Transcript PowerPoint
AGAINST MORAL CHARACTER EVALUATIONS: The undetectability of virtue and vice Peter B. M. Vranas Iowa State University Conference on Virtue Ethics and Moral Psychology, 7 October 2005 THE EPISTEMIC THESIS Epistemic thesis: Moral character evaluations are almost always epistemically unwarranted. Definitions: Moral character evaluations: evaluations of people as good, bad, or intermediate. A person is indeterminate iff the person is neither good nor bad nor intermediate. A person is fragmented iff the person would behave deplorably in an open list of situations and admirably in another such open list. THE ARGUMENT FOR THE EPISTEMIC THESIS (P1) Most people are fragmented. (L1) The prior probability that a person is fragmented should be high. (P2) The posterior probability that a person is fragmented shouldn’t differ much from the prior. (L2) The posterior probability that a person is fragmented should be high. (P3) Fragmentation entails indeterminacy. (C1) The posterior probability that a person is indeterminate should be high. ___________________________________________________________________________________________________________________________________________________________ ___________________________________________________________________________________________________________________________________________________________ ___________________________________________________________________________________________________________________________________________________________ OVERVIEW Part 1: Review of previous results Most people are fragmented (P1) Fragmentation entails indeterminacy (P3) Part 2: Posterior probability of fragmentation Approximate IndependencePosterior @ Prior Approximate Independence holds Part 3: Objections to the epistemic thesis The triviality objection The objection from ought-implies-can The objection from comparative evaluations (P1) MOST PEOPLE ARE FRAGMENTED (1) There is an open list of situations in which most people would behave deplorably: Obedience experiments (Milgram) Stanford Prison Experiment (Zimbardo) Seizure experiments (Latané & Darley) (2) There is an open list of situations in which most people would behave admirably: Electrocution experiments (Clark & Word) Theft experiments (Moriarty) Rape experiments (Harari et al.) (P3) FRAGMENTATION ENTAILS INDETERMINACY (P0) If A behaves much better than B in an open list of situations and much worse in another list, then A is neither better nor worse than B. (1) Every fragmented person is neither better nor worse than some intermediate person (who never behaves deplorably or admirably). (2) Every good person is better, and every bad person is worse, than any intermediate person. (3) No fragmented person is good or bad. Similarly, no fragmented person is intermediate. __________________________________________________________________________________________________________________________________________________________ __________________________________________________________________________________________________________________________________________________________ PART 2 Part 1: Review of previous results Most people are fragmented (P1) Fragmentation entails indeterminacy (P3) Part 2: Posterior probability of fragmentation Approx. IndependencePosterior@Prior Approx. Independence holds Part 3: Objections to the epistemic thesis The triviality objection The objection from ought-implies-can The objection from comparative evaluations THE POSTERIOR PROBABILITY OF FRAGMENTATION Notation: F: person p is fragmented. Ds: p behaves deplorably in situation s. Approximate Independence Condition: P(Ds|Ds) @ P(Ds)... The argument: (P4) If Approximate Independence holds, then P(F|Ds) shouldn’t differ much from P(F). (P5) Approximate Independence holds. (P2) P(F|Ds) shouldn’t differ much from P(F). ________________________________________________________________________________________________________________________________________________________ (P4) INDEPENDENCE POSTERIOR @ PRIOR Theorem. Consider S independent and identically distributed random variables, each of which can take the values -1, 1, and 0 with probabilities pD ,pA , and 1-pD-pA respectively. Let ND and NA be the numbers of these variables which take the values -1 and 1 respectively, and ND and NA the corresponding numbers for (any) S-1 of the S variables. Let F = ND > sD & NA > sA and let Ds = the s-th variable takes the value -1. Then: P(F|Ds)-P(F) = (1-pD)P(ND = sD & NA > sA) -pAP(ND > sD & NA = sA). (P5) APPROXIMATE INDEPENDENCE HOLDS Argument 1: Average correlation coefficients low. E.g., Hartshorne & May (1928, 1930). Objection: We can predict our friends’ actions. Reply: Behavior may be temporally stable (in recurring situations) but not crosssituationally consistent. In everyday life we observe our friends in recurring situations. Argument 2: Personality characteristics like religiosity, authoritarianism, introversion don’t predict who obeys in Milgram’s experiments. PART 3 Part 1: Review of previous results Most people are fragmented (P1) Fragmentation entails indeterminacy (P3) Part 2: Posterior probability of fragmentation Approximate IndependencePosterior @ Prior Approximate Independence holds Part 3: Objections to the epistemic thesis The triviality objection The objection from ought-implies-can The objection from comparative evaluations OBJECTION 1: TRIVIALITY The objection: (1) We seldom make MCEs, so (2) the epistemic thesis is uninteresting. Reply 1: (1) is false. Empirical evidence: Never Almost never Rarely Somewhat Somewhat Frequ- Very Almost rarely frequently ently frequently always _______________________________________________________________________________________________________________________________________________ #1: 3 20 30 46 54 26 5 5 #2: 0 5 10 25 53 53 34 8 Reply 2: (2) does not follow from (1). MCEs sometimes have important functions: they underlie esteem and contempt, praise and blame, explanations of behavior, and advice. OBJECTION 2: OUGHT-IMPLIES-CAN The objection: (1) We cannot avoid making MCEs (cf. Spontaneous Trait Inferences, STIs). (2) We have no epistemic obligation to avoid believing what we cannot avoid believing. (3) The epistemic thesis implies that we have an epistemic obligation to avoid making MCEs. Reply 1: (1) is false. STIs may be controllable. Reply 2: (2) is false. We have an epistemic obligation to avoid believing P iff our evidence makes P unlikely, but this can happen even if we cannot avoid believing P. OBJECTION 3: COMPARATIVE EVALUATIONS The objection: (1) If p is better than at least one good person, then p is good. (2) P(p is better than at least one good person|p is better than most people) high. Thus: (3) P(p is good|p is better than most people) should be high. So MCEs are justified if comparative MCEs are. Against (2): We don’t know what percentage of people are good, so even if p is better than (e.g.) 80% of people we can’t be confident that p is better than at least one good person. CONCLUSION: THE PRAGMATIC THESIS Local evaluations: referring to behavior in a relatively narrow range of situations. The pragmatic thesis: there is good pragmatic reason to prefer local to global evaluations. The argument: (1) Local evaluations avoid costs of global ones: tempting one’s luck, testing people. (2) Local evaluations preserve benefits of global ones: decisions about association, regulation of emotions. EPSTEIN AND AGGREGATION Epstein: Not surprising that correlations low. If you measure once, large error. You need to average many measurements. Correlations between aggregated measures will be high. Reply 1: Epstein’s argument presupposes constant true score, as when measuring length. But this presupposes cross-situational consistency. The issue is empirical, not a priori. Reply 2: Empirical evidence against Epstein; e.g., Hartshorne, May, & Shuttleworth 1930. THE VALIDITY OF THE ARGUMENT FOR P1 Theorem 1. Consider P people, SD situations in each of which at least pD people behave deplorably, and SA different situations in each of which at least pA people behave admirably. Let F be the number of people each of whom behaves deplorably in more than sD of the former SD situations and behaves admirably in more than sA of the latter SA situations. Then: F p D / P s D / SD p A / P s A / S A 1. P 1 s D / SD 1 s A / S A THREE CONCEPTIONS OF CHARACTER EVALUATIONS Q6 true (consistency of bad people) Q5 true (con- Consistency sistency of conceptions good people) (middle line) Q5 false (no ______ consistency of good people) Q6 false (no consistency of bad people) Impurity conceptions (hard line) Averaging conceptions (soft line) (P3) FRAGMENTATION ENTAILS INDETERMINACY (Q5) A person who often behaves deplorably is not good. (Q6) A person who often behaves admirably is not bad. (Q7) A person who often behaves deplorably and often behaves admirably is not intermediate (between good and bad). (P3) A person who often behaves deplorably and often behaves admirably is neither good nor bad nor intermediate. ___________________________________________________________________________________________________________________________________________________________ (Q7) FRAGMENTATION PRECLUDES “INTERMEDIACY” (Q10) Every good person is better than any intermediate person. (Q11) For any fragmented person f there is a good person g who is not better than f. (Q7) No fragmented person is intermediate. In support of Q11: Take f and g who behaves (a) admirably when f behaves deplorably, and (b) neutrally when f behaves admirably or neutrally. Then g is good but, by the Incommensurability Argument, g is not better than f. ____________________________________________________________________________________________________________________________________________________________ OBJECTIONS 1-2 TO Q5 Objection 1: Goodness of character depends on motives, regardless of whether acts deplorable. Reply: (a) I’m not denying that motives matter. (b) If the claim is that only motives matter, I disagree: weakness of will no adequate excuse. Objection 2: Counterfactual behavior is irrelevant to character because of “moral luck”. Reply: Counterfactual behavior is irrelevant to responsibility but can be relevant to character. OBJECTIONS 3-4 TO Q5 Objection 3: Certain counterfactuals irrelevant. You would have committed atrocities had you been raised in Nazi Germany. Reply: Only actual dispositions relevant to Q5. Objection 4: Even some actual dispositions irrelevant to character. You would kill if tortured. Reply: Irrelevant only if torture excuses killing; if it does, then killing after being tortured is not deplorable, so counterfactual irrelevant to Q5. OBJECTIONS 5-6 TO P5 Objection 5: Only extremely deplorable behavior precludes compensation. Crushing ants? Reply: Disagreement about antecedent of P5. Objection 6: People choose their situations. Reply: Take p1 and p2, who (know they) would kill if drunk. p1 drinks, p2 does not. Then p1 cp worse than p2, but both cp worse than p3, who would not kill if drunk. Disposition to choose situations does matter, but counterfactual behavior even in unchosen situations also matters. OBJECTION 7 TO P5 Objection 7: There is no fact of the matter about how you would behave in various situations. Reply: Only counterfactuals about whose truth there is a fact of the matter are relevant to P5. Rejoinder: What if all dispositions probabilistic? Reply: Two extreme cases. (1) For most people high probabilities. Then most people still fragmented. (2) For most people low probabilities. Then most people not fragmented but still indeterminate. AN OBJECTION TO Q6: WHAT ABOUT HITLER? Maybe Hitler was bad; consider Ted Bundy. Being nice to your mother is not admirable. A few isolated instances are no open list. Bundy was not better than an intermediate person who never behaves admirably. I’m not saying Bundy was good; only not bad. So badness cannot be fully compensated. Anchoring bias: good or bad news first? Symmetry argument: why give greater weight to deplorable than to admirable behavior? THE INCREDIBILITY OBJECTION First form: It’s incredible to deny that a brutal serial killer is bad just because in a psychological experiment he would (e.g.) stop a thief. Reply: (a) I’m not saying he’s good (or not bad). (b) I’m not relying just on experiments. (c) Anchoring bias: bad or good news first? Second form: I know my wife is good. It doesn’t matter that she would behave sadistically in an experiment; I don’t even believe she would. Reply: -You can’t predict. -Experiments matter.