Case 7: …a judge is sentencing a convict. The judge happens to know that the convict is an innocent man who has been unjustly framed. The law requires a ten year sentence, but morality requires that the judge help the innocent convict to escape. So the judge, moved by compassion for the convict, does just that.
FAM: x is merciful to y at t =df there are two alternative acts, A and B, available to x at t such that (1) x performs A, (ii) X's performing B would have resulted in more harm to y than x's performing A, (iii) x is motivated by compassion for y to choose A over B, and (iv) A and B are both permitted under the relevant notion of permissibility.
So FAM does indeed imply that the act in Case 7 is not merciful. For neither morality nor legality succeeds in making permissible both sentencing the prisoner and helping the prisoner to escape. Legality disallows helping the prisoner escape, while Morality disallows leaving the prisoner with a ten year sentence.
Markosian acknowledges that this is counterintuitive, but bites this bullet. His account after all seems to work well as a unified account of the majority of intuitions about mercy—and also suggests a very plausible explanation as to why those intuitions often differ.
I am very strongly inclined to think that the judge in Case 7 is merciful. Markosian offers an alternative account of mercy which would turn out a judgment that the act is merciful:
DFAM: x is merciful to y at t =df there are two alternative acts, A and B, available to x at t such that (1) x performs A, (ii) X's performing B would have resulted in more harm to y than x's performing A, (iii) x is motivated by compassion for y to choose A over B, (iv) two different notions of permissibility, are somehow picked out by the context, p1 and p2, (v) A is permitted under p1, and (vi) B is permitted under p2.
In my view it is difficult to see how a single context of utterance could render relevant two distinct notions of permissibility. Part of what constitutes a context of utterance is the speaker of the utterance and her audience. (Markosian often refers to this audience as though they were a group of lawyers, or a group of moral saints, or some other such unified group.) Indeed, to picture a single context of utterance with two distinct audiences is strange. Not that it's impossible, but rather, that calling this a "single" context doesn't seem useful—rather, it would seem a single utterance is being dealt with in two different contexts in such a case. And surely in each individual context of utterance, if the audience is in fact supposed to be able to come to a rational and unified decision as to the truth of the utterance, there must be just one notion of permissibility that applies. If the audience is divided, then (again, assuming they are to come together rationally to a unified decision) then they must have a way of resolving that division. This procedure would constitute a zeroing in on a single notion of permissibility.
Perhaps the resulting notion would have a disjunctive character. (Perhaps a group of lawyers and moral saints mixed together would decide that, for their purposes, an action is merciful so long as all the obvious requirements apply and each of A and B is either legal or moral.) That's still a single notion of permissibility. And in fact one could use this fact to argue that FAM itself can turn out a judgment of "merciful" on the action described in case 7—at least in those cases where some such disjunctive notion of permissibility turns out to be the right one. (One begins to suspect here, though, that any act could be merciful given the right audience—which seems strange enough to count at least a little against FAM.)
But it's not clear at all that a disjunctive notion of permissibility is ever the right one, much less when it would be the right one. (The lawyer/saint scenario briefly alluded to above seems very implausible on examination. Plausibly they would, or should, actually simply agree they have no common ground upon which to make a single judgment about the act, and would disband.)
So then, why do I think the action in Case 7 is merciful? Basically, it is because my idea of a judge includes the idea that a judge is trained to very carefully set aside his own biases and outside information, and instead to carry out his function as a representative of the government. A principled judge, for example, will uphold a guilty verdict on appeal even if he thinks the convict to be actually innocent, so long as all the rules of the courtroom were followed during the convict's trial. One could question whether upholding the verdict is moral (and one could argue that it is, but this is not relevant for our purposes) but there is, in any case, certainly a principle at work, moving the judge to make such a decision, which is not simply a principle of survival or comfort-seeking.
So I propose a different account of mercy, which I'll call the Internalized Principles Analysis of Mercy:
IPAM: x is merciful to y at t =df x believes there are two alternative acts, A and B, available to x at t such that (1) x performs A, (ii) x believes x's performing B would have resulted in more harm to y than x's performing A, (iii) x is motivated by compassion for y to choose A over B, (iv) two unselfish principles p1 and p2 move x, such that p1 moves x to choose A and p2 moves x to choose B. (p1 and p2 may be identical, as some principles allow for more than one option in a given situation.)
What does it mean for a principle to move a person to do an action? I am not sure if this is something that needs a lot of clarification or not. As a first pass, I'd say it just means that the principle is part of the rational calculus that seems to the person to count in favor of doing the action. I don't claim that this must be explicit or occurant to the person, just that it is part of the rational calculus that goes on in the person's mind in some way.
This account of mercy, unlike Markosian's, is non-contextualist. I have nothing in particular against contextualism per se, but mercy does not appear to me to be a contextual concept. I offer this account as an alternative to the contextual account. A strength of the contextual account is that it explains why intuitions about mercy differ. (People in different contexts will make different correct judgments.) I suspect (but have not really thought deeply about this) that IPAM also offers an explanation for this along the following lines: People may have different judgments about mercy because they bring to the judgment different assumptions about which principles genuinely move the agents described in various cases.
In near future posts, I'll discuss some of those cases, see how IPAM applies to them, and also explore the question whether IPAM may help explain differing intuitions about those cases.