Ethics Review Forum: John Oberdiek: ‘Imposing Risk: A Normative Framework’. Review by Madeleine Hayenhjelm

11 Replies to “Ethics Review Forum: John Oberdiek: ‘Imposing Risk: A Normative Framework’. Review by Madeleine Hayenhjelm”

1. Let me thank Madeleine for her generous and challenging review of Imposing Risk and the editors at PEA Soup for providing this forum.

   Madeleine identifies three ideas that form the “backbone of the overall framework” that I defend: (1) understanding risk to be the probability of harm; (2) taking cases of pure risk imposition–those that do not result in the harm that they risk, instill fear in those put at risk, nor otherwise disrupt the lives of the people exposed to risk–as the central moral puzzle; and (3) locating the primary moral significance of (pure) risk impositions in their negative effect on people’s autonomy.

   Although Madeleine expresses concerns about (1), saying that defining risk as probability of harm is “slightly too quick,” most of the concerns she raises bear on (3), my view that the primary moral significance of imposing risk consists in the detrimental effect it can have on the autonomy of those subject to the risk.

   Madeleine raises three worries about my focus on autonomy:

   First, she maintains that not all intuitively problematic risk impositions affect autonomy negatively. What about adding options to one’s option-set which are themselves risky? Second, Madeleine contends that not all intuitively problematic risk impositions affect autonomy or autonomous agents full stop, and if that is the case, assessing climate change risks, risks imposed on children, or on those in a coma, all would demand different accounts. Third and finally, if the primary moral significance of a risk imposition consists in its potential deleterious impact on autonomy, then this should “point to a defining difference between acts that impose a risk of harm, on the one hand, and acts that mitigate or counteract harm, on the other.” And yet mitigating risk can curtail autonomy more than the risk whose mitigation is sought—Madeleine cites as an example autonomy-curtailing curfews that might appropriately be put in place when there are high levels of air pollution.

   To substantiate her first worry, that not all morally problematic risks affect autonomy negatively, she imagines the risky boat rides offered by human smugglers to migrants. The case is complicated in a sense. There’s a lot wrong with refugee smuggling, after all, so it is hard to isolate the contribution that risk imposition makes to the scenario. But if we can in fact just focus on the option that is offered, I see no reason why I can’t reject it as exploitative. I certainly do not want to endorse a conception of autonomy whereby autonomy is enhanced by an option that looks attractive only if one is desperate. But I also concede that I do not provide an account of exploitative options that would permit me to draw the distinction I’m appealing to in a principled way. Madeleine’s initial example of a rickety bridge that’s built between two otherwise unconnected remote regions is more illuminating in this sense, as it strips away the exploitative aspect of human smuggling, and may show that exploitative “options” are actually just one species within the genus of options that might not be worth having. When I think about the rickety bridge example, I’d like to know whether it could have been built more safely—is it really a safe option?—and how important it is to connect these otherwise unconnected regions—is this what tort lawyers call an “attractive nuisance” or otherwise exploitative after all.

   Madeleine’s second concern is that my autonomy-based account cannot make sense of what is morally problematic about a wide range of risk impositions, for example, risks imposed on children or climate change risks. As it happens, I was initially prompted to develop my autonomy-based account of the moral significance of risk imposition by the case of children—Feinberg’s paper “The Child’s Right to an Open Future” was a touchstone for me. That children are not autonomous agents does not, I think, show that my account cannot cover them, and Feinberg shows why: children have futures, and they have a claim on us that we leave them “open.” I take my idea of options to track what Feinberg had in mind in referring to open futures. But I also do not want to over-claim. For example, there are no doubt many reasons to object to the risks that we are imposing on future generations by triggering and failing to mitigate climate change. I do not pretend to account for every problematic feature of that risk imposition. It does seem to me, however, that a central objection to the way we are affecting the climate is that it will deprive future generations of worthwhile options—through our risky actions, we are narrowing the otherwise open futures of the generations that will follow our own.

   Regarding Madeleine’s third concern, as a threshold matter, I do not believe that risk impositions are uniquely morally objectionable by virtue of diminishing autonomy—other conduct might be objectionable for the same reason. So, it may well be that efforts to mitigate risk could also have this effect and by virtue of that be objectionable. But I don’t think I have to declare a moral stalemate if it turns out that a given risk and the efforts to mitigate it diminish autonomy. If some proposal to mitigate risk would curtail the autonomy of individuals more than the risk being mitigated, it seems to me that there is good reason not to mitigate the risk, and on grounds of autonomy. For this reason, it’s hard for me to see how the curfew Madeleine imagines would be justifiable: if a remedy is more harmful than the illness it remedies, one should not pursue the remedy. More generally, that autonomy is the basis for a significant interpersonal objection to risk imposition does not mean that it isn’t the basis for objections to other forms of conduct.

   But it’s clear to me that I need to grapple more with all of these issues, and I really am grateful to Madeleine for bringing them to the fore, and for engaging with my work so thoroughly. I look forward to this discussion.

2. Hello everyone. I greatly enjoyed the book and the review. I found some of the worries raised by the review compelling, so I’m looking forward to seeing what people have to say about them.

   I am interested in the last argument in the book (147-153). The argument goes like this: there is a worry that the contractualist framework for risk analysis is too constricting. If each risk must be justified to each person, then most or all risks will be ruled out because they will not be justifiable to some people. The paradigm example is justifying highways to the Amish. They get nothing out of highways except cars that crash into them, so it’s got to be curtains for the highways. That is not the result we want, so there are three arguments which aim to show that the contracutalist framework doesn’t deliver this result.

   The first argument is that “contractualist agreement is hypothetical, not actual” and so we can ignore “recalcitrant and unreasonable hold-outs” (147). But I do not see why the Amish are recalcitrant or unreasonable holdouts. They do not wish to be killed by fast cars. That does not seem unreasonable.

   The second argument is that we are not supposed to be evaluating “specific instances of risk imposition” (147). Instead we evaluate “broader justifiable principle[s]” (148). Thus the Amish might object to highways but they can’t object to “all risky instrumentalities and activities having a roughly similar profile” to highways (149). I don’t understand how this works. Say that I endorse 5 risk impositions, all with a roughly similar profile, because I prefer life with these 5 risks to life without these 5 risks. Together they give me a 5% chance of dying on any given day. If someone comes along and proposes to impose 94 more risks, all with a roughly similar profile to the existing 5, and which will collectively increase my chance of dying on any given day to 99%, I do not see why I must accept these new impositions on pain of inconsitency. Why can’t I be happy with some level of overall aggregate risk, and why can’t I reject impositions that raise my life above that level, even if these impositions have a roughly similar profile to the risks I accept? So, why can’t the Amish accept the risk impositions that are inherent to Amish society (barn raising, unpasteurized milk consumption, etc.), but reject further impositions which raise the total level of risk beyond what they wish to accept? There has got to be some number of additional risks the Amish don’t have to take on, even if they have a roughly similar profile, because eventually a large number of risks will basically spell certain doom. But once we admit that we can rule out risks even if they have a roughly similar profile to risks we endorse, why can’t the Amish use this excuse to rule out highways?

   The third argument is that we all have reasons to want the option of living lives that are possible only if risk impositions are possible (149-53). Specifically, we value the autonomy that risk imposition makes possible (149-50). So nobody can demand that everyone live a life free of risk (149). But the Amish are not asking that we live a life free of risk and the autonomy afforded by risk. They are asking that we live a life free of highways and the autonomy afforded by highways. If the contractualist procedure is to have any sort of bite at all, it has to give individuals some sort of voice. Short of stipulating by fiat that the Amish have it wrong and the highway lovers have it right, I don’t see what would rule out the Amish demands which would not rule out perfectly reasonable demands. Perhaps my confusion here is due to my unclarity about what count as reasonable demands. Examples elsewhere in the book suggest that driving drunk, or in a risky fashion, are ruled out. But I am not sure what principle any given individual would rely on to rule out these risk impositions which the Amish could not avail in order to rule out highway risk impositions.

3. Hi all,

   We’re having technical difficulties posting John’s reply to the review. So sorry about the delay! Hopefully it’ll be addressed soon.

4. What a delightfully long and interesting reply from John!
   There is a lot here to comment on, but I wish to first say something about the first point, on the concept of risk that I did not have much space to develop in the review. I have been very influenced by papers such as Hansson’s piece “Philosophical Perspectives on Risk” (Hansson 2004) on five different understandings of risk in the literature (as an unwanted event that may or may not occur; as the cause of an unwanted even that may or may not occur; as the probability for an unwanted event that may or may not occur; as the statistical expectation value of unwanted events that may or may not occur; and as decisions taken under risk, i.e. with known probabilities). What Hansson’s paper highlights, I find, is the existing plurality of risk conceptions and the differences in terms of stress between them: on the possible outcome (of harm), on the likelihood of such an outcome, on the combined measure of likelihood and outcome, and so on. Now, it seems to me that when we talk about the moral aspects of “risk impositions” what we think of seems to be a class of actions of a certain nature. The notion of probability however is not pointing to a class of actions, but more towards a numeric description (of the likelihood for a certain outcome to occur given a particular type of action). My own thoughts are that perhaps for moral purposes the most fruitful concept of “risk” may be one that focuses on the property of adding a possibility for harm rather than probability. To me it seems to fit better with the focus on risk impositions and something being imposed (a possibility or potential contributory cause of harm under some epistemic description) rather than increases or decreases of likelihood of harm. This would also have the additional benefit of being able to discuss what is imposed and the best measure of likelihood, separately. I am well aware that my view is a minority view here but it would be interesting to hear John’s thoughts on why “risk” for moral purposes ought be identified with probability rather than say a possibility or a (contributory) cause of harm.

   The grand question of the book I find is the question about the moral problem with risk impositions. What is the problem of imposing a “pure risk” on someone? It cannot be that it “portends to harm”, it is argued, since this would only lend a secondary moral significance to the risk imposition itself. This seems to me to be absolutely right. This is where the elegant solution of autonomy diminishments enters the picture. Perhaps, what makes risk impositions potentially morally wrong comes from the fact that they diminish the number of choice-worthy options for the person subjected to the risk. This seems to seek to explain two things: a) that what a morally problematic risk imposition does in some sense is to diminish autonomy and b) what makes such a morally problematic risk imposition morally problematic is that it diminishes autonomy in this way. This, at least is my reading of it, that there is both a descriptive and a normative explanatory part of the idea, and, presumable that the latter depends on the former.

   What is unclear to me is what are to say about cases where the set of choice-worthy options in some respect is empty, when a person in a particular context only faces unchoice-worthy options. This is not clear to me.
   If we remove a safe bridge from a remote village then would obviously decrease the number of choice-worthy options for the affected villagers (but not necessarily add any risk). If, by contrast, we were to lead a villager across a very unsafe but also the only bridge from such a village this would certainly add risk, but it is unclear to what extent, if any, it would diminish autonomy given that such a risk may be instrumental to greater autonomy at the other side. This may perfectly track the moral conclusions about the case: when a risky means may lead to a chance for overall improvements in terms of choice-worthy options (as when an unsafe way is offered as an escape route from a helpless situation) this may on the whole indicate a morally permissible or even morally praiseworthy action if those prospects are great enough. But here it seems that there is something of the descriptive part of the risk and its moral properties that fall out of the picture. It seems that there is little room to describe the moral significance of the riskiness if there is no diminishment to the autonomy (given that there were no safe options available for the agent.) But perhaps this is to be understood against an ideal rather than actual baseline of choice-worthy options?

5. Madeleine is right to point out that there are more conceptions of risk than the probabilistic one that I explicitly entertain and adopt in the book and to flag that I do not do enough to justify the important first move I make in proceeding with that conception. I’m drawn to a probabilistic understanding of risk for two reasons. First, when I think about what risks are permissible to impose on others, it seems nearly beyond argument to me that doing something that is more likely to harm another is harder to justify than something that is less likely to do so, other things being equal. And it seems to me that the probabilistic conception of risk is what underwrites that intuition. There’s a second reason as well, and I think it works in conjunction with the first one. It has long been observed that experts and non-experts have widely divergent beliefs about risks. It is this fact, for example, that prompted Paul Slovic to develop his well-known psychological work on the perception of risk. I recall a chart I used to use in class that gave regulatory experts’ rankings of risks alongside those of laypeople, and at the top of the experts’ list was, I think, radon, whereas something like a plane crash was at the top of the layperson’s list. It seems to me that the reason that the experts’ list is more credible, and so should be what informs risk regulation, has to do with its statistical credentials, and more broadly is evidential credentials. If it’s just a fact that there is a higher probability that one will be killed by radon than a plane crash, that’s really important. So, I am led to the probabilistic conception of risk, then, because it seems to me that probabilities represent the truth of the matter about risk better than anything, and that fact seems morally relevant as well insofar as it is hard to shake the thought that conduct posing a higher probability of harm to another is harder to justify than conduct that poses a lower probability of harm.

   But what’s interesting, as I think more about the multiple conceptions that Madeleine cites, is that my account of the moral significance of risk impositions ultimately relies on a modal idea. The reason why pure risk imposition calls for justification, in my view, is that imposing risk can foreclose an option that was otherwise available. And otherwise-available options connote possibilities. It would seem, then, that I end up not subscribing exclusively to a probabilistic notion of risk after all.

   Switching gears now, the provision of risky options is a great puzzle. If it’s important to connect two remote areas but the only feasible way to do so is by a rickety bridge, then I can see why, though the means are risky, the bridge could nevertheless be justified. It would not follow, though, that the risk posed by the bridge couldn’t still be explained in terms of the options it forecloses, and for just the reason that Madeleine suggests: one could compare the actual bridge to a safer one that would be safer precisely because it would be impervious to a wider variety of stressful conditions than the actual bridge. So, one could use the safer bridge in windy conditions, heavier users could cross it, etc. The fact that the rickety bridge enhances autonomy overall, in other words, does not show that it is still risky in virtue of foreclosing certain more specified options. But again, it is a fascinating problem and I would love to hear what others have to say about it.

6. I share other commentators’ view that John’s book is an impressive and important contribution to the nonconsequentialist literature about the justifiability of risk-impositions. And I agree with John’s explanation of why a probabilistic conception of risk is appropriate in analyzing this question. But I would like to raise an additional question about John’s provocative view that the imposition of a pure risk (a risk of which the patient or victim is unaware) is wrongful, not because of its potential to cause material harm, but because it diminishes the potential victim’s autonomy.

   In John’s view, pure risks can diminish autonomy by foreclosing options that the victim would otherwise have. Consider two of his examples.

   1. Imposing risk is like laying a trap: “the trap takes away the option, or more accurately renders unacceptable the exercise of the option, of stepping where the trap has been set.” (P. 86.)

   2. If speeding driver A is weaving in and out of traffic, a nearby driver B’s freedom of safe movement has been significantly restricted, even if B is unaware of A’s conduct, since a small turn of the wheel by B *could* have resulted in catastrophe.

   I am not entirely persuaded by these examples. In the trap example, the force of the intuition that the trapper acts wrongfully may simply reflect the high level of culpability of the “trapper.” One who lays a trap intends to limit the freedom of another, precluding their movement by ensnaring them within the trap. But risky conduct can be impermissible even if it is not *intended* to limit the autonomy of others. A speeding driver might simply intend to arrive at her destination more quickly or to enjoy the feeling of wind flowing through her hair; it does not follow that the risk she imposes is therefore permissible. John’s real concern is, or should be, the effect of risky conduct on others’ options, not the intention of the actor.

   The speeding driver example is more persuasive: after the fact, driver B would be justifiably resentful that if B had only slightly changed his driving behavior, driver A would have hit him and possibly killed him. But this raises a more fundamental question. Is pure risk wrongful only because of the range of options that it forecloses? I am doubtful.

   Suppose I am sitting in a chair and you fire a gun at me, and suppose the bullet will kill me only if I sit perfectly still. You have foreclosed only one of my options—remaining still in the chair. Suppose instead you set up a contraption such that if I were to move the slightest bit, the contraption will kill me. In this case, you have foreclosed every single one of my options except the one I am presently exercising, of sitting still in the chair. According to John’s analysis, the risk posed by the contraption has a much more serious effect on my autonomy (in the sense of restricting my range of options), and is thus much more morally troublesome, than the risk posed by the aimed gun. And yet, if I really do intend to sit still in my chair for a while, and you really do intend to kill me before I move, this conclusion is implausible. The firing of the gun seems to interfere with my rights in a more serious way than does the contraption, precisely because the risk of serious harm that it poses is much greater. In short, although the autonomy-limiting effects of a risk imposition do seem relevant to its wrongfulness, the probability and seriousness of the harm posed by the risk also seem highly relevant.

   I have raised this concern and some others in a review of John’s book aimed at law professors here: https://torts.jotwell.com/the-morality-of-risking/

7. Thanks very much, Ken, for your probing comments (not to mention for your JOTWELL review). I’m not sure that we share the same intuitions about my and your cases.

   Let me begin with the trap thought experiment, but let me add a wrinkle to it to bring out what I take to be most important about it: imagine that the traps weren’t “laid” at all, but were a natural phenomenon. If that were true, it seems to me that we could still talk about the way in which the risk impinged on autonomy. One’s autonomy would still be limited by the traps even if one were able make one’s way through the traps without incident. Now, the effect on autonomy wouldn’t be morally significant in my amended hypothetical as the impingement wouldn’t have been due to anyone’s agency, but it would impinge on autonomy all the same. If that’s so, then that impingement becomes morally significant–in need of moral justification–if it is done either carelessly or intentionally.

   Ken is surely right that intentional risk impositions are morally freighted in ways that non-intentional risk impositions are not. The example I use in the book of the car weaving through traffic replaced a different example from the underlying article, which was of someone shooting a gun at another. I changed it precisely because I realized, upon reading Adriana Placani’s terrific Law and Philosophy piece, “When the Risk of Harm Harms,” that the intentionality of that risk imposition obscured rather than illuminated what I was after. If a risk imposition is intentional then it is clearly culpable–in addition to potentially diminishing the autonomy of the one subjected to risk. But I’m not primarily interested in culpability, I’m interested in permissibility.

   Regarding his own example, where I’m shot at and can survive only by moving a bit or I’m strapped in a chair that will kill me if I move at all, Ken says that according to my analysis “the risk posed by the contraption has a much more serious effect on my autonomy (in the sense of restricting my range of options), and is thus much more morally troublesome, than the risk posed by the aimed gun.” Ken is certainly right that I believe that being prevented on pain of death from ever moving seems like a far worse curtailment of autonomy in comparison with being forced to move for a moment. As it seems like a wash in terms of culpability–both autonomy impingements are intentional–then it does actually seem to me that the chair situation is indeed “morally more troublesome.”

8. Hi, I apologize that I haven’t read the book, but I’m finding this conversation fascinating. I just wanted to make a small suggestion on one point that has come up, about the provision of risky options and about regulations to restrict risky options. I was wondering if it would be compatible with Prof. Oberdiek’s framework to appeal to indirect consequentialist considerations for cases like these. Maybe putting up a risky bridge where there was none before gives me more options, but a general practice of putting up risky infrastructure may nonetheless diminish my autonomy.

   Why? The thought would be something like this. Acquiring information about how risky various options are is costly, it consumes time and other resources. So if I have to constantly make judgments for myself about how risky certain situations are, I may find myself able to pursue considerably fewer options than I would like, since my time and other resources have been chewed up on doing research about, say, which bridges are safe, or whether I can trust my toothpaste not to poison me. So it may be that even if putting up any individual risky bridge does give me a new option, on the whole my autonomy is enhanced by a general ban on risky bridges. I can simply trust that options created by others will be reasonably safe, and go about making choices without prior investigation. Similarly for curfews. Maybe the curfew reduces my autonomy, but the fact that I can count on the gov’t to issue a curfew whenever the air is unsafe saves me time and energy that allows me to pursue more of my valued options overall, rather than spending time trying to determine air quality for myself.

   Anyway, that’s just a thought. Like I said, this conversation has been really interesting.

9. Thank you, Derek, and certainly there’s no need to apologize for not reading the book!

   The kinds of considerations you adduce about the burden of gathering information and so forth are ones that I take seriously. But I try to incorporate those considerations at a different stage in my argument. I think the burden of information gathering needs to be factored in at the (earlier) stage of initially characterizing the risk–that is, before we are in a position to assess either the moral significance of risk or the permissibility of imposing risk. Specifically, those are some of the main considerations that lead me to accept an evidence-relative conception of risk, and to reject a fact-relative one, as I think the fact-relative conception asks too much of agents. (A chunk of chapter 2 is devoted to criticizing fact-relativity, and specifically Judith Jarvis Thomson’s arguments on its behalf.)

   With respect to Madeleine’s puzzle about the provision of a risk option, though, I’m not as attracted to a solution along the lines you offer. I’m not certain that I fully understand the solution, though: are you suggesting that the provision of a single risky option does not call for moral justification, and so is not morally significant, if in general the options provided meet whatever standard of justification we have? At this point in the argument, I’m simply trying to figure out why we care about risk as a moral matter, I’m not yet trying to assess when imposing risk is permissible.

   Certainly a practice of providing only risky options does not enhance my autonomy the way a practice of providing safer options would, but I don’t see how that fact bears on the moral significance of the provision of a single risky option. It seems to me that we should still want to be able to assess the moral significance of such an option independently. Put it this way: even if there were countless sturdy bridges linking the two areas, it seems to me that providing a seemingly superfluous rickety bridge would still crimp one’s autonomy to the extent one used it. Maybe one is too heavy for it or something. It is probably important to underscore that the risk at issue is a pure risk, so that one is unaware of the risk–perhaps the bridge looks like all of the others but is built with substandard concrete–which lends credibility to the possibility that people would still use this bridge. So, just as having lots of safe bridges in one place doesn’t diminish the autonomy-diminishing characteristics of the risky bridge–its provision remains morally significant–it doesn’t seem to me that a general practice of providing safe bridges will affect the moral significance of providing a risky bridge. I hope that makes sense.

10. Hi John,

    The solution was just meant to appeal standard indirect consequentialist considerations. Sure, giving you a risky option gives you an additional option, but a rule that says “Don’t provide options over a certain threshold of risk,” generally adopted by society, gives people even more options–because they don’t have to waste their time figuring out how safe every individual option is. Providing an individual risky bridge would probably call for moral justification, then, because you would need to offer some reason for violating an autonomy-enhancing rule.

11. Thanks, John, for your thoughtful response to my comment. I still believe that the “trap” metaphor is potentially misleading in its connotation that the actor must harbor an intention to trap or confine the other, but we agree that your “foreclosure of options” approach can apply when the “trap” is merely carelessly set.

    With respect to the chair examples, I should have been clearer. My point was not that there is never a moral difference between cases in which many options are foreclosed and those in which few options are foreclosed, but only that this difference is often much less significant than a difference in the probability of harm. In the chair examples, there was indeed a large difference in probability of harm that, in my view, is more significant than the difference in foreclosure of options.

    Here is another set of examples that might bring out the point. (And they involve trolleys!)

    Suppose D1 is driving a truck without paying sufficient attention to the icy conditions on the road. He loses control of the truck, which is headed directly towards a stationary trolley. V1 is the driver of the trolley, taking a brief nap before the trolley is to begin its daily route. If V1 were to awaken and see D1’s truck approaching, her only option for avoiding serious injury or death would be to turn the trolley onto a “single* available spur. Fortunately, the truck does not crash into the trolley but just misses.

    Compare the following otherwise identical example of truck driver D2 and trolley driver V2. If V2 were to awaken and see D2’s truck approaching, V2 has a choice of *five* spurs onto which she could turn the trolley. Choosing any of these five options would save V2 from injury or death. Again, the truck does not crash into the trolley.

    Assume that in both cases, the truck driver poses the same substantial chance of causing the harm, and the trolley driver, if awake, would have the same chance to avoid the harm. Is there really a moral difference between the unjustified risk that D1 imposed and that D2 imposed? The availability of additional options in D2/V2 seems to make little or no genuine difference. Furthermore, imagine that the ex ante risk imposed by D1 is significantly greater (say, a 50% risk of death) than the risk imposed by D2 (say, 10%). Shouldn’t we then conclude that D1’s risk imposition is morally more problematic than D2’s?