The Golden Rule
The Golden Rule (GR) is probably the most well-known ethical principle among non-philosophers. Its standard formulation is familiar, but I reproduce it here for convenience:
(GR) Do unto others as you would have them do unto you.
As stated, GR does not tell us what actions are right. Nor does it tell us what actions are wrong. It is an imperative, and so it merely tells us what to do. On one plausible interpretation, however, GR is intended to tell us to perform an action just in case it is right. If so, we can use GR to formulate an analysis of rightness. [First, though, some preliminaries concerning actions are in order. (1) I take actions to be events of a certain sort (presumably they are to be distinguished from events that are non-actions by their special relation to intentions). On this view A-Team's kicking T-Ball at time t is an action, but kicking, kicking T-Ball, and kicking T-Ball at time t are not actions. (2) Every action has associated with it an agent. The agent of an action is the object who does or performs that action. In the case of A-Team's kicking T-Ball at time t, the agent is A-Team. (3) Some actions have associated with them a patient. The patient of an action is the object that action is done to. In the case of A-Team's kicking T-Ball at time t, the patient is T-Ball. (4) A single action may be of many different action types. So, for instance, A-Team's kicking T-Ball at time t is a kicking action, a kicking-T-ball action, a kicking-T-ball-at-time-t action, an action that occurs at time t, an action whose agent is A-Team, an action whose patient is T-Ball, etc.] This may be done as follows:
The Golden Analysis of Rightness (GAR): Necessarily, for all x, x is right iff: (i) x is an action and (ii) there are a y and a z such that: y is the agent of x, z is the patient of x, and there is an action type T such that x is of type T and y would have z be the agent of an action of type T whose patient is y.
It seems to me, however, that GAR is false. Regardless, it is certainly imprecise. In particular, consider the phrase 'y would have z be the agent of an action of type T whose patient is y'. Superficially, this phrase appears to be the consequent of a counterfactual conditional. But we are not given its antecedent.
In addition, there is some reason to think that, despite its superficial appearance, 'y would have z be the agent of an action of type T whose patient is y' is not intended to be the consequent of a counterfactual conditional at all. One reason to think this is that we are never given the associated antecedent. Another is that, colloquially, the 'would have' locution is often used to express desire. Suppose, for instance, that A is a student and B is A's professor. If A complains of one of B's policies, B might respond to A by saying: 'What would you have me do instead?' Plausibly, B's question can be paraphrased as: 'What do you want me to do instead?' Supposing that the 'would have' in 'y would have z be the agent of an action of type T whose patient is y' is to be taken in the same manner, we may paraphrase GAR as:
GAR.1: Necessarily, for all x, x is right iff: (i) x is an action and (ii) there are a y and a z such that: y is the agent of x, z is the patient of x, and there is an action type T such that x is of type T and y wants z to be the agent of an action of type T whose patient is y.
GAR.1 is certainly false, however. For, as noted above, for every action A, there is a time t such that A is of the action type being an action that occurs at t. But then, if GAR.1 is correct and A-Team wants T-Ball to be the agent of an action of the action-type being an action that occurs at t1 whose patient is A-Team himself, then for all actions A such that (i) A is of the action type being an action that occurs at t1, (ii) A-Team is the agent of A, and (iii) T-Ball is the patient of A, A is right. But clearly this is not so.
To remedy this problem, the action types in question ought to be restricted to what we might call 'relational action types'. To illustrate: the actions A-Team's kicking T-Ball at t1 and Smith's kicking Jones at t2 are both of the relational action type kicking. Taking this restriction into account, we get:
GAR.2: Necessarily, for all x, x is right iff: (i) x is an action and (ii) there are a y and a z such that: y is the agent of x, z is the patient of x, and there is a relational action type T such that x is of type T and y wants z to be the agent of an action of type T whose patient is y.
Unfortunately, GAR.2 can also be shown to be false. For take any action types T1 and T2, and suppose that all and only x1, ..., xn are actions of type T1 and all and only y1, ..., yn are actions of type T2. Then there will be an action type such that all and only x1, ..., xn and y1, ..., yn are actions of that type; namely, a disjunctive action type formed from action types T1 and T2. So, for instance, all and only the actions of the relational action types kicking and shooting are actions of the action type kicking or shooting. But now suppose that A-Team wants T-Ball to be the agent of an action of some relational action type T whose patient is A-Team himself. Either A-Team has the further desire that T-Ball be the agent of an action of relational action type T or shooting which has A-Team himself as its patient, or he could easily form that desire by reasoning from the desire he already has. But then, for any T or shooting action, if A-Team is the agent of it and T-Ball is the patient of it, then (given GAR.2) that action either is or easily could be right (if A-Team goes through the necessary reasoning process). And that includes any shooting actions among the T or shooting actions. And surely that is false. In general, if GAR.2 is true, then for all x and for all y, if x wants y to do some action of some relational action type T to x, then for any action A (of action type T') whose agent is x and whose patient is y, that action is right (or easily could be right if x reasoned to a desire for y to do an action of type T or T' to x).
One might try to patch GAR.2 by restricting the action types in question once more. So, for instance, we might offer GAR.3:
GAR.3: Necessarily, for all x, x is right iff: (i) x is an action and (ii) there are a y and a z such that: y is the agent of x, z is the patient of x, and there is a non-disjunctive relational action type T such that x is of type T and y wants z to be the agent of an action of type T whose patient is y.
However, GAR.3 is surely false as well. For suppose that there are many action types such that A-Team wants T-Ball to be the agent of an action those types whose patient is A-Team himself, but that all of those action types are disjunctive. Then (given GAR.3) for every action A such that A-Team is the agent of A and T-Ball is the patient of A, it is not the case that A is right. And, of course, that is false: A-Team can perform a right action whose patient is T-Ball even though he has no desires regarding T-Ball's performing an action on him of a non-disjunctive relational action type.
As a matter of fact, variations of the problem highlighted in the last paragraph can be offered against both GAR.2 and GAR.1, as well. In response to GAR.2: A-Team can perform a right action whose patient is T-Ball even though he has no desires regarding T-Ball's performing an action on him of any specific relational action type. In response to GAR.1: A-Team can perform a right action whose patient is T-Ball even though he has no desires regarding T-Ball's performing an action on him of any specific action type. (Suppose, for instance, that A-Team has no desires regarding T-Ball's performing any actions.)
So, as shown by the problems mentioned in the previous two paragraphs, the conditions on the laid down in GAR.1-GAR.3 are not necessary for an action to be right. In addition, as the other problems raised for GAR.1 and GAR.2 showed, the conditions they lay down are not sufficient for an action to be right. (I also highly suspect that the conditions laid down by GAR.3 are not sufficient for an action to be right. However, I will not argue for that claim here since doing so would require performing the quite difficult task of finding an uncontroversial example of a non-disjunctive relational action type.) But for any proposed analysis of the property being right, showing either that the conditions laid down in that analysis are not necessary for an action to be right or that they are not sufficient for an action to be right is sufficient to show that that proposed analysis is incorrect. So, GAR.1-GAR.3 are all incorrect. In addition, enough has been said in the past few paragraphs to show that no version of GAR that interprets the phrase 'y would have z be the agent of an action of type T whose patient is y' in the way that GAR.1-GAR.3 all do (that is, as involving y's having a desire towards z to the effect that z perform an action of a certain sort on y) is correct. None of them will provide a necessary condition on an action's being right, since someone can perform a right action on someone else without having any desire to the effect that the second person perform an action of any type on the first.
[Notice that even if the defender of a GAR.1-style analysis of rightness can overcome the mainly technical problems thus far noted against his view, other problems will remain. Suppose, for instance, that A-Team wants T-Ball to shoot him; that is, suppose that A-Team wants T-Ball to be the agent of an action of relational type shooting whose patient is A-Team. Then, on a GAR.1-style analysis, any action of the relational action type shooting with A-Team as its agent and T-Ball as its patient is a right action. But that's false. For suppose that A-Team has a fatal disease that causes him chronic pain, and wants T-Ball to shoot him to end his life. Suppose also that T-Ball is a good man who has done nothing wrong, is quite happy, is in no pain, etc. Then it is wrong for A-Team to shoot T-Ball, contra a GAR.1-style analysis.]
The conclusion of the discussion thus far, then, is that all attempted paraphrases of GAR in which the phrase 'y would have z be the agent of an action of type T whose patient is y' is interpreted in the way in which all of GAR.1-GAR.3 interpret it are false. Thus, to give GAR (and GR) their due, we ought to explore how to spell out GAR in a way that gives that phrase its other plausible interpretation; namely, as the consequent of a counterfactual conditional. Let us call all such ways of spelling out GAR its "counterfactual readings".
Upon hearing objections to GR similar to those given above against GAR.1-GAR.3, some people respond by saying that the objector is misunderstanding GR. And, in fact, they offer the rudiments of a counterfactual reading of GAR. According to these people, what GR asks you to do when performing an action is to imagine that your situation is reversed with that of the patient of that action, and to do to that person what you would want that person to do to you if your situations were reversed.
One way to understand this proposed reading of GAR is as follows:
GAR.4: Necessarily, for all x, x is right iff: (i) x is an action and (ii) there are a y, a z, and a T such that: y is the agent of x, z is the patient of x, T is a relational action type such that x is of type T, and if it were the case that x has every property y in fact has and y has every property x in fact has, then x would want y to be the agent of an action of type T whose patient is x.
Unfortunately, it seems that any instance of 'if it were the case that x has every property y in fact has and y has every property x in fact has, then x would want y to be the agent of an action of type T whose patient is x' in which x is distinct from y will have a necessarily false antecedent. For instance, take
(S) 'if it were the case that A-Team has every property T-Ball in fact has and T-Ball has every property A-Team in fact has, then A-Team would want T-Ball to be the agent of an action of type T whose patient is A-Team'.
Now one property that A-Team in fact has is the property being identical to A-Team and one property T-Ball in fact has is the property being identical to T-Ball. But it is not the case that possibly, A-Team has the property being identical to T-Ball, and it is not the case that possibly, T-Ball has the property being identical to A-Team. So, (S) has a necessarily false antecedent. Therefore, given the standard Stalnaker-Lewis semantics for counterfactuals, (S) is vacuously true. But then, if GAR.4 is true, for any action performed by anyone on anyone, that action is right. But some actions are wrong. So, it is not the case that GAR.4 is true.
There are two sorts responses that a proponent of GAR.4 might make to this worry. The first sort of response is to call for a re-evaluation of the Lewis-Stalnaker semantics for counterfactuals. To make this response palatable, the proponent would probably cite instances in which that semantics makes seemingly false predictions.
Now I am, to some extent, sympathetic to the view that the Stalnaker-Lewis semantics for counterfactuals is mistaken. However, this first sort of response cannot work because there is a distinct problem with GAR.4 that would not be solved by offering an alternate semantics for counterfactuals. Suppose, for instance, that T-Ball in fact wants T-Ball to be the agent of an action of type T whose patient is A-Team. Then, if it were the case that A-Team has every property T-Ball in fact has and T-Ball has every property A-Team in fact has, then A-Team would want T-Ball to be the agent of an action of type T whose patient is A-Team. If GAR.4 is true, then in such a case, for any action of type T whose agent is A-Team and whose patient is T-Ball, that action is right. But it is not the case that in such a case, for any action of type T whose agent is A-Team and whose patient is T-Ball, that action is right. So, it is not the case that GAR.4 is true. (For instance, suppose that T-Ball wants to kill A-Team. It is not the case, simply in virtue of this fact, that it is right for A-Team to kill T-Ball. For it might be the case that T-Ball would never act upon his desires or, supposing that he did, it might be the case that there is a way for A-Team to prevent T-Ball from killing A-Team short of killing T-Ball.)
The second sort of response that a proponent of GAR.4 might make is to restrict the properties in question. Suppose, for instance, that he restricts the properties in question to sort S. Then he might offer the following revision:
GAR.5: Necessarily, for all x, x is right iff: (i) x is an action and (ii) there are a y, a z, and a T such that: y is the agent of x, z is the patient of x, T is a relational action type such that x is of type T, and if it were the case that x has every property of sort S y in fact has and y has every property of sort S x in fact has, then x would want y to be the agent of an action of type T whose patient is x.
The main problem with GAR.5 is that, presumably, some action that A-Team performs on T-Ball can be right even if: if it were the case that A-Team has every property of sort S T-Ball in fact has and T-Ball has every property of sort S A-Team in fact has, then it would be the case that there is no relational action type T such that A-Team wants T-Ball to be the agent of an action of type T whose patient is A-Team. If so, though, the conditions laid out by GAR.5 are not necessary for an action to be right. Therefore, GAR.5 is not true.
Notice, though, that any analysis of rightness along the lines of GAR.4 and GAR.5 succumbs to a version of the problem presented in the last paragraph. Therefore, the counterfactual readings of GAR, like the other readings, fail to correct analyses of rightness. But there are no readings of GAR besides these two sorts of readings. Therefore, GAR is false and GR fails to provide a reliable guide to right action.
(GR) Do unto others as you would have them do unto you.
As stated, GR does not tell us what actions are right. Nor does it tell us what actions are wrong. It is an imperative, and so it merely tells us what to do. On one plausible interpretation, however, GR is intended to tell us to perform an action just in case it is right. If so, we can use GR to formulate an analysis of rightness. [First, though, some preliminaries concerning actions are in order. (1) I take actions to be events of a certain sort (presumably they are to be distinguished from events that are non-actions by their special relation to intentions). On this view A-Team's kicking T-Ball at time t is an action, but kicking, kicking T-Ball, and kicking T-Ball at time t are not actions. (2) Every action has associated with it an agent. The agent of an action is the object who does or performs that action. In the case of A-Team's kicking T-Ball at time t, the agent is A-Team. (3) Some actions have associated with them a patient. The patient of an action is the object that action is done to. In the case of A-Team's kicking T-Ball at time t, the patient is T-Ball. (4) A single action may be of many different action types. So, for instance, A-Team's kicking T-Ball at time t is a kicking action, a kicking-T-ball action, a kicking-T-ball-at-time-t action, an action that occurs at time t, an action whose agent is A-Team, an action whose patient is T-Ball, etc.] This may be done as follows:
The Golden Analysis of Rightness (GAR): Necessarily, for all x, x is right iff: (i) x is an action and (ii) there are a y and a z such that: y is the agent of x, z is the patient of x, and there is an action type T such that x is of type T and y would have z be the agent of an action of type T whose patient is y.
It seems to me, however, that GAR is false. Regardless, it is certainly imprecise. In particular, consider the phrase 'y would have z be the agent of an action of type T whose patient is y'. Superficially, this phrase appears to be the consequent of a counterfactual conditional. But we are not given its antecedent.
In addition, there is some reason to think that, despite its superficial appearance, 'y would have z be the agent of an action of type T whose patient is y' is not intended to be the consequent of a counterfactual conditional at all. One reason to think this is that we are never given the associated antecedent. Another is that, colloquially, the 'would have' locution is often used to express desire. Suppose, for instance, that A is a student and B is A's professor. If A complains of one of B's policies, B might respond to A by saying: 'What would you have me do instead?' Plausibly, B's question can be paraphrased as: 'What do you want me to do instead?' Supposing that the 'would have' in 'y would have z be the agent of an action of type T whose patient is y' is to be taken in the same manner, we may paraphrase GAR as:
GAR.1: Necessarily, for all x, x is right iff: (i) x is an action and (ii) there are a y and a z such that: y is the agent of x, z is the patient of x, and there is an action type T such that x is of type T and y wants z to be the agent of an action of type T whose patient is y.
GAR.1 is certainly false, however. For, as noted above, for every action A, there is a time t such that A is of the action type being an action that occurs at t. But then, if GAR.1 is correct and A-Team wants T-Ball to be the agent of an action of the action-type being an action that occurs at t1 whose patient is A-Team himself, then for all actions A such that (i) A is of the action type being an action that occurs at t1, (ii) A-Team is the agent of A, and (iii) T-Ball is the patient of A, A is right. But clearly this is not so.
To remedy this problem, the action types in question ought to be restricted to what we might call 'relational action types'. To illustrate: the actions A-Team's kicking T-Ball at t1 and Smith's kicking Jones at t2 are both of the relational action type kicking. Taking this restriction into account, we get:
GAR.2: Necessarily, for all x, x is right iff: (i) x is an action and (ii) there are a y and a z such that: y is the agent of x, z is the patient of x, and there is a relational action type T such that x is of type T and y wants z to be the agent of an action of type T whose patient is y.
Unfortunately, GAR.2 can also be shown to be false. For take any action types T1 and T2, and suppose that all and only x1, ..., xn are actions of type T1 and all and only y1, ..., yn are actions of type T2. Then there will be an action type such that all and only x1, ..., xn and y1, ..., yn are actions of that type; namely, a disjunctive action type formed from action types T1 and T2. So, for instance, all and only the actions of the relational action types kicking and shooting are actions of the action type kicking or shooting. But now suppose that A-Team wants T-Ball to be the agent of an action of some relational action type T whose patient is A-Team himself. Either A-Team has the further desire that T-Ball be the agent of an action of relational action type T or shooting which has A-Team himself as its patient, or he could easily form that desire by reasoning from the desire he already has. But then, for any T or shooting action, if A-Team is the agent of it and T-Ball is the patient of it, then (given GAR.2) that action either is or easily could be right (if A-Team goes through the necessary reasoning process). And that includes any shooting actions among the T or shooting actions. And surely that is false. In general, if GAR.2 is true, then for all x and for all y, if x wants y to do some action of some relational action type T to x, then for any action A (of action type T') whose agent is x and whose patient is y, that action is right (or easily could be right if x reasoned to a desire for y to do an action of type T or T' to x).
One might try to patch GAR.2 by restricting the action types in question once more. So, for instance, we might offer GAR.3:
GAR.3: Necessarily, for all x, x is right iff: (i) x is an action and (ii) there are a y and a z such that: y is the agent of x, z is the patient of x, and there is a non-disjunctive relational action type T such that x is of type T and y wants z to be the agent of an action of type T whose patient is y.
However, GAR.3 is surely false as well. For suppose that there are many action types such that A-Team wants T-Ball to be the agent of an action those types whose patient is A-Team himself, but that all of those action types are disjunctive. Then (given GAR.3) for every action A such that A-Team is the agent of A and T-Ball is the patient of A, it is not the case that A is right. And, of course, that is false: A-Team can perform a right action whose patient is T-Ball even though he has no desires regarding T-Ball's performing an action on him of a non-disjunctive relational action type.
As a matter of fact, variations of the problem highlighted in the last paragraph can be offered against both GAR.2 and GAR.1, as well. In response to GAR.2: A-Team can perform a right action whose patient is T-Ball even though he has no desires regarding T-Ball's performing an action on him of any specific relational action type. In response to GAR.1: A-Team can perform a right action whose patient is T-Ball even though he has no desires regarding T-Ball's performing an action on him of any specific action type. (Suppose, for instance, that A-Team has no desires regarding T-Ball's performing any actions.)
So, as shown by the problems mentioned in the previous two paragraphs, the conditions on the laid down in GAR.1-GAR.3 are not necessary for an action to be right. In addition, as the other problems raised for GAR.1 and GAR.2 showed, the conditions they lay down are not sufficient for an action to be right. (I also highly suspect that the conditions laid down by GAR.3 are not sufficient for an action to be right. However, I will not argue for that claim here since doing so would require performing the quite difficult task of finding an uncontroversial example of a non-disjunctive relational action type.) But for any proposed analysis of the property being right, showing either that the conditions laid down in that analysis are not necessary for an action to be right or that they are not sufficient for an action to be right is sufficient to show that that proposed analysis is incorrect. So, GAR.1-GAR.3 are all incorrect. In addition, enough has been said in the past few paragraphs to show that no version of GAR that interprets the phrase 'y would have z be the agent of an action of type T whose patient is y' in the way that GAR.1-GAR.3 all do (that is, as involving y's having a desire towards z to the effect that z perform an action of a certain sort on y) is correct. None of them will provide a necessary condition on an action's being right, since someone can perform a right action on someone else without having any desire to the effect that the second person perform an action of any type on the first.
[Notice that even if the defender of a GAR.1-style analysis of rightness can overcome the mainly technical problems thus far noted against his view, other problems will remain. Suppose, for instance, that A-Team wants T-Ball to shoot him; that is, suppose that A-Team wants T-Ball to be the agent of an action of relational type shooting whose patient is A-Team. Then, on a GAR.1-style analysis, any action of the relational action type shooting with A-Team as its agent and T-Ball as its patient is a right action. But that's false. For suppose that A-Team has a fatal disease that causes him chronic pain, and wants T-Ball to shoot him to end his life. Suppose also that T-Ball is a good man who has done nothing wrong, is quite happy, is in no pain, etc. Then it is wrong for A-Team to shoot T-Ball, contra a GAR.1-style analysis.]
The conclusion of the discussion thus far, then, is that all attempted paraphrases of GAR in which the phrase 'y would have z be the agent of an action of type T whose patient is y' is interpreted in the way in which all of GAR.1-GAR.3 interpret it are false. Thus, to give GAR (and GR) their due, we ought to explore how to spell out GAR in a way that gives that phrase its other plausible interpretation; namely, as the consequent of a counterfactual conditional. Let us call all such ways of spelling out GAR its "counterfactual readings".
Upon hearing objections to GR similar to those given above against GAR.1-GAR.3, some people respond by saying that the objector is misunderstanding GR. And, in fact, they offer the rudiments of a counterfactual reading of GAR. According to these people, what GR asks you to do when performing an action is to imagine that your situation is reversed with that of the patient of that action, and to do to that person what you would want that person to do to you if your situations were reversed.
One way to understand this proposed reading of GAR is as follows:
GAR.4: Necessarily, for all x, x is right iff: (i) x is an action and (ii) there are a y, a z, and a T such that: y is the agent of x, z is the patient of x, T is a relational action type such that x is of type T, and if it were the case that x has every property y in fact has and y has every property x in fact has, then x would want y to be the agent of an action of type T whose patient is x.
Unfortunately, it seems that any instance of 'if it were the case that x has every property y in fact has and y has every property x in fact has, then x would want y to be the agent of an action of type T whose patient is x' in which x is distinct from y will have a necessarily false antecedent. For instance, take
(S) 'if it were the case that A-Team has every property T-Ball in fact has and T-Ball has every property A-Team in fact has, then A-Team would want T-Ball to be the agent of an action of type T whose patient is A-Team'.
Now one property that A-Team in fact has is the property being identical to A-Team and one property T-Ball in fact has is the property being identical to T-Ball. But it is not the case that possibly, A-Team has the property being identical to T-Ball, and it is not the case that possibly, T-Ball has the property being identical to A-Team. So, (S) has a necessarily false antecedent. Therefore, given the standard Stalnaker-Lewis semantics for counterfactuals, (S) is vacuously true. But then, if GAR.4 is true, for any action performed by anyone on anyone, that action is right. But some actions are wrong. So, it is not the case that GAR.4 is true.
There are two sorts responses that a proponent of GAR.4 might make to this worry. The first sort of response is to call for a re-evaluation of the Lewis-Stalnaker semantics for counterfactuals. To make this response palatable, the proponent would probably cite instances in which that semantics makes seemingly false predictions.
Now I am, to some extent, sympathetic to the view that the Stalnaker-Lewis semantics for counterfactuals is mistaken. However, this first sort of response cannot work because there is a distinct problem with GAR.4 that would not be solved by offering an alternate semantics for counterfactuals. Suppose, for instance, that T-Ball in fact wants T-Ball to be the agent of an action of type T whose patient is A-Team. Then, if it were the case that A-Team has every property T-Ball in fact has and T-Ball has every property A-Team in fact has, then A-Team would want T-Ball to be the agent of an action of type T whose patient is A-Team. If GAR.4 is true, then in such a case, for any action of type T whose agent is A-Team and whose patient is T-Ball, that action is right. But it is not the case that in such a case, for any action of type T whose agent is A-Team and whose patient is T-Ball, that action is right. So, it is not the case that GAR.4 is true. (For instance, suppose that T-Ball wants to kill A-Team. It is not the case, simply in virtue of this fact, that it is right for A-Team to kill T-Ball. For it might be the case that T-Ball would never act upon his desires or, supposing that he did, it might be the case that there is a way for A-Team to prevent T-Ball from killing A-Team short of killing T-Ball.)
The second sort of response that a proponent of GAR.4 might make is to restrict the properties in question. Suppose, for instance, that he restricts the properties in question to sort S. Then he might offer the following revision:
GAR.5: Necessarily, for all x, x is right iff: (i) x is an action and (ii) there are a y, a z, and a T such that: y is the agent of x, z is the patient of x, T is a relational action type such that x is of type T, and if it were the case that x has every property of sort S y in fact has and y has every property of sort S x in fact has, then x would want y to be the agent of an action of type T whose patient is x.
The main problem with GAR.5 is that, presumably, some action that A-Team performs on T-Ball can be right even if: if it were the case that A-Team has every property of sort S T-Ball in fact has and T-Ball has every property of sort S A-Team in fact has, then it would be the case that there is no relational action type T such that A-Team wants T-Ball to be the agent of an action of type T whose patient is A-Team. If so, though, the conditions laid out by GAR.5 are not necessary for an action to be right. Therefore, GAR.5 is not true.
Notice, though, that any analysis of rightness along the lines of GAR.4 and GAR.5 succumbs to a version of the problem presented in the last paragraph. Therefore, the counterfactual readings of GAR, like the other readings, fail to correct analyses of rightness. But there are no readings of GAR besides these two sorts of readings. Therefore, GAR is false and GR fails to provide a reliable guide to right action.
posted by Greg at 5:23 PM
0 comments
