Not for the Sake of Selfishness Alone
Related: Fake Selfishness, Not for the Sake of Pleasure Alone, Not for the Sake of Happiness (Alone), Value is Fragile, Fake Fake Utility Functions
No one deserves thanks from another about something he has done for him or goodness he has done. He is either willing to get a reward from God, therefore he wanted to serve himself. Or he wanted to get a reward from people, therefore he has done that to get profit for himself. Or to be mentioned and praised by people, therefore, it is also for himself. Or due to his mercy and tenderheartedness, so he has simply done that goodness to pacify these feelings and treat himself.
- Mohammed Ibn Al-Jahm Al-Barmaki
In a 1990 experiment, Jack Dovidio made subjects feel empathy for a young woman by asking subjects to imagine what she felt as she faced a particular problem.1 Half the subjects focused on one problem faced by the woman, while the other half focused on a different problem she faced. When given the opportunity to help the woman, subjects in the high empathy condition were more likely to help than subjects in the low empathy condition, and the increase was specific to the problem that had been used to evoke empathy.
What does this study say about altruism and selfishness?
Some people think that humans are purely selfish, that we act for selfish motives alone. They will re-interpret any counter-example you give (“But wouldn’t you sacrifice your life to save the rest of the human species?”) as being compatible with purely selfish motives.
Are they right? Do we act for selfish motives alone?
Let’s examine the evidence.2
We begin with a rough sketch of human motivation. We have ‘ultimate’ desires: things we desire for their own sake. We also have ‘instrumental’ desires: things we desire because we belief they will satisfy our ultimate desires.
I instrumentally desire to go to the kitchen because I ultimately desire to eat a brownie and I believe brownies are in the kitchen. But if I come to believe brownies are in the dining room and not the kitchen, I will instrumentally desire to walk to the dining room instead, to fulfill my ultimate desire to eat a brownie. Or perhaps my desire to eat a brownie is also an instrumental desire, and my ultimate desire is to taste something sweet, and I instrumentally desire to eat a brownie because I believe that eating a brownie will satisfy my desire to taste something sweet.
Of course, desires compete with each other. Perhaps I have an ultimate desire to taste something sweet, and thus I instrumentally desire to eat a brownie. But I also have an ultimate desire for regular sex, and I believe that eating a brownie will contribute to obesity that will lessen the chances of satisfying my desire for regular sex. In this case, the ‘stronger’ desire will determine my action.
The full picture is more complicated than this,3 but we only need a basic picture to assess the claim that we only act for selfish motives alone.
We might categorize ultimate desires like this:4
Psychological egoists think all ultimate desires are of type 2. Psychological hedonists are a subset of egoists who think that all ultimate desires are of type 1. Psychological altruists think that at least some ultimate desires are of type 4. If some ultimate desires are of type 3, but none are of type 4, then both egoism and altruism are false.
Previously, I presented neurobiological evidence that psychological hedonism is false. In short: desire and pleasure are encoded separately by the brain, and we sometimes desire things that are not aimed at producing pleasure, and in fact we sometimes desire things that do not produce pleasure when we get them.
But can we also disprove the claim that we act for selfish reasons alone (psychological egoism), by showing that normal humans have desires for the well-being of others?
The standard theory of altruism in psychology is the empathy-altruism hypothesis, which says that altruism exists in humans and is often the result of an empathic emotional response to another’s distress.5 Experiments have uncovered a plausible causal chain from perspective-taking to empathy to the motivation of helping behavior, and shown that this chain is probably incompatible with psychological egoism.
Of course, showing that empathy causes helping behavior does not win the day for altruism over egoism. The egoist may say that, for example, empathy causes sadness and that people are motivated to help because they believe helping is the best way to alleviate their own sadness. Or perhaps people believe that helping people in some circumstances will make them feel good, or that failing to help will make them feel bad, and this is what motivates them to help.
Daniel Batson and other researchers have spent several decades running experiments that allow the empathy-altruism hypothesis to be compared to specific versions of the egoism. We can’t review all those studies here,6 but let’s examine a few of them.
Testing the Aversive-Arousal Reduction Hypothesis
One egoistic hypothesis explains helping behavior by saying that the sight of someone in distress causes an aversive reaction, and this (and not empathy) causes a desire to relieve the aversive emotion by helping the person in distress.
The Dovidio experiment described above undermines the naive version of aversive-arousal reduction hypothesis, but the egoist may insist that one’s distress is increased when one feels empathy, and that the increased helping that follows empathy is due to increased distress. In contrast, the altruist maintains that empathy evokes an ultimate desire to help, which leads to helping behavior.
How can we test these competing hypotheses?
Batson argues that manipulating difficulty of escape allows us to compare these two hypotheses experimentally. The central idea is that if a subject is motivated by an ultimate desire to help the target, that desire can be satisfied only by helping. However, if a subject is motivated by a desire to reduce his own distress, that desire can be satisfied either by helping or by merely escaping from the distress-inducing situation—for example, by leaving the room so that one is no longer confronted by the needy target. Assuming that subjects do whatever is easier and less costly, the aversive-arousal reduction hypothesis thus predicts that even subjects experiencing empathy will simply leave the needy target, provided escape is made easy enough.
...To determine whether empathy is playing a role in producing helping behavior, he has to compare the behavior of low-empathy and high-empathy subjects. To determine whether ease of escape has any effect on the likelihood of helping behavior, he must arrange things so that leaving is significantly more costly for some subjects than for others. So there are four experimental conditions: low-empathy subjects where escape is either (1) easy or (2) hard, and high-empathy subjects where leaving is either (3) easy or (4) hard.7
Batson says that in the condition with high-empathy where escape is easy (condition 3), the aversive-arousal reduction hypothesis predicts a low level of helping behavior, while the empathy-altruism hypothesis predicts a high level of helping behavior.
Batson conducted 6 experiments to test these predictions. All 6 experiments confirmed the empathy-altruism hypothesis over the aversive-arousal reduction hypothesis. Let’s consider one of them.8
In one experiment
student subjects were required to watch, via what they believed to be closed-circuit TV, as another student subject, Elaine, attempted to perform a task while receiving electric shocks at random intervals. Observer subjects were told that their task would be to form and report an impression of how Elaine performs under aversive conditions. Actually, what the subjects were viewing was a videotape. On the tape, Elaine is clearly finding the shocks very uncomfortable, and after her second trial at doing the task, she explains to Martha, the assistant overseeing the experiment, that she is unusually sensitive to mild electric shocks because of a childhood trauma. Martha then suggests that perhaps the observer subject might be willing to help Elaine by taking her place, and the experimenter asks whether the subject is willing to do that. To manipulate ease of escape, some subjects are told that if they decide not to take Elaine’s place, they will be required to watch eight additional trials, while other subjects are told that if they decide not to take Elaine’s place, they will be free to go, although Elaine will have to endure eight more trials. To manipulate the level of empathy that subjects feel for Elaine, subjects are given a copy of a personal values and interests questionnaire, allegedly filled out by Elaine, in order to help them form an impression of her performance. In the high-empathy condition, Elaine’s values and interests are very similar to the subject’s (which had been determined in a screening session several weeks before), while in the low-empathy condition, they are very different.9
The results confirmed the empathy-altruism hypothesis over the aversive-arousal reduction hypothesis:10
Testing the Empathy-Specific Reward Hypothesis
Another egoistic hypothesis claims that helping behavior is motivated by the expectation of a reward. But this doesn’t explain Dovidio’s results—that empathy increases helping behavior. Still, the egoist might maintain that helping behavior is especially rewarding when one feels empathy for the distressed person. Call this the empathy-specific reward hypothesis.
This version of egoism predicts the helper will not be rewarded (by a jolt of pride, or whatever) if she is unable to relieve the target’s distress, either because there’s nothing she can do or because someone else helps the target before she can.
The empathy-altruism hypothesis says that people are motivated by an ultimate desire that the target’s distress be alleviated. The empathy-specific reward hypothesis (a version of egoism) predicts that the helper will not be rewarded if she is unable to personally relieve the target’s distress.
To test between the two hypotheses, Batson told11 participants that
they would likely have the chance to perform a simple task that would reduce the number of electric shocks that a peer would receive… Somewhat later, half of the participants learned, by chance, that they would not be performing the helping task after all, and thus that they could not help the other student. This divided the participants into two experimental conditions, ‘perform’ and ‘not perform.’ Subsequently, half of the participants in each condition learned that, by chance, the peer was not going to get the shocks, while the other half learned that, by chance, the peer would still have to get the shocks. This yielded two more experimental conditions, ‘prior relief’ and ‘no prior relief’. All participants were also asked to self-report their level of empathy for the peer, so that high- and low-empathy participants could be distinguished. To assess mood change, the moods of all participants were measured both before and after the experimental manipulation. As we saw above, the version of the empathy-specific reward hypothesis that we’re considering predicts that participants in the perform + no prior relief condition should indicate an elevated mood, since they were able to help the peer; it also predicts that participants in all the other conditions should not have an elevated mood, since for one reason or another they were unable to help, and thus were ineligible for the reward. The empathy–altruism hypothesis, by contrast, predicts an elevated mood in all three conditions in which the peer escaped the shocks: perform + no prior relief, perform + prior relief, and not perform + prior relief. The only condition in which empathy–altruism predicts low mood is the one in which the peer gets the shocks: not perform + no prior relief.12
Again, the results confirmed the empathy-altruism hypothesis over the empathy-specific reward hypothesis.
Conclusion
In these and many other experiments, the empathy-altruism hypothesis has been confirmed over a wide variety of naive and sophisticated versions of egoism.
Batson concludes:
In study after study, with no clear exceptions, we find results conforming to the pattern predicted by the empathy–altruism hypothesis, the hypothesis that empathic emotion evokes altruistic motivation. At present, there is no egoistic explanation for the results of these studies… Pending new evidence or a plausible new egoistic explanation for the existing evidence, the empathy–altruism hypothesis… seems to be true.13
Thus, it seems we probably do not act for the sake of selfishness alone.
Notes
1 This technique for inducing empathy in subjects had previous been tested by Stotland (1969) and others.
2 This article draws heavily from the more detailed review of the evidence available in Stich et al. (2010). That article also reviews evolutionary hypotheses about altruism, which the authors find (as yet) unpersuasive. Also see Eisenberg & Miller (1987); Dovidio et al. (2006).
3 For a fuller discussion of concepts of desire and human motivation, see Schroeder (2004). For a more recent explanation of how human motivation works, see Glimcher (2010).
4 Image from Stich et al. (2010).
5 By ‘empathy’, I mean something like Batson’s (1991: 86) stipulated definition in terms of “feeling sympathetic, compassionate, warm, softhearted, tender, and the like.” By ‘distress’ I mean something like Batson’s (1991: 117) stipulated definition in terms of “self-oriented feelings such as upset, alarm, anxiety, and distress.”
6 But, see Batson (2011); Batson et al. (1991, 1998).
7 Stich et al. (2010), pp. 177-179.
8 Batson et al. (1981), experiment 1.
9 Stich et al. (2010), pp. 180-181.
10 Image from Stich et al. (2010), p. 181.
11 Batson et al. (1988), experiment 1.
12 Stich et al. (2010), p. 198.
13 Batson et al. (1991), p. 174.
References
Batson (2011). Altruism in Humans. Oxford University Press.
Batson, Batson, Slingsby, Harrell, Peekna, & Todd (1991). Empathic joy and the empathy–altruism hypothesis. Journal of Personality and Social Psychology, 61: 413–426.
Batson (1998). Altruism and prosocial behavior. In Gilbert & Fiske (eds.), The Handbook of Social Psychology, Vol. 2 (pp. 282-316). McGraw-Hill.
Batson, Duncan, Ackerman, Buckley, & Birch (1981). Is empathic emotion a source of altruistic motivation? Journal of Personality and Social Psychology, 40: 290–302.
Batson, Dyck, Brandt, Batson, Powell, McMaster, & Griffitt (1988). Five studies testing two new egoistic alternatives to the empathy–altrusim hypothesis. Journal of Personality and Social Psychology, 55: 52–77.
Dovidio, Allen, & Schroeder (1990). The specificity of empathy-induced helping: Evidence for altruistic motivation. Journal of Personality and Social Psychology, 59: 249–260.
Dovidio, Piliavin, Schroeder, & Penner (2006). The Social Psychology of Prosocial Behavior. Lawrence Erlbaum Associates.
Eisenberg & Miller (1987). Empathy and prosocial behavior. Psychological Bulletin, 101: 91–119.
Glimcher (2010). Foundations of Neuroeconomic Analysis. Oxford University Press.
Schroeder (2004). Three Faces of Desire. Oxford University Press.
Stich, Doris, & Roedder (2010). Altruism. In Doris (ed.), The Moral Psychology Handbook (pp. 147-205). Oxford University Press.
Stotland (1969). Exploratory studies of empathy. In Berkowitz (ed.), Advances in Experimental Social Psychology, Vol. 4 (pp. 271–313). Academic Press.
I actually think a lot of people who advance the selfishness hypothesis are instead confused about selfishness.
On one hand, the word “selfishness” has a bunch of negative connotations and associations. But on the other hand, the revealed preferences theory holds that, in the end, we do what we do because we in some sense want to. Excepting reflexes and autonomic behavior, this is both trivially true and utterly useless. Saying that we do something “because we want to” has no explanatory power. But people who are confused about this take “we do what we want to”, together with “to do something merely because you want to is selfish” together to say that “all behavior is selfish.” This smuggles in all the connotations of selfishness, without explaining anything.
This post does a good job of explaining why our behavior doesn’t match some of the connotative meanings of selfishness, but doesn’t address the common confusion. That’s probably because you wrote it for the Less Wrong audience, who should already have the tools to avoid the common trap.
Good post, and I agree with your conclusion, but I’d try a different tack to get there—one that’s probably the same argument encoded differently.
Instead of talking about “desires for the well-being of others” vs. “self-interested desires”, I’d throw out the whole desire language and use the language of behaviorism here. Seeing others in pain is negatively reinforcing. Helping others is positively reinforcing. Therefore we help others.
Instead of the question “what desire is at the root of your helping behavior?” this suggests the question “why is helping others reinforcing?” to which the answer is probably evolutionary and has nothing to do with the character of the person in question.
If you then ask the altruist what desires motivated their action, they’ll make something up, the same way people usually make up their reasons for stuff.
Also, am I understanding the experiment table right in saying that when subjects felt high empathy for Elaine, they were more willing to stay when escaping was easy than when it was difficult? Any ideas why that might be?
Stylistically, I would organize the article around the cleverness of the experiments. I’d invite the reader to pause and think of how to test the questions at hand, with hints. This has the benefits of stretching people’s minds, making people less prone to arguments from ignorance because it confronts them with the fact they couldn’t think of how to test something and thought it impossible, making them feel good when they can think of how to test it, and it might even lead to someone thinking of a novel and superior way to do so.
Then I’d linger on how details of the experiment disambiguates subjects’ reasons behind actions. This makes the issue an exercise in problem solving rather than receiving a teacher’s password from authority.
It is of course lukeprog’s show, and he’s doing a great job. His way has its advantages.
The numbers are in that direction, but I’m pretty sure the difference is not statistically significant.
I am confused as to how the experiment was set up.
In one case, the subject chose between watching eight sessions of shocks or being shocked in eight sessions, in the other, the subject chose between leaving, in which case the actor would be shocked for eight sessions, or being shocked for eight sessions.
In the latter case, did the subject have the option to watch eight sessions? If not, what reason was given to the subjects, as it makes no sense from their perspective for the actor to be obligated in the experiment wile they are exempt when it is the actor who is squeamish under electricity. If so, then the cases are not symmetrical, right? There would be three options for the “easy” escape condition, but only two for the “difficult” one.
Perhaps this construct of options directed people to switch. For example, they may have felt more guilty about taking the most self-serving of three options than of two.
The span of the numbers in the experiment with Elaine were kind of amazing, and seem to me like they could easily be interpreted to lead to the conclusion that most people are quite amoral, which is a bit distressing. Specifically, the amazing part is the dramatic swing from 18% helping to 91% helping on the basis of discovering that “Elaine’s values and interests are very similar to the subject’s… while in the low-empathy condition, they are very different”.
The effect of the similarity manipulation in the easy escape condition appears to be huge.
One way to reconcile my surprise is to imagine that the value differences the experimenters showed the subjects are really dramatic, like to the extent that “low empathy Elaine” is some kind of Baby-eating monster who is dissimiliar to most people and who many would feel moral righteous in helping to punish. That would explain the results in a way that would be moderately comforting, in the sense that the 18% helping rate might not simply show that only about 18% of the study population is what I would think of as “basically moral”. If low empathy Elaine was a monster then some fraction of the 82% could reasonably have been the sort of people who also tend to enact personally costly punishment to maintain a moral order. If that was coming through in the study as a confound, that would be kind of comforting :-)
On the other hand, if “low empathy Elaine” just likes a different genre of music, has different dietary practices, and has the opposite tendencies from the subject on an introversion/extroversion scale then I find myself kind of horrified by the results of the low empathy condition… if low empathy Elaine is still a basically good person who is “just different” but not bad then it seems to indicate that something like 82% of the study population were assholes who wouldn’t help someone in distress if the distressed person is even a small distance outside their monkeysphere.
Help distinguishing between these interpretations would be appreciated.
The following is from the Batson, et al. (1981) paper.
I’ll email the paper to anyone who PMs me.
Magazines and where she lived produced that difference? Oh my stars. I’m surprised this experiment doesn’t get cited as much as Milgram.
I think this is pretty easily distinguished, though. In Milgram, the subject was in a position to prevent the “victim”’s extreme pain and possible death, at no cost to the subject. In Batson, the subject is in a position to prevent the “victim” from receiving shocks by volunteering to receive those shocks themselves. (Yes, the victim does claim to be unusually averse to shocks, but there’s no real reason for the subject to believe that claim.) I see “help someone by hurting yourself in equal measure” as being a very different ethical proposition from “help someone at no cost to yourself”.
Those details are pretty much exactly the sort I was looking for. Legwork very much appreciated.
Well, the Milgram experiment already tells us that no more than 35% of people are basically moral, and that’s in a case of life or death. In the Batson experiment the stakes are lower.
Before I learned about the Milgram experiment I thought of people as being moral or immoral. I don’t think that way anymore.
Couldn’t an egoist still answer, “Yes, my mood is elevated from seeing/knowing others escape pain, but that doesn’t mean I place (or should place) a terminal value on other people’s pain in addition to the value I place on my mood. In fact I don’t / shouldn’t.” Scope insensitivity seems incompatible with valuing other people’s lives directly, but perfectly compatible with this egoist position. See Boredom vs. Scope Insensitivity where I point out why this position is not easy to dismiss as a normative (ethical) theory. And given widespread scope insensitivity in people’s actual behavior, I don’t see how it can be dismissed as a positive (psychological) theory.
I don’t think scope insensitivity sheds much light on this topic. People’s attitudes toward their own lives exhibits a similar insensitivity. For example, if you convince me that my life could end now, or extend for another 50 years, depending on what I do, I will gladly work very hard and endure many hardships to get the latter result. If you convince me that my life could extend for 500 more years, I will probably not work ten times as hard and endure ten times as much hardship. (Or if I did, I would have to talk myself into it.) And that’s not because I expect life to get boring. And if you convince me that my life could end now or in 5 years, I will probably work much more than 1/10th as hard for the extra 5 as I would for the extra 50.
Motivation doesn’t scale linearly with payoff. That fact does not distinguish selfish from other-regarding motivation.
Luke—your typology of ends reminds me of something I was reading recently by Jonathan Edwards. I know this is not an atheology post, and the Edwards work isn’t particularly empirical, but I thought it might be an interesting antecedent besides.
Aversive-arousal: so it distresses me more to see someone suffer who I imagine to be more like me, and, that distress really lingers once I escape—knowing that I did nothing to help hurts. Someone who I don’t identify with creates a low level of distress that fades much more easily.
Empathy-altruism: not enough details about how the reslts favored empathy-altruism for me to rebut. But it seems like knowing someone like you is suffering hurts, and hearing that they won’t suffer is a relief, whether or not you expect to get credit for saving them. I might feel relieved that they’ll be saved 100% perfectly without my being responsible for any mistakes that cause them to suffer.
I am not sure where all this is supposed to lead. What is necessary is to put real people into brain scanners and expose them to various moral dilemmas and then derive mathematical models of ethical properties, statements, attitudes, and judgments from the data. If you try to figure out the meaning of terms like “altruism” or “selfishness” by contemplation you will end up introducing new and malleable definitions for those terms instead of figuring out what both, our conscious and subconscious selfs, our cultural high-level cognition and evolutionary heritage, mean by those terms. That can only be done by looking directly at the point where recursive justification hits bottom, by collecting data and turning it into mathematical precise definitions.
Doesn’t this entirely depend on how you define “selfish”? It’s not clear to me why you set up these two hypotheses as the difference between egoists and altruists, as one could easily explain the empathy-altruism hypothesis in the language of selfishness.
Indeed, I would take the experiment whose results you quote as telling evidence that human altruism is selfish, powered by the degree to which we recognize ourselves in others. Yvain seems to strike at the heart of the issue much more closely by talking about ‘reinforcement’ rather than ascribing political names to desires.
Could you give the specific results, as in the aversive-arousal reduction experiment?
Amazing article… especially when I consider that I’m working on one right now, that is questioning whether altruism (in the sense people imagine it) exists ;)
So let me play devil’s advocate here and say that I can’t see how the experiments you described imply altruism. Given that we evolved in groups of moderate sizes, of course one would expect that people help someone in distress instead of just fleeing the situation—such dishonorable behavior would severely tarnish one’s reputation, even if the person-to-be-helped was a stranger. The study is just too simplistic and doesn’t factor in important things like social status and tit-for-tat expectations, which surely play a role in “altruistic” decisions as well.
I am quite prepared to believe that the person who’s helping believes he or she does it for altruistic reasons, but I think more realistic is an explanation that includes some kind of concern about social status. I’m not convinced by the study you cited, but since this is a matter that interests me a lot, I’ll definitely check your references and take a look what other studies Batson et al. came up with. As always, thanks for your efforts luke!
Altruism may exist (I think it does), but I’m not sure I fully understand the distinction which is being made here. Does the aversive-arousal hypothesis say that people only feel bad when they see people suffer? (Not a serious suggestion, since then all blind people would be sociopaths.) Is it the immediate sensory perception of suffering that, according to this hypothesis, causes discomfort? The article talks about “escaping the situation,” so is it some aspect of the social position in which a person finds himself, the being-expected-to-help, which is understood to cause the unpleasant sensations? For me, just knowing that another person suffers is enough to upset me—I suppose that doesn’t count as aversive-arousal? Does it fit the definition of altruism that you are using here?
I often feel very intense, arational urges to help when I become aware that other people are in trouble. These feelings do seem to follow from, or in some way depend on, empathy with the needy party, but I’m not sure I would classify them as altruism; they don’t necessarily agree with what I believe to be the right thing to do, and sometimes I wish I could get rid of them more easily.
When I cannot gratify this need-to-help, I often suffer rebound depressions in which my ability to learn and do difficult work is significantly impaired. For example, I recently read about this couple, and found myself losing sleep over their predicament, and feeling the urge to dip into my emergency fund in order to send them more money than I could otherwise afford, even though I knew that I could do much more good by sending the same money to VillageReach and had already made the choice not to do so. Nor do I think this was really a case of buying fuzzies—the satisfaction that my small donation gave me was much smaller in magnitude than my distress at being unable to fix the situation; it was even less than the satisfaction I feel when I send a similar amount to VillageReach. But the article made me empathize with them, and now I was stuck with the empathy.
I could try to train myself not to feel this need-to-help so intensely, but I’m not sure I really want to. I think I like being the kind of person who will drive through the night to help a friend in trouble, or take responsibility for the well-being of an elderly neighbor—I just don’t want to have to do the same for strangers. Because of this, I have come to rely on scope insensitivity as a defense mechanism. I try to avoid empathizing excessively with random needy people, because once Helping Girl Mode is engaged, I have a tough time turning it off, or even wanting to turn it off. When dealing with statistics instead of people, I can allocate my charitable giving according to what seems most effective, rather than to address what I perceive as urgent needs.
(I hope no one interprets this comment as bragging. I’m pretty sure I’m confused somehow, and I know it’s silly to care that much about random internet people.)