1. Conceptos fundamentales
1.1 Teoría de la consolidación unidimensional de Terzaghi (1923)
In this section and the next, I will criticize Levy’s rejoinders to Robichaud’s two- pronged attack. Levy’s view is overly restrictive and unable to sustain or justify our social practices of holding each other accountable for many of those acts and omissions for which we do in fact hold people responsible.84
Recall that Robichaud’s first prong involves accepting Levy’s internalist conception of rationality, while arguing that internalism does not entail epistemically deficient agents are akratic. We can hold agents accountable for failing to avail themselves of opportunities for the improvement of their beliefs even if they do not believe all-things-considered they ought to avail themselves of such opportunities. This will occur when an agent has sufficient reasons to take advantage of an opportunity to meet her epistemic obligations, yet she has sufficient reasons of equal force that underwrite not taking advantage of that opportunity. Limiting ourselves to two choices, the agent will either take advantage of the opportunity or not. Either way, she will do something rationally (i.e., supported by reasons). Yet, if the agent fails to take advantage of the opportunity we can still hold her accountable because, given her sufficient internalist reasons, she had the capacity to take advantage of the situation for improving her epistemic position but failed to do so. When this results in
84 I thank Hanser for requesting clarification on the nature of my criticism. Should the objection be that any view that fails to vindicate our ordinary practices is defective, which is something Levy would just deny? Or, is there something more? In reply, the objection is that responsibility as accountability, which is the conception of responsibility at issue, is a social concept that accords with our practices of holding people accountable for their wrongdoings. Any view that does not capture such ordinary practices is revisionary of the concept of responsibility at issue. There are theorists that endorse this approach to moral responsibility, such as Manuel Vargas (2012). But, that is not what Levy is up to in embracing an internalist version of responsibility as accountability.
ignorance—moral or circumstantial—the agent is culpable for that ignorance even though the ignorance itself is not the result of an akratic act.
Levy (2016) begins his response to Robichaud’s first prong by reiterating the meaning of ‘capacity’ that he is working with. Recall Levy’s chain of nested requirements: a requirement for it being reasonable to expect an agent to remedy her ignorance is that she has the capacity to correct her ignorance, an agent has a capacity to correct her ignorance only if a reasoning procedure is available to the agent that would enable her to remedy her ignorance, and this reasoning procedure is available to the agent only if she possesses internalist reasons that rationalize using such a procedure. For Levy, this ultimately means that, “it is reasonable to expect someone to do something only if they have the skills, and the opportunity to intentionally do it (or do something that entails doing it) via a reasoning procedure; where to do something via a reasoning procedure requires that the agent does not do it by chance or through a glitch in their agency” (Levy 2016). Levy argues that the case Robichaud has in mind of possessing sufficient but nondecisive reasons results in the agent improving her epistemic position by way of chance or glitchy agency. When, for Levy, is an event chancy?
Inspired by the work on luck by Duncan Pritchard (2005) and E. J. Coffman (2007), Levy (2011) develops a modal understanding of when an event is chancy:
Chance: Event E is chancy if it occurs in the actual world at t1, but it fails to
more than a small change to the actual world at t0; and the agent lacks direct
control over E’s occurrence.(Levy 2011: 19)85
In figuring out whether an event E in the actual world is the result of chance or a glitch in one’s agency we construct nearby possible worlds that are reachable by only making a small tweak in the conditions just prior to E happening. We press play on the causal sequence of those nearby possible worlds and see if an event other than E occurs in those worlds. If a different event occurs in a large enough proportion of those possible worlds, then event E in the actual world is a chancy event. What about the qualifier “large enough”? What constitutes a failure of E to happen in a large enough number of nearby possible worlds?
Levy’s view of chance is an improvement over Pritchard and Coffman’s view. Pritchard and Coffman think luck requires both significance and chance. The notion of luck is agent-relative. It is not purely a matter of statistics and the frequency of occurrence of events. An event must have significance to an agent for that event to be lucky relative to the agent. If I do not care whether a specific leaf falls into my backyard, then, when it falls into my backyard, it would be odd to say that it is lucky that the leaf fell into my backyard. Even if the event is chancy in that in most worlds reachable by making a minor tweak in the physical conditions (i.e., leaf position, wind direction and speed, and so on) the leaf does not fall into my backyard, the lack of significance prevents the event from being lucky when it occurs in the actual world. Levy improves on Pritchard and Coffman’s view of luck by integrating the significance
85 Quoted with the time dimension (t) signified with subscript numbers instead of symbols. I will largely bracket the issue of control in the Chance principle. The thought behind the control qualification in Chance is that an event is not chancy if the subject possesses direct voluntarily control over the event. Events subject to chance are typically outside this form of agential control.
requirement into the chance requirement. Levy makes “large enough” co-vary with the significance of the event for the agent. To see how this works, consider this case involving Russian roulette:
Russian Roulette: Samuel plays Russian Roulette with a revolver that holds six
bullets. He loads a single bullet in the revolver, spins the revolver’s bullet chamber, points the gun at his head, pulls the trigger, and lives.86
For Levy, Samuel living through playing Russian roulette is a chancy event. Even though Samuel survives the game in a high proportion (5/6) of possible worlds, this ratio is not high enough given what is at stake, namely Samuel’s life. Coffman considers an event chancy so long as it fails to occur in at least half of the nearby possible worlds reachable by making a small tweak in actual conditions. For Coffman, Samuel is not lucky to survive playing Russian roulette, as the event is not chancy. Instead, Samuel is fortunate. Against Coffman’s position, Levy argues:
Though I think that Coffman may be right in distinguishing between luck and fortune by reference to the chanciness condition, I doubt we can establish even a rough threshold, below which an event is not chancy enough to count as lucky, which will hold for all possible cases. It is more plausible to hold that the degree of chanciness necessary for an event to count as lucky is sensitive to the significance of that event for the agent. Surely if Samuel is lucky to survive playing a round of Russian Roulette (as Coffman must concede) he is also lucky to survive a single round with a gun which has a 49.5 percent probability of firing….On the other hand, there is a threshold, even for an event as significant
for the agent as the death Samuel risks, below which surviving is merely fortunate and not lucky (suppose that the probability of the gun’s firing was 0.00001%); conversely, for relatively trivial events, the probability of the event’s occurrence might have to be well below 50 per cent for it to count as lucky (suppose that the gun would give Samuel a mild and brief headache; in that case the probability of its failing to fire might have to be quite remote for this failure to count as lucky for the agent). (Levy 2011: 17)
Levy’s view of luck improves on Coffman’s view because it does not set a hard threshold above which an event counts as chancy. Even if, as in Levy’s case above, Samuel plays Russian roulette and the gun is slightly more likely than not to fail to fire, Samuel living through playing the game is still lucky. Luck is involved even though he survives in (slightly) more of the nearby possible worlds than he dies (and hence is not chancy for Coffman). Alternatively, for an event of low significance for the agent to count as lucky the event may need to have a low chance of occurring (i.e., a probability well south of 50%). We can imagine a game of Russian roulette played with a toy revolver gun and bullets that if fired at one’s head at close range gives one a minor headache. Imagine that 5 toy bullets are loaded in the gun. We can look at this case and a case of standard Russian roulette to illustrate the relation between significance, chance and luck for Levy. Consider the cases in this table:
Gun Fires Gun Fails to
Fire Levy’s Verdict
Russian Roulette: 1 Real Bullet
Death Odds:
1/6 (16.7%) Survival Odds: 5/6 (83.3%) Lucky to survive, yet odds favor this outcome.
Russian Roulette: 5 Toy Bullets
Minor Headache
Odds: 5/6 (83.3%) No Headache Odds: 1/6 (16.7%) Lucky no headache, yet odds against this outcome.
The odds in the chart are not hard and fast thresholds, but they illustrate how Levy thinks of luck. An event that is hugely significant for the agent, such as death, must have vanishingly small odds of occurring for the failure of its occurrence (e.g., surviving) to count as fortunate, not lucky (as when there’s only a 0.00001% chance of the gun firing). When a very significant event has non-insubstantial odds of occurring, such as the 16.7% chance of death in the standard Russian roulette case, then the failure of its occurrence—which is largely more probable than not—can still qualify as a lucky event for the agent. Conversely, when an insignificant event has substantial odds of occurring, such as the 83.3% chance of a minor headache in the gun with five toy bullets case, then the failure of its occurrence—which is largely less probable than not—can qualify as a lucky event for the agent.
With this background in mind, let’s consider Levy’s rejoinder to Robichaud’s first prong of attack on Levy’s internalism. Recall, we’re considering a case in which an agent has sufficient, non-decisive reasons that equally favor one of two options: either taking steps to improve one’s epistemic position or not taking such steps. Robichaud claims that an agent can “rest on her epistemic laurels,” be blameworthy for doing so, yet fail to act akratically. Levy replies that Robichaud’s analysis only works if it is reasonable to expect the agent to improve her epistemic position. But in the case imagined by Robichaud the agent does not have the capacity to comply with such demands. Why not? The agent complying with such demands is an event that is chancy, and we cannot reasonably expect agents to do things the occurrence of which is a matter of luck.
That an agent with sufficient, nondecisive reasons for complying with her procedural epistemic obligations complies with those obligations is a chancy event in the actual world. This is because the agent’s reasons are equally sufficient for rational compliance and non-compliance. Yet, this assumes the factors are insignificant. If the agent decides one way instead of the other, and significant or stable factors determine this choice, the event will not be chancy on this view.87 Insignificant factors, such as
“situational and internal primes,” can generate cross-world instability for that event. The event can fail to occur in a large proportion of possible worlds reachable by making small tweaks in the state of the actual world just prior to the event occurring, which would make the event in the actual world chancy. Levy explains how such cross- world instability between the event in the actual world and events in nearby possible worlds can occur:
Perhaps the decision will be caused by a stochastic brain mechanism that cuts short deliberation between options; in that case, the agent will just plump for one option or the other. Perhaps it will be situational influences, of the kind that social psychologists have studied, that will be decisive. Or perhaps it will be the force with which a consideration strikes her, or the order in which they occur to her. All of these are, or are the product of, trivial aspects of the environment or of the agent herself. It follows that in a large proportion of nearby possible worlds, the agent will choose differently; thus her actual choice is chancy. (Levy 2016)88
87 Thanks to Greco for bringing to my attention the need for this qualification.
88 I adopt Levy’s use of the phrase “actual choice,” but this phrase is ambiguous between choice in the actual world versus how the agent in fact choses in the actual world. Chanciness is not determined by how the agent in fact choses, it is determined (in this context) by whether a choice in the actual world
Situational and internal primes can cause a person to choose one way as opposed to another when her reasons between the choices are equally sufficient and forceful in supporting the respective choices. Levy takes this point a step further:
More generally, when it is genuinely the case that an agent has sufficient but not decisive reasons to choose between two or more conflicting options, chancy factors will play a decisive role in how she chooses. It is only when an agent has decisive reasons to choose a particular option that her choice will be resistant to the influence of chance events89….In such cases, how the agent decides is
subject to chance and it is not reasonable to expect (in the relevant sense of ‘expect’) that someone be subject to chance. (Levy 2016)
There are several ways of calling into question Levy’s response to the first prong of Robichaud’s objection to Levy’s reply to the externalist about culpable ignorance. First, it is unclear that “trivial aspects of the environment or of the agent herself” will prove decisive in choosing between options that are each supported by reasons the agent takes strong enough to support two distinct courses of action. It matters why the agent takes herself to have strong reasons supporting both choices. The internal reasons that rationalize the distinct actions may involve deep aspects of the person’s psychology and values. To see how this works, let’s borrow an example that Robichaud uses to illustrate when a subject possesses sufficient but nondecisive reasons. This case involves someone deciding to quit smoking or keep smoking:
to meet her epistemic duty would fail to obtain in a significant number of nearby possible worlds. It is important to keep in mind that “actual choice” is referring to whether a choice in the actual world subject to modal evaluation is chancy, which, again, may not be how the agent actually chooses. 89 This is because, “the stronger her reasons for a particular choice, the larger the proportion of possible worlds that differ only trivially from the actual world in which she will choose accordingly, such that if she makes that choice in the actual world, her choice is not chancy” (Levy 2016).
an agent has sufficient, nondecisive motivating reasons to quit smoking when she takes herself to have reasons that are strong enough to make quitting a rational option, but not strong enough as to decisively support quitting. Such an agent might simultaneously take herself to have reasons to continue smoking that she (wrongly, let’s say) takes to be just as strong as her reasons to quit. These other sufficient though nondecisive reasons would make it rational for this agent to continue with her smoking habit. (2014: 142-3)
We can fill in the case a bit. This is one way a person might take herself to have internal reasons that simultaneously rationalize quitting smoking and keeping smoking. An agent might have motivating reasons she takes as strong enough to make quitting smoking rational because smoking is a somewhat nasty habit and she values beauty and good aesthetics. For instance, she values pleasant smells and foods that taste good. Smoking makes her not smell good and it dulls her taste buds. Yet, an equally deep value for the agent might be independence and autonomy. The agent might value not following the crowd and smoking is largely frowned upon in her culture. For her smoking is a way to not conform to societal norms. The values of aesthetics and personal autonomy might be both deeply held and equally strong.90 They will not be
overridden (i.e., the tie will not be broken) between the sufficiently strong reasons favoring acts that reflect those values by random brain glitches, the order of presentation of considerations, or other situational influences. Such chancy factors likely will not make a difference in deciding between two options, given that the
90 Notice I am not claiming that the aesthetic value and the personal autonomy value are objectively strong values. The context of discussion is motivating reasons, which are reasons the agent takes to be sufficiently strong. Objectively speaking the reasons may not be sufficiently strong but that is not what is at issue here.
reasons are sufficiently strong for the agent because tied to things she deeply values. It is reasonable to think chancy factors will not necessarily break the tie and this counters Levy’s claim otherwise when he states, “when it is genuinely the case that an agent has sufficient but not decisive reasons to choose between two or more conflicting options, chancy factors will play a decisive role in how she chooses” (2016).91
However, if situational and internal primes are truly insignificant features of the agent, then they will not necessarily make one option decisive over another equally weighty and sufficiently supported option, especially when the two options are tied to deeply held values.
Secondly, for the sake of argument, we can allow that the situational and agential primes generate choices in nearby possible worlds that differ from the choice in the actual world but argue that such primes do not entail the claim that the agent’s choices will differ in a “large proportion of nearby possible worlds.” Again, one could support this contention by way of claims regarding how some of the primes will not secure different choices in enough nearby possible worlds. The nearby possible worlds are reachable by making minor changes to the initial conditions right before the