Most people are better judges of other people's moral character than they are of their own.

Experiments conducted at Cornell University and reported in the Journal of Personality and Social Psychology (Vol. 79, No. 6, pp. 861-875) found many people making an error in self-assessment. Participants in the survey consistently tended to overestimate their own generosity, but actually were quite accurate when predicting the generous behavior of others, according to Nicholas Epley, a graduate student of psychology, and David Dunning, a professor of psychology in the College of Arts and Sciences at Cornell.

"Most people are overly optimistic about themselves," Dunning says. "As it turns out, we don't know as much about our own moral nature as we know about others'."

It has long been known – both in psychological research and in everyday life – that people commonly feel "holier than thou," believing that they are more moral, kind and altruistic than the typical person, the psychology professor says.

"We knew something had to be wrong when the average person thinks he or she's a better person than the average person, when the majority of Americans consider themselves to be members of an elite moral minority," says Epley, who conducted a series of revealing experiments about subjects' perception versus the reality of their moral behavior. "We wanted to know whether people feel holier than thou because they underestimate others' moral goodness, or because they overestimate their own."

Each experiment asked college students at Cornell to predict how they would behave when faced with one of several moral dilemmas, and to make the same predictions for their peers. Again and again, people predicted that they would be more generous and kind than others. Yet when the time came to put their money where their mouths were, most kept their wallet in their pockets.

This inaccuracy in self-prediction was matched by an impressive ability in predicting how other people would respond to the same morality moments. In psychological terms, the experimental subjects were successfully anticipating the base rate of moral behavior and accurately predicting how often others, in general, would be self-sacrificing. The problem is that people don't use their wisdom about others to successfully predict their own actions, Dunning notes. "Even when we know that other people will be selfish, we think we're special, that the rules don't apply to us," Dunning explains.

For example, in one study students said they would donate roughly half of their $5 fee for participating in the experiment to charity, if given a chance, but that other students on average would donate only about $1.80. When students were actually given a chance to donate their fee, the average donation was only $1.53.

In another particularly telling experiment, students were given the choice of performing an onerous, time-consuming task themselves or assigning the work to someone else. The "someone" in some cases was another college student and, in others, a 10-year-old girl who presumably would have great difficulty with the task.

Many students predicted they would take on the onerous task themselves, particularly when the other person was a little girl. However, most students facing an actual decision chose the easier job for themselves and were just as likely to do so whether they were assigning the difficult task to another college student or to the youngster.

"The only thing they were responding to was self-interest," Epley says. "Whether they were dealing with another college student or a little girl made no difference."

The "holier than thou" experiment was supported by a grant from the U.S. National Institute of Mental Health for studies of human capacity for self-insight.

A total of $19.89 was actually donated to charities from the experiment asking subjects to predict how much of their participation fee they would give to the needy. Had the subjects followed through with their good intentions, the charities would have received $31.72.

Brief descriptions of experiments

• Daffodil Days: How many daffodils would you buy to help the American Cancer Society? What percentage of your peers will buy flowers for charity and how many, on average, will each buy? These questions were asked of some 250 students in a psychology course five weeks before the annual Daffodil Days on campus.

  More than eight out of 10 students polled said they would buy flowers (on average, two flowers each) but they expected that their peers would not be as generous, predicting 56 percent would buy (an average of 1.5 flowers apiece). Three days after the daffodil sale, the same students were asked: How many charity flowers did you actually buy? Only 43 percent had helped the cancer society, buying, on average, 1.2 flowers each. The students had greatly overestimated their own kindness, but they were reasonably good at predicting the kindness of others.
   

• A Saint's Dilemma: Playing the "prisoner's dilemma" game, 97 Cornell undergraduates were told the rules and asked to predict how each of them and how a fellow (but unseen) "prisoner" would respond. In the game, undergraduates could choose to cooperate with another person, in which case both would receive equal amounts of money, or choose to "defect," giving them a chance to gain a lot of money, while the other person would receive nothing.

  Despite the potential big payoff for a selfish decision, 84 percent of the students predicted they would cooperate so that they and their partners could receive equal amounts of money. The same students expected only 64 percent of their peers would be as cooperative. Again, they were wrong about themselves and nearly right about their peers. Sixty-one percent of an equivalent group actually cooperated when playing the game for real.
   

• Can You Spare a Dime? Thirty-eight high school and college students attending Cornell's summer session were paid five $1 bills each for participating in a fictitious psychology experiment, but the psychologists were more interested in the students' generosity with their payments. Students were asked to predict how much of their payment they and their peers would donate to charity. An equivalent group of subjects were given a real opportunity to donate their money to charity, anonymously, by placing money in sealed envelopes.

  If all the well-intentioned experimental subjects had lived up to their plans, charities would have received $2.44, on average, from each student. If the predictions of peer generosity had come true, each student would have donated an average of $1.83. When the envelopes were opened, they contained, on average, $1.53 per subject.
   

• Can You Spare the Time? In a more complicated study designed to see how well people anticipate the impact of self-interest and moral concerns on their behavior versus that of others, a total of 61 Cornell undergraduates were offered several options: They could selflessly accept a time-consuming task themselves and assign a less burdensome job to another person, or they could selfishly choose the short task for themselves. To examine how much self-interest influenced their choices, the onerous task in one condition was only slightly longer than the alternative, but in the other condition it was substantially longer. To make moral concerns either more or less salient, participants in one condition believed the other person was a random college student and in another condition that the person was a 10-year-old.

  Once again, the participants erroneously thought they would behave more admirably than they actually did, and again were reasonably close when predicting the behavior of others. Participants predicted that they would be less influenced by self-interest and more by moral concerns than others. Indeed, for self-predictions, subjects stated that they would be influenced only by moral concerns (for example, if the other person was a college student or a little girl). In fact, when given an actual choice, only self-interest mattered. Only the length of the onerous task made any difference in people's actual choice.