Over the past few years, an interesting development in experimental philosophy has been work on the “ought implies can” principle (OIC) in commonsense morality. Several research teams have investigated whether patterns in commonsense moral judgment are consistent with a commitment to OIC, understood as a conceptual entailment from having a moral responsibility to being able to fulfill it. Across a variety of contexts and testing procedures, the principal finding has been very consistent: people are definitely willing to attribute moral responsibilities to agents unable to fulfill them. Based on these findings, I and others have concluded that there is no conceptual entailment from “ought” to “can.” But there is a lingering question. If there is no conceptual entailment, then what is the source of the intuitive link, which many theorists seem to sense, between “ought” and “can”? A new paper might provide at least part of the answer.

One theory, proposed by Walter Sinnott-Armstrong, is that (moral) “ought” claims can function as both descriptions of responsibilities and as encouragement toward their fulfillment. Their descriptive function is perfectly consistent with inability, but their encouragement function is not. When I describe you as having a responsibility to fulfill your promise, my remark is perfectly consistent with your being unable to do so. In this respect, “ought” exceeds “can.” But when I encourage you to fulfill your promise, the natural interpretation of my remark suggests that I believe you can do so. In this respect, “ought” implies “can.” (Charles Pigden has traced some of this theory’s historical antecedents.)

So the theory really has two parts. One part, the dual-function hypothesis, states that “ought” claims have at least two functions: description and encouragement. Another part, the interaction hypothesis, states that the two functions relate differently to an agent’s inability to perform.

To test the theory, I ran a series of behavioral experiments. It turns out that “ought” statements can function to both describe responsibilities and encourage their fulfillment. (They can also be used to cast blame in some contexts.) This supports the dual-function hypothesis. Moreover, the two functions relate differently to whether an agent is able or unable to perform. Regardless of whether an agent can perform, people consistently interpret “ought” statements as describing responsibilities. And while people consistently interpret “ought” statements as encouragement when the agent is able to perform, they do not do so when the agent is unable to perform. This supports the interaction hypothesis.

But it’s not just that inability depresses an “ought” claim’s ability to function as encouragement. It seems to do so for a specific reason, namely, that it is not worthwhile to encourage somebody to do something that they cannot do. This is best illustrated by the following study.

Participants were assigned to one of two conditions, in which they read a story about Alex (my adorable nephew’s namesake). Alex promised to deliver a document by 4pm. In the email condition, the document is digitized and has to be emailed to the clerk’s office. Given the strong and reliable internet connection, Alex can email the document all the way up to 4pm and still deliver it on time. In the drive condition, the document is hand-signed and has to be physically delivered by driving it to the clerk’s office. Given the driving distance, if Alex leaves any later than 3pm, it is impossible for him to deliver the document on time.

So the two conditions are matched in terms of Alex’s promise to deliver the document by 4pm. But they differ in the time at which Alex becomes incapable of fulfilling the promise: 4pm in the email condition compared to 3pm in the drive condition.

Participants then performed a timeline graphing task. They were presented with a series of statements and instructed to construct timelines depicting the period during which the statements are true in the scenario. The three key statements were these:

  1. Alex has a responsibility to deliver/email the document by 4pm. (description)
  2. It is still worthwhile to encourage Alex to deliver/email the document by 4pm. (encouragement)
  3. Alex can still deliver/email the document by 4pm. (ability)

As you can see in the figure below, in the email condition, participants depicted all three statements as true until basically the same time: 4pm. By contrast, in the drive condition, participants depicted the responsibility as lasting until 4pm, but ability expired much earlier, at basically the same time that it was no longer worthwhile to encourage Alex to deliver — literally and figuratively! — on his promise.

figure - timeline graphs
Mean estimates on a timeline graph task. Participants constructed bar graphs representing the time during which each statement was true in the scenario. Error bars +/- SEM.

Overall, then, these findings support a specific diagnosis of the intuitive appeal of OIC: it is intuitive, at least in part, because “ought” claims can function to encourage behavior, but it isn’t worthwhile to encourage people to do things they can’t do.

The findings also address some concerns that others have expressed about prior findings undermining the status of OIC in commonsense morality. The concerns pertained to alleged oddness or task demand brought on by certain questioning procedures, which allegedly complicates the interpretation of very clear and consistent patterns whereby the overwhelming majority of people attribute moral responsibility in tandem with inability. For instance, participants might agree with responsibility attributions if at any point in the scenario the agent had a responsibility, while agreeing with the ability attribution if at any point in the scenario the agent was able to perform. If so, then there might be no point in the scenario where participants attribute both responsibility and inability. But this concern cannot apply to the results from the timeline-graphing study just described. Instead of providing verbal answers that could potentially be implicitly indexed to different times in the scenario, participants graphed timelines to visually represent the time during which a statement was still true. In other words, people drew simple pictures showing that, on their understanding, some responsibilities outlast the agent’s ability to fulfill them.

4 Replies to “How “ought” exceeds but implies “can”

  1. Thanks for the interesting post. Is there a reason why you say that ought statements have two functions, rather than simply that “ought” is ambiguous? My initial reaction on seeing one of your other papers on this was precisely that “ought” is ambiguous between a sense which is insensitive to ability and a “deliberative” sense which is sensitive to ability. Some distinction along these lines seems to appear in various authors’ work. On this view, ought-implies-can is true after all, but only of one of the concepts we can express using the word “ought”.

  2. Hi Alex,

    There are a few reasons for that. First, the theory I was testing is stated in terms of functions rather than senses. Second, and relatedly, the theory in question does not require any particular explanation for the multi-functionality of “ought” claims, so it didn’t seem advisable to saddle it with that additional assumption. And I’m unaware of any evidence that “ought” is ambiguous in the relevant way, although a potential explanation for its having multiple functions is that it has multiple senses.

    Beyond all of that, I don’t think that the sort of proposal you’re suggesting can salvage a sense in which OIC is true (understood as a conceptual entailment). Suppose that “ought” has multiple senses, and according to one of them “You ought to keep your promise” means “I encourage you to keep your promise.” (According to the results I obtained, ability constrains the encouragement function.) There is definitely no conceptual entailment from “I encourage you to keep your promise” to “You can keep your promise.” It’s too bad that there isn’t, though, because it would make all sorts of things possible!

  3. Thanks for that interesting data and observation. It is helpful to see such clear cases of responsibility and ability coming apart in a test case.

    I’m wndering about something though: Is there a possibility that the subjects may think of a responsibility that was acquired when Alex was able to fulfill it as not being extinguished by Alex’s failure to fulfill it within contingently constrained time intervals, but only by metaphysical impossibility? That everyone thinks the responsibility ends at 4 PM must have an explanation, and the best I can think of is that it is metaphysically impossible to deliver a document by 4 PM when it is now 4:10 PM. That Alex can’t do that isn’t her fault. But that at 3:15 Alex can’t deliver the document in the drive case isn’t because of some law about time, but because Alex has let contingent things like driving distance culpably make it impossible to deliver, but she’s not off the hook until the clock has struck 4 at which point she’s just laden with an unfulfilled past responsibility.

    Of course, that is just speculating from the Armchair, but I’d be curious if you could test the following:

    If Alex, not knowing the time, promises to deliver the file by 4, does that still create a responsibility if the promise is made at 3:15 PM in the drive case so that it is already contingently impossible to deliver on time?
    Compare that to Alex, not knowing the time, promising to deliver the file by 4 via email when it is already 4:15 PM.

    If Alex is judged to have a responsibility in neither case, then this points to some problem with the previous setup where Alex is judged to have the responsibility still at 3:15. But if the judgments here are that Alex has the responsibility in the first, but not the second case, then that undermines the explanation from left over responsibility (you once could, and it isn’t yet metaphysically impossible) I was postulating.

    I’m also curious if there is a way to reverse the temporal order, such that you promise to do something that is at first impossible and only later becomes possible, to see if that stays consistent with your earlier result.

    Maybe something like the following: Alex promises to have a document delivered as soon as it has been signed. The document is signed at 3PM. What time is such that Alex is responsible for the document being in the hands of the recipient then? (That sounds awful and makes for a terrible vignette, but it’s important not to just ask when the document has to be in transit). If there is a good way of asking this, I wonder if people still think Alex is responsible for the document arriving at 3 PM in the drive case, even though that’s impossible. But it’s just as impossible as the document arriving at 4 PM in the firs case if Alex hasn’t left yet at 4 PM. If in the promising to deliver once signed + drive case subjects think that Alex isn’t responsible for the document being in the hands of the recipient until 4 PM then this would point at some “can” driving the ought after all, and that “once could and still metaphysically could” explains the judgments in your case.

    I don’t know if testing any of this is worthwhile, but I would like to know what other than an OIC principle, even a one restricted to metaphysical can, explains why the responsibility cuts of at 4 PM in your case.

    – Kolja

  4. Hi Kolja,

    Why does everyone think that the responsibility expires at 4pm? Good question! You suggest that this is because people think it is metaphysically impossible to fulfill the responsibility by 4pm once 4pm has passed. I have not tested this, so I can’t rule it out, or in, with data. Based on my experience, though, I’d say that “metaphysical impossibility” is not a concept that people are pre-theoretically going to apply. It’s a technical concept. It’s hard to imagine a study providing evidence that people were implicitly reasoning from an assumption about metaphysical impossibility.

    An alternative explanation, requiring no technical concepts, is that people understand *the promise*, and thereby the responsibility, to expire at 4pm.

    A further question might then be why is that a feature of our institution of promise-making, and it would not be surprising if our institutions were broadly sensitive to general facts about human abilities.

Comments are closed.