Can Moral Beliefs be Supported by Observation?

Photo by Lucas Brandão from Pexels

The observation of cake in the fridge is good evidence that there is in fact cake in the fridge (absent defeaters). Observational evidence of this sort is paradigmatically good evidence for belief. Can there be observational evidence for moral beliefs? For example, are there any possible observations that could confirm a belief like it is wrong to eat animals raised in factory farms (P)? Here's an easy "yes" to that question: you could get observational evidence that factory farms subject animals to immense pain and suffering. The problem is that the answer can be explained in the following way: your observational evidence is justification for the empirical premise that factory farms subject animals to immense pain. That premise, in conjunction with the premise that it is prima facie wrong to eat animals that have been subjected to immense pain justify P. Hence, your evidence about animal suffering in factory farms only indirectly confirms; the support your empirical evidence gives to P is mediated through other non-moral premises. That isn't surprising, however. The really interesting question is whether you can get "direct" empirical support for a moral claim like P––support that is not mediated through non-moral premises or non-moral beliefs. 

Sarah McGrath thinks this kind of support is possible (see Moral Knowledge, ch. 4). Start with the Bayesian notion of evidential support: E supports H just in case Pr(H|E) > Pr(H). As others have noted, Bayesianism offers a "weak" theory of evidential support: very small boosts in the credibility of H count as evidential support. Can observational evidence provide this kind of "boosting" for moral beliefs? McGrath invites us to reflect on a case like this:

• The Story of Ted: Ted believes that inter-racial marriages are morally problematic. Because of this moral belief, he thinks that a proliferation of inter-racial marriages will have adverse consequences for inter-racial marriages and for society at large. Importantly, he does not take these adverse consequences as the grounds of the wrongness of inter-racial marriages. Ted is not a consequentialist. Instead, he thinks there will be adverse consequences because he believes that inter-racial marriages are morally problematic. The moral status of the union is part of what explains the ensuing adverse consequences; the adverse consequences do not explain the wrongness of the unions, on his view. He thinks that families will start to deteriorate, the divorce rate will increase, more diseased and feeble minded children will be born, individual wellbeing will diminish, and so on (these very arguments against inter-racial marriage were kicked around in the late 19th century and into the early 20th century! 😔). The number of inter-racial marriages begins to grow. However, contrary to his expectation, Ted fails to observe these expected outcomes. He observes the opposite: inter-racial marriages seem to improve society and are no more prone to end in divorce than other marriages.

Plausibly, it is reasonable for Ted to revise his beliefs about the moral status of inter-racial marriages (call his view about the morality of inter-racial marriages "M"). M is disconfirmed somewhat. That is, where O is Ted's observational evidence concerning inter-racial marriage: Pr(M|O) < Pr(M). But it follows from the Bayesian apparatus that O disconfirms M just in case O confirms ~M. Hence, Pr(~M|O) > Pr(~M). A moral claim, ~M, is supported by a piece of observational evidence.

The same conclusion seems to apply to a case like this:

• Ted's Surprising Discovery: Ted believes that inter-racial marriage is morally problematic (M). Moreover, he thinks highly of his good friend, Jan. He has reason to believe that Jan would never repeatedly engage in anything morally problematic (maybe he thinks Jan is a moral exemplar). He then makes multiple ongoing observations of Jan in an inter-racial marriage (O*). 

Again, it seems reasonable that Ted's confidence in M should diminish somewhat (perhaps substantially!). His surprising observation about Jan, O*, disconfirms M (and, hence, confirms the moral claim that inter-racial marriage is not morally problematic).

An obvious response points out that support is a three place relation, between evidence, hypothesis and background assumptions, B. O supports ~M only relative to Ted's background beliefs (like his beliefs about the adverse consequences of inter-racial marriage or about the moral character of Jan).  Hence, it is misleading to say that O supports ~M. McGrath argues convincingly that this doesn't pose a problem for her view. By parity of reasoning, the same argument undermines the view that empirical observations sometimes support rich scientific theories. We tend to think that rich scientific theories (like atomic theory) can be supported by observational data. But, just as with moral claims, this support is obtained relative to a set of background assumptions. If the presence of background assumptions is consistent with observational support for scientific theories, the presence of background assumptions is consistent with observational support in the case of moral claims. McGrath concludes:

Even a relatively fundamental moral conviction (of a kind whose truth or falsity seems maximally independent of any contingent matters of fact) can give rise to expectations about how the world will turn out to be when it is conjoined with other views. When those expectations are disappointed, it can make sense to treat that outcome as evidence against the views that gave rise to them. When those expectations are fulfilled, it can make sense to treat that outcome as evidence in favor of those views.

My suspicion: the relation of support between O and ~M (and B) can be explained by appeal to special items in Ted's background beliefs that aren't paralleled in cases of support between observation and scientific hypotheses. First, it is important to note the following: the Bayesian formalism models support; it does not explain how the support arises in the first place. It is true that Pr(there is milk in the fridge | I remember seeing milk in the fridge) > Pr(there is milk in the fridge). But this inequality modeling support does not tell us how the support relation obtains. Is the support grounded in explanatory relations between my memorial evidence and the relevant claim? Or is the support grounded in relations of immediate support? And so on. Hence, although Pr(~M|O) > Pr(~M) might be true, that fact alone does not fill in the details about how O supports M. Put differently, merely pointing out that Pr(~M|O) > Pr(~M) does not tell us anything about how O transmits support to M.

With that in mind, some kinds of support can arise for reasons that have more to do with "bridge principles" than with "pure" relations of support between evidence and hypothesis.  If Ted's beliefs (about inter-racial marriage or Jan) serve as "bridge principles" between O and ~M, then there is nothing surprising about O supporting ~M. If it's the case that "If O, then ~M" is in my background knowledge, then I gain support for ~M by learning that O. Pr(~M| O & [If O, then ~M]) = 1. Supposing Pr( ~M) ≠ 1, Pr(~M| O & [If O, then ~M]) > Pr(~M). But that isn't surprising. That's modus ponens doing the epistemic lifting, it would seem, and we already knew that it could do that. The real question is whether O can confirm ~M in the absence of any such connecting principles. Moreover, if observational support for scientific hypotheses is our comparison, then we can ask: do scientific hypotheses need these type of bridge principles (present in B) in order to receive support from observational evidence? That seems doubtful. On the other hand, if we imagine that there are no principles connecting O to ~M in Ted's background knowledge, do we really intuit that O supports ~M? It's hards to say. Unless the same type of logical relations between empirical observations and scientific hypotheses are needed to explain how empirical observations can support scientific hypotheses, then there might be a crucial difference between ordinary cases of observational support for scientific hypotheses and the kind of observational support for moral claims that McGrath identifies (sorry, this last sentence was a mouthful!).