramaterized length concept is perfectly servicable and entrenched in their everyday life, engineering and science.
Some of the more reflective metricians start looking for a metaphysical grounding for the metricians’ length-concept. They ask for a story about where in the world to locate this relation that pairs that table, say, with the number 2. We know the sort of resources they could use—they could talk about the ratios between the length of the table and the length of one of their standard measuring sticks, and note that this is the same ratio as that between 2 and 1. They could show how the system of ratios can be represented by real numbers.
So the metricians have a successful up-and-running practice of length measurement and a story about the metaphysical basis of it. Does that mean that we should think that there’s an especially interesting relation between the table and 2, as opposed to between the table and 6.xyz? Surely not—after all, the imperialists in the neighbouring country could have a corresponding practice of reporting lengths, again without explicit reference to units, but which corresponds systematically to our measurements in feet and inches. There’s nothing wrong with the metricians and the imperialists ways of measuring lengths, or the metaphysics they offer. But their practices are parochial. When they use the relation that maps the table onto the number 2, they’re not talking purely about length—length is the intended subject-matter, but the relation they use to glom onto it has an admixture of an arbitrary choice of unit, in a way that isn’t explicitly acknowledged in their talk.
In contrast to the metricians and imperialists, the synthethisians don’t report lengths by appeal to numbers at all—not even via an explicit choice of unit as we do. They talk explicitly about the ratios and comparative distance relations between various objects. There’s nothing parochial about their talk—they capture the subject-matter directly (though it may from a practical point of view be a clunky way to proceed).
Might the same be true of representation? Might our representation talk (that I see this table or believe of it that it is solid) be similiarly parochial and embed an unacknowledged element of arbitrariness? This sort of questions has been raised before—Davidson used an analogy to the theory of measurement as a way into his concerns about indeterminacy, and Field in 1975 put his finger on exactly the issue as I understand it. But I think the message hasn’t been fully absorbed. The point, generalized, is that even if you having a successful theory of some aspect of the world, and even if you have a naturalistic and vindicatory metaphysics of the same, then there’s still a question about whether you are purely reporting on the facts about representationor whether there’s a perspectival element to the whole theory-and-metaphysics shabang. In short, there’s still a question about the objectivity of the theory of representation.
That’s got implications for the way we think about the philosophy and metaphysics of representation. By now, we know many candidate ways to defeat the letter of the Quine-Putnam-Davidson indeterminacy arguments, even if we may squabble over the details (shall we throw causation into the mix, or would an appeal to natural properties be better?). But suppose you’ve got your candidate—a naturalistic and vindicatory metaphysics of entrenched theorizing about representation. The real hard task is to say why appeal to causality or naturalness isn’t an arbitrary admixture to the theory, grounding only a parochial representation-relation—representation from a certain point of view. If you back down at this point, conceding that there’s no objectivity in the inclusion of causation rather than some permuted variant or naturalness vs its variant, then I think in spirit, if not in letter, you’re with the Quineans and Davidsonians, thinking there’s no real interesting sense in which we’re any more representationally-related to balls in front of our eyes than small furry creatures on another galaxy. It’s just as staggeringly implausible as it is in the original formulations. So the stakes for our conception of what we’ve achieved are very high.
There are in fact two really hard tasks here—first to articulate what being “purely about X” amounts to; and second to apply that understanding to the debate about representation, and see if which if any extant theories pass the test.
That is where debates about the explanatory role of representation becomes vital, in my view. I’ve been defending the line that this question of the objectivity of representation depends on teleology of the concept—the point and purpose which it serves for us. An arbitrary admixture to the metaphysics of X is one that can be varied while still achieving the purposes in question. And I’ve been arguing that considerations of charitable interpretation (as opposed to causality or naturalness) can be defended as non-arbitrary on these grounds.
LB: As you know, we are working on a project investigating the potential benefits of beliefs that are inaccurate in some important respect, so we find some aspects of your project very interesting and relevant. In particular, your project considers different ways in which content should be ascribed to beliefs in an interpretive framework, including informativeness, accuracy, and simplicity. Do you think accuracy should play a fundamental role?
RW: That’s an interesting question! Davidson thought one of the norms of interpretation was to make the target a “believer of the true, a lover of the good”. By contrast, Lewis doesn’t have this emphasis on *truth* or *goodness* in his official account of interpretation. For him (as I read his “Radical Interpretation”—the passage is a bit subtle) it’s all about rationality: depicting the believer as responding rationally to evidence (that’s his formulation of charity) and as acting rationality (that’s his rationalization principle). Attributing more accurate but *unjustified* beliefs, is no gain from this perspective.
How could we resolve this debate about what the norms of interpretation should be? The kind of perspective outlined above suggests that we go back to the question of the point and purpose of interpreting and ascribing beliefs in the first place. I think there’s a lot to be said for rationality as tied up with the telos of ascribing attitudes to people. If a person is rational, then the sort of person you can engage with in a distinctive way, as part of a process of joint inquiry or action.
The distinctive engagement is by presenting them with reasons to believe or do something---and only if they're rational (at least to the relevant extent) can you trust them to respond to reasons in the right kind of ways. So detecting a rational pattern is detecting an opportunity for a distinctive kind of person-to-person engagement. That kind of thing isn’t secured just by accuracy—you might pat someone on the back for getting things right, but if you think of someone has having hit on the right idea by pure chance, the preconditions for full collaborative engagement with them on theoretical inquiry and practical projects haven’t been met.
If we do think that rationality has the primary role here, perhaps there’s room for accuracy as supplementary. After all, in order to engage in joint cognitive or practical projects with another person, being able to interact with them in rational ways is required, but you’ll typically also want some shared starting points. And interpreting others as accurate (by your lights) is exactly to identify common ground. So—tentatively—I think that both rationality and accuracy have a fundamental role in interpretation. But rationality is more fundamental than accuracy: an absence of rationality means that you can’t engage with them *at all* in the ways we want; an absence of accuracy (by your lights) just means that it’s harder to do so.
(Just to mention another theme that lurking in this area. Williamson has suggested that the right way to do something like radical interpretation is to maximize (wait for it!) knowledge, rather than true belief. And you can see why that might be related to the telos of interpretation, too, if we’re interested in interacting with others as part of a project of inquiry aimed at knowledge. Since knowledge entails truth (and justification) this kind of story might subsume the other two.)
Having said that, let me add an important caveat. From my perspective it’d certainly be legitimate to engage in the project of interpretation for purposes *other* than that of finding people with whom one can rationally collaboratively engage in inquiry, or practically collaboratively engage in joint projects to bring about the good. You could engage in interpretation of another because you want to predict what they’ll do next. Those different purposes would support different constraints on interpretation. So while normative stuff like accuracy and rationality may be fundamental relative to one kind of interpretative exercise, that doesn’t mean they’re fundamental for everything we might be doing around here.
If we do think that rationality has the primary role here, perhaps there’s room for accuracy as supplementary. After all, in order to engage in joint cognitive or practical projects with another person, being able to interact with them in rational ways is required, but you’ll typically also want some shared starting points. And interpreting others as accurate (by your lights) is exactly to identify common ground. So—tentatively—I think that both rationality and accuracy have a fundamental role in interpretation. But rationality is more fundamental than accuracy: an absence of rationality means that you can’t engage with them *at all* in the ways we want; an absence of accuracy (by your lights) just means that it’s harder to do so.
(Just to mention another theme that lurking in this area. Williamson has suggested that the right way to do something like radical interpretation is to maximize (wait for it!) knowledge, rather than true belief. And you can see why that might be related to the telos of interpretation, too, if we’re interested in interacting with others as part of a project of inquiry aimed at knowledge. Since knowledge entails truth (and justification) this kind of story might subsume the other two.)
Having said that, let me add an important caveat. From my perspective it’d certainly be legitimate to engage in the project of interpretation for purposes *other* than that of finding people with whom one can rationally collaboratively engage in inquiry, or practically collaboratively engage in joint projects to bring about the good. You could engage in interpretation of another because you want to predict what they’ll do next. Those different purposes would support different constraints on interpretation. So while normative stuff like accuracy and rationality may be fundamental relative to one kind of interpretative exercise, that doesn’t mean they’re fundamental for everything we might be doing around here.