Diagnosis precedes prescription

The title principle seems obvious enough. Which makes it all the more puzzling that most normative political theorists ignore it in practice. Why is this? What are the implications?

Let me elaborate. Consider the global poverty literature. Two classes of prescriptions are made here. Examples from the first class — call them “general prescriptions”, for lack of imagination — include:

  • “We (citizens of affluent countries) ought to alleviate global poverty.”
  • “We ought to take care of domestic poverty before dealing with global poverty.”
  • “We ought to work toward a more equal global distribution of wealth.”
  • “We ought to bring people up to some minimum threshold of well-being but no more.”
  • While I think diagnosis of a sort precede these sorts of prescriptions, this might be more controversial then the principle I had in mind when I wrote the title to this post. Anyway, these aren’t the sorts of prescriptions referred to in the title.

    Call the second class of prescriptions “specific prescriptions” (again, for lack of imagination). Examples include:

  • “We ought to donate more money to NGOs who are working to alleviate global poverty.”
  • “We (our governments) ought to give more development assistance to poor countries.”
  • “We ought to tax international financial transactions/carbon emissions/sales of resources extracted from the global commons and use to revenue to finance development in poor countries.”
  • “We ought to make international trade fairer by, e.g., abolishing agricultural and industrial subsidies in developed countries.”
  • It’s this second class of prescriptions that I’m interested in. It’s this second class of prescriptions that must be preceded by diagnosis.

    Diagnosis is the practice of explaining an outcome by identifying the causal process generating it. The sort of diagnosis with which many of us are most familiar is medical diagnosis. The aim of medical diagnosis is to identify the causal process producing the medical abnormality. Most obviously, this requires finding a causal process that can explain the medical symptoms. But (as any dutiful viewer of House will know), the symptoms aren’t the only relevant data points. The symptoms that are absent can be just as informative as the symptoms that are present. The point is this: medical diagnosticians seeks to identify the causal process generating the data so that they can identify the places in the causal chain where intervention is most likely to be effective, as well as identify the intervention possibilities most likely to be successful.

    Similarly, one would think that identifying the causal process generating problematic social outcomes we observe — like global poverty — would be a key step in the task of proposing intervention. Once we’ve identified our general duties — suppose we have an obligation to alleviate global poverty — it would be helpful to know what causes global poverty so that we can figure out how to most effectively discharge our general duties. Indeed, the point seems obvious and is likely obvious to most political philosophers. Yet, the practice of political philosophers shows, at worst, a startling disregard for the importance of the diagnostic task, at best, a naïveté about what is required of a diagnosis.

    To avoid making this too long, I’ll give an example of each sort of flaw and say a few things about what is required of a diagnosis. The examples:

    Peter Singer has, for nearly 40 years, been a prominent advocate of charitable donations to NGOs. Specifically, he recommends that residents of affluent countries give a modest percentage of their income to aid organizations according to a sliding scale based on income. (See ch. 10 of Singer 2009 for details.) This prescription is motivated by Singer’s oft-cited “drowning child” thought experiment: You’re on your way to an important meeting wearing a nice suit and shoes; you see a child drowning in a shallow pond; the child’s life is vastly more important than your suit and shoes; you ought to wade into the pond and save the child.

    A key feature of Singer’s motivating thought experiment is the absence of background context. Most importantly, we don’t know how the child came to be drowning in the pond in the first place. Such information might be unnecessary when determining whether we should wade into the pond to pull out the child because it doesn’t matter how the child got there; it’s immediately obvious that the way to save the child is to pull her out of the pond. But global poverty is not nearly so easy to figure out. And this is where the analogy to the drowning child breaks down. The drowning child scenario is an emergency; most cases of global poverty are chronic. Emergencies are relatively easy to solve: if possible, remove the immediate threat. Chronic cases are much more difficult to solve. They’re like weeds in a garden: to prevent them from coming back, we need to pull them out by the roots. Thus, if we are to figure out how to prevent global poverty, it matters how people became impoverished. Singer’s failure to recognize this means he is effectively shooting first and aiming later (if at all) with his prescriptions. Sure, some might hit, but this seems a terribly ineffective way of treating an illness.

    Leif Wenar (2008) proposes that we alleviate global poverty by enforcing international property rules. Affluent countries should cease importing resources (such as oil and mining products) from corrupt and authoritarian rulers (who have no legal right to sell these resources without their constituents’ consent), place tariffs on imports from countries who purchase resources from such states (as, e.g., China does from Sudan), and place the revenue generated by the tariffs in a “Clean Hands” trust fund, whose contents are to be returned to the citizens of states like Sudan once they establish a democratic government. Wenar’s proposal is based on the following diagnosis of global poverty. Social scientists have noted a correlation between natural resource abundance and negative development outcomes (the so-called “resource curse”). This correlation seems to be explained by three “causal pathways”: authoritarianism, increased civil war incidence, and low economic growth rates. Each of these are statistically correlated with both resource abundance and negative development outcomes. Thus, according to Wenar, if we can break the connection between resource abundance on the one hand and authoritarianism, civil war, and low growth on the other, we should be able to break the connection between resource abundance and negative development outcomes. Resource income can then be invested in development and poverty will be overcome in resource abundant countries, which is where a significant percentage of the global poor reside.

    Wenar should be praised for doing his social science homework. But identifying a series of correlations doesn’t amount to an adequate diagnosis. The correlations identified by Wenar aren’t “causal pathways”, so they don’t explain the resource curse. They are simply additional correlations that themselves require explanation. An adequate diagnosis would explain the data; it would identify the causal process that connects resource abundance with authoritarianism, civil war, low growth, and poverty. Wenar’s diagnosis is akin to delineating which symptoms are correlated with which diseases. This can be useful if these correlations lead to further insight about the operative causal processes. The correlations identified by Wenar give only limited such insight. But without this further insight, we’re left treating symptoms.

    I’ve already hinted at what I think counts as an adequate diagnosis. Before we make specific prescriptions about how to best discharge our general duties, we need to identify the causal processes generating the outcomes we seek to address. There are some things to be said here about appropriate engagement with and application of social scientific research, but we can leave further discussion of the details to the comments thread.

    Two further things we can take up in comments:

    First, why do normative theorists neglect the diagnostic task? My guess is that is has something to do with an appeal to some notion of the “appropriate division of labour”, where diagnosis is farmed out to social scientists. I think this is untenable, but we can talk about it.

    Second, what are the implications for philosophers’ proposals of neglecting the diagnostic task? I think it puts us in the position of a blindfolded archer: a desire to shoot but no idea where to aim. Unfortunately, our blindfolds haven’t prevented us from shooting. Not surprisingly, this dramatically reduces the chances of making effective prescriptions.

    Advertisements

    11 Responses to Diagnosis precedes prescription

    1. Shen-yi Liao says:

      Does Singer intend for his prescription to be a solution to global poverty? The analogy certainly makes it sound more like the prescription is how to immediately alleviate the human suffering associated with, and likely caused by, global poverty right now.

      It’s funny that you mention House. On the show, often the prescriptions are themselves part of the diagnosing process. Time’s short and people’s lives are on the line, let’s try whatever MIGHT work and see which does. The human sufferings associated with global poverty certain bear resemblance to House’s patients in this respect. In which case, perhaps the philosophers aren’t in the position of a blindfolded archer, but in the position of a cranky but oh-so-attractive doctor: shooting toward some broad direction is a way of narrowing down the target. Moreover, it is the only way we have that satisfy other important desiderata, such as having relatively good chance to alleviate human suffering NOW.

    2. Shen-yi Liao says:

      Of course, a relevant disanalogy is that where as doctors are the ones who are doing both the prescribing and the executing, the normative theorists are mostly only doing the latter. What is involved in turning normative proposals into policies, however, might themselves again introduce desiderata for the prescriptions to satisfy.

    3. David says:

      The Singer point is fair enough. Say he only means aid donations to alleviate immediate human suffering. There remains the question of whether the proposed intervention effectively achieves its objective. There’s a growing body of evidence that suggests that aid programs don’t generally reduce the amount of suffering associated with poverty, while it has been effective in particular cases. But whether it does so depends on whether the intervention targets the causal process at places where it is likely to be effective. Determining that requires adequately diagnosing the problem.

      I like the point about House using speculative interventions as part of the diagnostic process. Although (importantly, I think) I see his motivation as less “these people are in trouble; we need to do whatever we can to help them out” and more “if we try this treatment, we are likely to generate new data that will help us make a better diagnosis”. The practical implications of the two are importantly different. The former attitude gives up on diagnostics and just starts trying interventions in the (sometimes blind) hope that something works; the latter insures that the diagnostic phase never stops. Because I think social causal processes are complex enough to preclude the possibility of an ex ante complete diagnosis, I’m in favour of experimental interventions that are motivated by the latter attitude. My criticism of philosophers, then, is that their prescriptions are often motivated by the former attitude. That is, philosophers’ sense of moral urgency often leads them to infer from “we need to do something” that “we ought to do anything that sounds like it might plausibly work”. This inference is a mistake. Precisely because these problems are morally urgent, we should commit to being earnest about the diagnostic task so that we can cut down on the amount of wasted effort, time, and resources poured into interventions we could have seen would be ineffective had we done a little more homework.

    4. Shen-yi Liao says:

      Those responses seem fair and I am generally on board with what you say. I just want to push from a different angle.

      It is possible that no individual philosopher work on the topic take diagnosis as seriously as they should, but collectively they do. To continue on our extended House analogy, it would be like when House’s team throw out ideas and argue against each other from available evidence. I am unfortunately ignorant of the literature. What I would like to know is whether diagnoses, in your sense of identifying relevant causal paths, are ever used to disprove proposed theories? For example, Philosopher A proposes Prescription A, and Philosopher B brings up Evidence B to show that Prescription A won’t do the work intended. If that happens, maybe that is some reason to think that collectively, philosophers do take diagnoses into account. And while it would come after the initial prescriptions, it would help the participants of the debate to converge on a final (or at least more considered) prescription that gets put into policy.

      • David says:

        “What I would like to know is whether diagnoses, in your sense of identifying relevant causal paths, are ever used to disprove proposed theories?”

        Yes. Pogge’s work provides the clearest example of this. He thinks that too many philosophers prescribe various sorts of assistance to poor countries, which is based on their adoption of the “purely domestic poverty thesis”, viz., that cross-national variations in development outcomes are explained entirely by domestic factors. But, he argues, this conceals the fact that international institutions are important for shaping development outcomes, so assistance is insufficient for addressing global poverty. What we need instead (or in addition?) is international institutional reform.

        While I think Pogge has done the debate a great service, “international institutions are important too” is not a diagnosis that can ground any prescription. International institutions could matter in any number of ways, each of which implies different interventions. To be sure, Pogge has specified some of the ways that institutions matter. One example is his criticism of the “international resource privilege”, which permits any group who holds a monopoly over the use of force in a territory to be the legally recognized vendors of the natural resources in that territory. Pogge claims that this international rule harms the poor by creating an incentive for coups and enabling authoritarian leaders to consolidate power. While this looks like a diagnosis, I think it’s not, not yet anyway. Why? Because there are several paths from resource abundance to poverty that go through “incentive for coups and ability to consolidate authoritarian rule”. Which of these paths is the operative one? That’s the question a diagnosis should answer.

        But I get the spirit of your response, viz., that as a body, perhaps philosophers are progressively sharpening the diagnosis and thus making better prescriptions. I think this is partly true; the Pogge example shows that philosophers do occasionally point to potentially important causal factors that were previously overlooked. And this is definitely pushing prescriptions in the right direction. I guess my complaint is one about the resolution of the diagnoses provided. The causal claims philosophers make (when they do so) are not nearly sharp enough. I’ll elaborate on this resolution point below in my response to Jonathan.

    5. Jonathan Livengood says:

      Thanks for an interesting, albeit puzzling, post. I have a bunch of questions. Which global poverty literature are you talking about? A quick glance at the social science articles cited by Wenar — thanks for the reference, by the way — shows that the economists and political scientists working on these questions *are* interested in causation and report their results in causal terms. It is true that the first article cited by Wenar (Wantchekon 2002) reports a simple correlation, but many other papers he cites make causal claims. (See Ross 2001, 2006; Jensen 2004; Collier 1998, 2004; and Sachs 2001; all cited by Wenar.) Do you object to the particular statistical techniques used by the social scientists? I’d really like to hear your positive proposals for discovering the relevant causal structures. Do you propose conducting controlled experiments in the third world? That might be what Greg House would do, but would it be ethical? (I’m not even sure that House’s use of experimentation is ethical on individuals, let alone if we ramp up to experimentation on entire nations.) And would it be feasible? I somehow doubt that we could get enough control of third-world nations to run the required experiments, especially not if we’re actually required to get informed consent.

      • David says:

        First, the quick answers. I’m talking about the normative political philosophy global poverty literature, not the social science global poverty literature. I like statistical techniques just fine, although I occasionally have criticisms of particular model specifications. I’m not proposing controlled experiments in the third world, not least because I’m skeptical about how useful they would be due to problems associated with trying to scale up the results.

        Now, on to the real issue. I agree that the social scientists Wenar cites make causal claims. I’d even go further than that and say that Wenar makes causal claims. (The relevant part of Wenar’s paper here is sec. 1.) But I never denied that Wenar or the papers he cite make causal claims. I presented the Wenar paper as an example of philosophers’ naïveté about the requirements of diagnosis. In other words, I don’t think Wenar makes the kind of causal claims that could count as adequate diagnosis. My main problem is with the resolution of Wenar’s causal claims. They’re not sharp enough.

        What do I mean by “not sharp enough” in this context? Without much argumentation, I think that an adequate explanation of a social outcome identifies two things: (1) the individual-level behaviour(s) that generate(s) the social-level outcome, and (2) the aggregative process by which the relevant behaviour is translated into social outcomes. To illustrate this point (and simultaneously show where I think Wenar goes wrong), I’ll press on Wenar’s diagnosis a little.

        I think Wenar’s discussion of resource abundance and authoritarianism (p. 3f) comes closest to the model of diagnosis I have in mind, so I’ll focus on that. (The civil war and low growth discussions are pretty easily shown to be wide of the mark I have in mind.) First, Wenar sets the backdrop with the claim that resources generate massive revenue for the vendor. Against this, Wenar makes the following causal claims:

        (1) Dictators can use resource revenue to buy whatever he needs to prevent successful domestic challenges to his rule.

        (2) Resource revenue can release dictators from the need to raise revenue through taxation, undermining an important accountability mechanism.

        (3) Dictators can use resource revenue to buy political support, thereby keeping himself in power.

        These claims are headed in the right direction (according to me). But they’re not quite there yet. First, each of the three claims identifies a distinct causal mechanism — a fact Wenar either misses or ignores. This is important because different causal mechanisms imply different interventions. (Cf. Ross 2006, pp. 266, 280 onward on this point.) I’m not saying all three couldn’t be operative; they might be. But Wenar presents these as possible mechanisms linking resources with authoritarianism. (That’s how I’m reading the “can” in each of the claims.) Some might call this nitpicking, but I don’t think it is. The fact that Wenar is content to identify the menu of possible mechanisms linking resources with authoritarianism suggests that he doesn’t think it matters how the two are linked. But to determine which intervention is likely to be effective, that’s precisely the information we need. We don’t merely want to know the menu of causal mechanisms that are possibly operative.

        But this isn’t my main criticism. My main criticism is that Wenar’s causal claims aren’t yet sharp enough. To get sharp enough, an adequate diagnosis would further identify the mechanism(s) that enable(s) a government leader to use resource revenue to buy the requisite tools for putting down domestic challenges or buy political support, or the mechanism(s) that enable(s) resource revenue to undermine a leader’s accountability. (To be clear, by “mechanism”, I mean the two things I identified above as components of an adequate social explanation: (1) the individual-level behaviour(s) that generate(s) the social-level outcome, and (2) the aggregative process by which the relevant behaviour is translated into social outcomes.)

        The need to come up with sharper explanations is motivated by the recognition that resource revenue doesn’t uniformly result in leaders (violently) suppressing domestic challenges or avoiding accountability or buying political support. For example, Norway and Canada are among the top 15 oil exporters; Norway, Canada, Netherlands, U.S., and Australia are among the top 15 natural gas exporters; and Australia, U.S., Canada, Poland, Czech Rep., and Netherlands are among the top 15 coal exporters. So at least Norway and Canada, and perhaps also the U.S. and Australia are pulling in considerable resource revenue, on par with many countries who suffer from the resource curse. And yet none of these countries suffer from the resource curse. This raises the question: Why does the causal process that links resource abundance to authoritarianism operate in some places and not others? What causal mechanism(s) enable(s) resource abundance to lead to authoritarianism in some instances and not in others? The answer to these questions yields a much sharper diagnosis than the causal claims offered by Wenar.

        This response has become pretty long, so I’ll wrap up with a few comments indicating the direction I’d like to see things go (and open up further discussion).

        First, I think we’ll get sharper diagnoses if we start to employ more formal game-theoretic models and then test the observable implications of those models. Such models make the causal logic much more explicit and enable us to identify much more clearly points in the causal chain where intervention is likely to be successful.

        Second, because I haven’t yet said this, my (tentative) diagnosis of global poverty is based the “selectorate model” of Bueno de Mesquita et al. (2003). This says that the number of people on which a leader depends to stay in office (the “winning coalition”) and the ratio between the size of the winning coalition and the “selectorate” (the set of people who could possibly wind up in the winning coalition) are important determinants of what a leader does with the revenue at his disposal — whether public goods that are important for development are provided; whether the leader steals state money for himself; etc.

        Third, the danger in asking for ever-sharper diagnoses is that I could be asking that we identify something like “causal foundations” where there are none to be found. For example, my preferred diagnosis of global poverty relies national selection institutions as a key causal factor. But cross-national variations in selection institutions need to be explained. Why doesn’t the mechanism that generates this variation provide the “ultimate” diagnosis of global poverty? I’ve only started thinking about this question recently, so I don’t have much to say yet. And this is officially too long, so I should stop.

    6. Jonathan Livengood says:

      David,

      I agree with your main claim — diagnosis first — at least in the kinds of cases you are considering. (In some other cases, one might expect treatment of symptoms to be the right course of action, say after a natural disaster.) And I like your point about resolution, though I’ll make some qualifications below. I’m still not clear on your proposal for *discovering* the fine-grained causal structures that you want. Could you give an example of the application of game theory to the problem?

      As to causal foundationalism, I think there is a prior problem. Long before you get to anything foundational, you will have lost your ability to make sense of proposed interventions — losing sight of the forest for the trees. That is, supposing you could fine-grain all the way down to fundamental physics, why would you want to? The problem of resolution, then, is to find the resolution appropriate to the policy problem at hand. I’m not sure what criteria one should use to determine the appropriate graining, but I don’t think it can be “get as fine-grained a picture as you possibly can.” And there is a related point. Suppose we know that increasing %GDP from natural resource exports for a nation causes increasing individual poverty for that nation. If we can intervene on %GDP, why do we need to know (for policy purposes) the mechanism in any more detail? One might worry, I suppose, that intervention will (or might) have negative consequences for another important variable — say individual liberty. But I’m not sure how having a more *fine-grained* model solves this problem. Rather, it seems that in order to answer that worry, one needs a more *complete* causal structure.

      Evidently, Ross (2006) agrees with you that the causal mechanisms haven’t been adequately specified (though he takes his work to be an improvement). He writes: “The fourth problem is the failure to determine the causal mechanisms that link mineral wealth to war. Different scholars offer different theories … But due to the aforementioned problems — measurement, endogeneity, and robustness — and due in part to a shortage of data, we have not been able to tell which mechanism (or mechanisms) is correct. Because each causal mechanism implies a different set of policy interventions, getting the mechanism right is critical.” (266) But this makes me wonder … if the social scientists are making these points and attempting to give us causal models explicitly for the purposes of policy prescriptions, what are normative political philosophers adding to the discourse? I’m simply ignorant here. I do work on causal inference and philosophy of social science, but I haven’t read anything in political philosophy in several years.

      • David says:

        Thanks for your further reply, Jonathan. I’ve only got a few minutes, so I’m going to set aside (for now) the first and third paragraphs of your response and just respond to the second for now.

        I agree that there’s an appropriate level where we can stop for the purposes of social explanation. Going all the way to fundamental physical particles is both unnecessary and unlikely to help us with the policy questions. That’s not quite the issue I was wanting to raise with the question of causal foundationalism. The issue I wanted to raise was that it seems like any social causal variable determined to be salient could have a further explanation at the same level, which could then be said to be “the real explanation”. Returning to the selectorate model. I think that the selectorate model uses differences in national selection institutions to arrive at a compelling individual-level explanation for a variations in social outcomes, including development outcomes. But the cross-national variations in selection institutions likely has a further individual-level explanation (rather than a fundamental physical particle explanation), which some might be inclined to see as the “real explanation” of global poverty. Does this make any sense?

        I think that staying at the level of %GDP from resource exports is too “coarse-grained”. Supposing with you that we know there’s a correlation between %GDP and individual poverty. This connection could arise as a result of any number of individual-level behaviours. Here are two:

        (1) High resource exports from Oiland increases demand for Oiland’s local currency, which increases the value of Oiland’s currency relative to the currency of their trading partners. As a result of the exchange rate increase, the price of Oilands’s manufacturing exports goes up, reducing demand for Oilandian manufactured goods. Since resource extraction can be done by relatively few workers,the collapse of the Oilandian manufacturing sector results in high unemployment. Since resource extraction can be done by low-skill workers, the government has no incentive to invest in public goods like health care and education, wiping out any social welfare net for the recently-unemployed manufacturing workers. A large percentage of the population plunges into poverty. (This is the so-called “Dutch Disease”.)

        (2) High %GDP from resources leads the Oiland government to nationalize the resource sector, putting the resource revenue directly in the government’s hands (instead of private hands). Oiland’s ruling elite then use this revenue to buy political support from, e.g., senior military leaders, who subsequently use force to insure the results of a rigged election; or from key tribal leaders, who deliver a bloc of votes. In any case, the number of people on whom the ruling elite depend for support to remain in office is effectively decreased. Since they rely on few people to remain in office, Oiland’s government has little incentive to invest in public goods or invest in development. As a result, a high percentage of people plunge into poverty. (This roughly follows the selectorate model mentioned above.)

        Since %GDP and poverty could be correlated by at least two distinct causal processes (I’m sure there are others), and since each causal process implies a distinct intervention, we need a diagnosis that identifies the causal processes at work at the level of the policy intervention.

        As I noted above, I think it’s the level of individual behaviour (what Elster calls the “microfoundations” of a social outcome) at which our explanations should reside. One reason (among others) is that political and/or economic policy interventions aim to alter individual behaviour. There is no way to alter the effect of %GDP on poverty without altering the way individual politicians, investors, workers, consumers, etc. respond to the policy. It’s these responses that a policy is aiming to shape, and it’s these responses that any policy proposal must account for.

        I’ll return to the points you raise in the first and third paragraphs later.

    7. Jonathan Livengood says:

      Dave,

      What I asked you to suppose was that %GDP from natural resources is a *cause* of individual poverty, not that it is merely correlated. You are right to say that different mechanisms might recommend different interventions, but from your descriptions of the mechanisms, I think these can only be on-path (from %GDP to poverty) interventions. If %GDP is really a cause and if one can intervene on it directly, then one knows enough of the mechanism to implement a successful policy.

      Maybe there are good reasons to think that %GDP cannot be intervened on directly? Still, if it is a cause, then what we would look for are causes of %GDP that *can* be intervened on directly. Having found one, provided we don’t care about side-effects from such a manipulation (or are convinced there are none or are convinced they aren’t too bad), we intervene to indirectly set %GDP. Then %GDP has its downstream effect on individual poverty.

      Or is your worry tied to discovery? Maybe you think we cannot know that %GDP is a cause of poverty until we know the mechanism connecting the two? I think this brings us back to (1) why you think the methods employed by social scientists–who think they are identifying causes, not mere correlations–aren’t good enough, (2) what criteria you have for picking out the resolution at which to observe the mechanism, and (3) what your positive alternative for causal discovery is.

      I’m guessing you were going to say something about some of this in your reply to the two points you didn’t cover in your last response, but I’m evidently not patient enough to wait!

      • David says:

        Hi Jonathan. Thanks again for your comments. Sorry for my delay.

        I have no idea what it would mean for %GDP from resources to be an unmediated cause of poverty. I happen to think that there are no direct social causal relations between macro-level social entities, such as %GDP from resources or poverty in the aggregate. The only social causal relations are those between individuals, each of whom is pursuing his goals (some of which coincide and some of which conflict) and adjusting the course of his pursuit in response to the behaviour of others. The macro-level outcomes are the result of aggregated individual-level behaviour. In the words of Thomas Schelling: “If we see pattern and order and regularity,… inquire first of all what it is that the individuals who comprise the system seem to be doing and how it is that their actions, in the large, produce the patterns we see” (Micromotives and Macrobehavior, p. 22). On my view, if there is a causal relationship between %GDP and poverty, it is mediated by individuals’ behaviour.

        Thus, I agree that we can’t know that %GDP has a causal relationship to poverty until we know the individual-level behaviour that connects the activity that results in a particular %GDP from resources with the activity that generates poverty. So, as you say, this brings us back to your (1), (2), and (3). I’ll respond to each in turn.

        Per (1): I don’t think the methods employed by social scientists are inadequate. I must have been unclear on this point if that’s the impression you got. By and large, I think they’re right on target. My post was meant to expose the following blindspot in the philosophical global justice literature, as I see it. Philosophers typically spend most of their effort making normative arguments about what sorts of institutional arrangements count as just in light of certain moral considerations. Many of them then proceed directly from the normative argument to institutional design proposals they ostensibly intend to be implemented in the actual world. Some recognize that social science might have some important insights here and so they do their duty and throw in some token references to some political science papers before making their prescriptions. The few really diligent ones do their social science homework and identify some important correlations but stop short of paying attention to what I take to be the really important contribution of social science: the identification of social causal processes. In other words, my post was admonishing philosophers who fancy themselves to be in a position to offer institutional reform proposals without adequately paying attention to the social scientific task of diagnosing the (morally) problematic outcome they seek to avert.

        Per (2): We should pick out mechanisms at the level that is most helpful for explaining social outcomes. I already said a little about this at the start of this reply. Since there are no social causal relations at any level greater than (i.e., more macro than) the individual level, the resolution must be at least that fine. Should we make the resolution any finer? I think not. My reason for thinking this is that going more fine-grained than individual behaviour won’t add anything to our explanation of the social outcome. Social outcomes are the result of aggregated individual behaviour. Perhaps individual behaviour is ultimately best explained by neurological and biochemical processes. That’s fine; then, ultimately perhaps, social outcomes are explainable in terms of neurological and biochemical processes. But I don’t think such an explanation is any more illuminating than the one given in terms of individuals; in fact, it is probably obfuscating. It seems to me we’d want translate the operative neurological and biochemical processes into their individual behavioural effects before going on to connect them to the social outcome. But this response is sort of off-the-cuff and less principled than one might like.

        Now, finally, on to (3). I’ll give a quick sketch of why I think formal game theoretic models are useful for illuminating social causal mechanisms. I’ve already said that I think social outcomes are the result of the aggregated behaviour of goal-directed individuals pursuing their goals and adjusting their behaviour in response to the behaviour of others. In other words, social outcomes are the aggregate result of strategic interactions between goal-directed individuals. And that’s what game theory deals with — strategic interactions between goal-directed individuals. The advantage of formally modeling these interactions is it forces us to be explicit about our premises: Who are the players? What are their goals? What actions are available to them? How are the players’ interactions structured? Formal models also enable us to use logic to deduce the consequences of various combinations of actions and make conjectures about which outcomes are most likely given certain premises about the players’ goals. Thus, models enable us to say that outcome O occurs when these players have these goals and act in these ways in pursuit of their goals. In other words, game theory enables us to explain social outcomes in terms of players, goals, actions, and interactive structure — i.e., as the aggregate result of strategic interactions between goal-directed individuals. And since I think that the latter is all there is to social causation, I think game theory enables us to illuminate social causal mechanisms. (By the way, while I think formal models are particularly useful, I don’t think all game theoretic arguments need to be formalized. I’m happy enough when the arguments are explicit about the players, goals, actions, and interactive structure and use compelling logic to get from these inputs to the social outcome.)

        Because game theoretic explanations must be explicit about players, goals, acts, and interactive structure, and because we can alter the model and compare the results, game theory has the additional advantage of enabling us to identify which interventions could be effective and which ones would likely be ineffective.

        I guess I still owe you a response to your question about what normative philosophers could add to the discourse about diagnosing social outcomes. I’ll have to get back to this one. I hope you still have some patience left…

    %d bloggers like this: