Want to know how much you'd enjoy an experience? You're better off asking someone who has been through it, even if they're a complete stranger, than to find out information for yourself. This advice comes from Daniel Gilbert from Harvard University, who espoused it in his superb book Stumbling on Happiness. Now, he has found new support for the idea by studying speed-daters and people receiving feedback from their peers.
In the first study, he found that female students were better able to predict how much they would enjoy a speed-date if they listened to the experiences of strangers than if they make their own assessments based on available information. Likewise, the second study found that people more accurately foresaw their reactions to criticism when they knew how someone else had reacted than when they had the information for themselves.
This interesting result masks a second one of equal importance - people don't believe that this works. In both cases, most volunteers felt that they would be better off relying on their own counsel than to lean on the opinions of others. In some ways, that's not surprising - it is, after all, a fairly counter-intuitive idea. But it also comes on the back of a huge amount of evidence that we humans are really very bad at predicting what will make us happy.
Time and again, psychological studies have found that we overestimate how happy we will be after winning a prize, starting a new relationship or taking revenge against those who have wronged us. We also overrate our disappointment at bad test results, disability or failure to progress at work. Try as we might, we consistently fail to forecast our own emotional reactions, and we even fail to accurately remember our past experiences to be used as guides.
Gilbert says that the main reason for this is an inability to accurately imagine future events. We can close our eyes and try to picture ourselves in the future but we focus on the wrong things, we predict that our emotions will last longer than they do, and so on. Some scientists have tried to improve things by training people to mentally time-travel with more accuracy but these attempts have been largely unsuccessful.
Gilbert advocates a different ethos - use other people as living simulators. In most cases, you can find someone who has been through the same experience, and they can act as a mental surrogate. Having actually lived through the experience you're trying to picture yourself going through, they are resistant to many of the mistakes we naturally make in our predictions.
In his first experiment, Gilbert sent 33 female undergraduates on short, five-minute speed-dates with one of eight men. Each woman was asked to predict how much they would enjoy the date based on either a personal profile and photo provided by the man himself, or the report of another stranger on how she had enjoyed her date. After they made their forecast, Gilbert gave the women the second set of information so that all of them went into the date with the same info.
When they later rated their experience, those who had relied on the stranger's information more accurately predicted their enjoyment of the date than those who used the profile and photo. In fact, relying on other women's experiences halved the error between prediction and reality.
The problem was that very few of the women believed this; 75% of them believed that getting information for themselves would have allowed them to better predict the outcome of the date they were sent on, and 84% thought that doing so would better allow them to forecast a future date.
Of course, it's possible that the men may have lied on their profiles, but Gilbert thinks it unlikely. If they exaggerated their qualities, the women would have overestimated how much they would enjoy the date. In fact, they actually underestimated, and ended up having a better time than they imagined.
In his second study, Gilbert told 88 students to write a story that someone would use to class them as one of three personality types - a positive A person, a neutral B person or a negative C person. Half of them were given complete descriptions of the three types and asked to predict how they would feel if they were labelled in each category. The other half were only told how another student (from a separate group of 25) felt when they were classified as type C. They too made their predictions, and were then shown descriptions of all three personality types.
Unbeknownst to them, all the volunteers were placed in the negative C group, described as people who "sacrifice their beliefs because they seek contentment rather than challenge" or whose "long-term relationships end usually because the person's partner has found a more suitable alternative."
Again, basing their calls on a stranger's reactions proved to be a better strategy than using the descriptions they read for themselves. It reduced the difference between imagined and real emotions by 63%. And again, people don't believe this. When Gilbert told a separate group of 63 students about the experiment, he found that by a wide margin, they believed that the complete descriptions of the personality types would be more useful than the responses of other recruits.
If anything, Gilbert's experiments have underestimated the value of listening to your neighbours. In both cases, recruits relied on the emotional reactions on complete strangers; in real life, we would most likely rely on the opinions of friends, who we presumably have more in common with. And the two scenarios that Gilbert used - dating and feedback from peers - are by no means unusual.
There are clearly some major disadvantages in relying on the experiences of others. The most obvious one is that people are different from one another and one person's pleasure may be another's disgust. But the point is not that a peer's prediction will be perfect, but that it's better than what you could do yourself. Gilbert notes that people's emotional reactions aren't as different as we may think. Many of them are based on ancient bodily processes that are consistent across cultures. And we're especially likely to share emotional reactions with those people most likely to share theirs with us - our friends and family. As he says:
"[Firstly, we prefer] warm to cold, satiety to hunger, friends to enemies, winning to losing, and so on. An alien who knew all the likes and dislikes of a single human being would know a great deal about the entire species. Second, people tend to marry, befriend, work with, and live near those who share their preferences and personality traits... In short, there is little disagreement among people about the sources of pleasure and pain, and even less disagreement among neighbours."
Gilbert's experiments are compelling and they are just the tip of the iceberg in similar research on the psychology of happiness (see chapter 11 of his book for more). But more pessimistically, they also suggest that people are unlikely to take full advantage of the "power of surrogation", of using other people's experiences to simulate your own. Instead, we tend to hold to the mistaken strength of our own predictions, misguided though they may be. As Gilbert says:
"When we want to know our emotional futures, it is difficult to believe that a neighbour's experience can provide greater insight than our own best guess."
More on happiness: Money can buy happiness... if you spend it on other people
Photo by Laughlin, found on Flickr
Reference: Science 10.1126/science.1166632
- Log in to post comments
It strikes me that these results may be demonstrating social persuasion rather than prediction.
I.e. a peer telling me to expect X may more effectively prime me to experience X than would my own expectation of X based on information.
In order to make this distinction, it would seem wise to test a third party observer: Subject: Dater predicts her satisfaction from speed dates based on informational documents, then goes on speed dates and report her satisfaction. Subject: Observer predicts Dater's speed date satisfaction based on reported experience by other Dater subjects or based on Dater's self-predicted expectation.
This way the Daters have more uniform predictive experiences and are not exposed to unbalanced social priming of their expectations, yet the data exists to compare information-based expectations, peer-based expectations, and actual outcome.
It's a bit confusing, because it involves a lot of predicting based on other people's predictions, but that's the point of the original.
Ah but remember that everyone entering the dates did so with both sets of info - the profiles and the ratings from other dates. The only difference was which of the two bits of information were given before they made their prediction. If the peer opinions persuaded them in a certain direction, you'd expect both groups to be swayed since both had access to that information.
I thought a similar thing to what Spaulding thought.
Since everyone entered the dates having read the peer evaluation, it's still possible that the peer evaluation influenced their future perception of the date more so than their own evaluation based on the profile and pic.
I think, from personal experience, although there may have been a study, that if you form your own opinion and voice it before hearing others' opinions, like in this study, your original opinion is less susceptible to influence by others' opinions, so social persuasion is reduced, but perhaps not completely gone, and it could have created the significant effect described.
I thought a similar thing to what Spaulding thought.
Since everyone entered the dates having read the peer evaluation, it's still possible that the peer evaluation influenced their future perception of the date more so than their own evaluation based on the profile and pic.
I think, from personal experience, although there may have been a study, that if you form your own opinion and voice it before hearing others' opinions, like in this study, your original opinion is less susceptible to influence by others' opinions, so social persuasion is reduced, but perhaps not completely gone, and it could have created the significant effect described.
The only thing I'd worry about with such a study is the following:
http://www.xkcd.com/185/
I maintain that there's a difference between merely offering info vs. saying "evaluate based on this info." Plus there's the issue of sequence - initial information likely has more impact than subsequent information, as Jake suggests. I think one would need to control for these things in order to distinguish a study of prediction from a study of influence.
Spaulding and Jake - it's a good argument but just trawling through the Supplementary info, I think I've found the result that falsifies it.
In the speed-dating experiment, there was no difference between how much the women in the two groups enjoyed their dates. It was only their predicted ratings, rather than their actual ones, that were significantly different. If your theory is right, and they were simply being primed by the opinions of others, you would expect an effect on their actual experiences, no?
The experiments are designed so that the experience provides much more information than the summary given to the subjects pre-experience does. So the only thing that may be surprising is that the subjects believe that they can predict their enjoyment of the experience from information better than from the ratings of other people.
Whether these experiments are interesting thus depends entirely on how the subjects were asked the question. If they were asked, before being given information or being told what that information would be, whether they could predict their response to an experience better by making their own judgement based on information, or from the responses of others, then the result is not interesting. The subjects did not know that they would be given only a trivial amount of information relative to those who had the experience. The result is only interesting if the subjects were given the information first, and then asked whether they could predict their response better from that information than from someone else's experience. This post doesn't say which of these things happened.
Of course my next step was to look up the original article, to see how the question was phrased. But I can't! You didn't mention what the original article was, or where or when it was published, Ed!
Thanks for the follow-up, Ed!
What, besides the reference at the end? ;-) Citation wasn't available at time of writing, but the DOI is right there.
Here's a link:
http://www.sciencemag.org/cgi/content/abstract/323/5921/1617
If anything, it's the opposite. In the first study, the experiences of the another woman amounted to just three of how much she liked the man, how much she wanted to befriend the man, and how much she wanted to have a romantic relationship with the man. These were made by marking three continuous 100 mm scales whose endpoints were labeled not at all and very much. Compare that to the profiles filled out by the men themselves.
The difference was even starker for the second experiment. For each personality type, there was only one rating, again on a linear scale from very bad to very good. In contrast, the full descriptions (say for personality type C) said: "These people are fairly competent and well-adjusted but have few qualities that distinguish them from others. They are generally well-liked, partly because they do not pose a threat to the competencies of others. These people tend to succeed in a wide variety of careers but, because of their ability to get along well with others, mostly excel in jobs requiring them to be part of a team rather than take on individual responsibilities. In their personal relationships, they will sacrifice their beliefs because they seek contentment rather than challenge or excitement. Their romantic relationships are fairly successful as long as they are with a person of the same type. Divorce rates do not differ substantially from the national average in this group and when long-term relationships end it is usually because the person's partner has found a more suitable alternative. These people tend to have a realistic picture of both their talents and their limitations and thus tend to structure their tasks quite appropriately."
Actually they were asked to predict which type of information would allow them to make the most accurate response *after* the date.
Sorry, I hadn't used DOIs before, and did not perceive it as a reference.
And, equally importantly, after being shown an example of the "simulationist" information. To quote the paper:
That's the crucial piece of information I wanted.