Illusion of External Agency: When Your Contentment Looks Like a Plan

People are better than they realize at being satisfied with what they get. A long tradition in social psychology shows that once an outcome is fixed, the mind reworks it: the unchosen option loses its appeal, the chosen one gains, bad news is reframed, and attention moves to whatever is good about the situation. Gilbert and his colleagues called the collection of these processes the psychological immune system, and showed in 1998 that people fail to anticipate it, predicting that bad outcomes will hurt longer and more than they do. They called that failure immune neglect.

The illusion of external agency is what immune neglect produces after the fact. If you do not know that your own mind improved an outcome, the improvement needs another explanation, and the natural one is that the outcome was good to begin with. A good outcome that someone else arranged then invites a further inference. For an agent to have produced the best outcome for you on purpose, the agent must have known which outcome was best, wanted you to have it, and been able to bring it about: insight, benevolence and influence. Gilbert and his colleagues argued that an agent known to have two of these is credited with the third whenever an outcome feels optimal, and that the feeling is often self-made.

The three experiments in the 2000 paper each tested one of the three inferences. Where an experimenter was known to be benevolent and insightful, satisfied participants concluded the experimenter had influenced their choice. Where a device was known to be benevolent and influential, satisfied listeners concluded it had insight into their taste. Where a giver was known to be influential and insightful, satisfied recipients concluded she was benevolent. In every case the satisfaction had been manipulated by the experimenters, not by the agent, by controlling whether participants learned their outcome before or after they had formed a view of the alternatives. Learning the outcome first makes it easy to make up one's mind in its favor; learning it afterward requires changing a mind already made up, which is harder.

The authors drew the wide conclusion that belief in an omniscient, omnipotent and benevolent agent, such as God, may draw part of its force from this mechanism, since a person whose outcomes keep turning out well may reasonably look for the cause outside themselves. They were careful to say that experiments cannot settle a theological question, and that the mechanism is a general one: it operates whenever a system, an institution, a stroke of fate or another person stands where the credit for self-generated satisfaction can land.

Daniel Gilbert, Ryan Brown, Elizabeth Pinel and Timothy Wilson published "The illusion of external agency" in the Journal of Personality and Social Psychology in 2000. All three experiments ran at the University of Texas at Austin, and in each the experimenters manipulated how easily participants could come to like their outcome by controlling whether they learned which outcome was theirs before or after forming impressions of the alternatives.

In the first experiment, 77 women chose a teammate for a game by picking one of four folders said to contain autobiographies of four other participants; all four folders held the same autobiography. Half had read and rated all four autobiographies beforehand, and half had not. Participants were then told that a subliminal message during an earlier task had been designed to steer them toward the most compatible teammate. After omitting ten participants for suspicion or procedural error, the 34 who had not rated beforehand liked their teammate more than the 33 who had, 6.03 against 5.70 on a 7-point scale, and rated the subliminal message as more likely to have influenced their choice, 4.00 against 3.18.

In the second experiment, 67 students were told that a prototype in-car system called a SmartRadio had selected one of two songs for them to hear three times. All received the less pleasant song. Half learned this before hearing excerpts of both songs and half afterward. With 61 in the final data set, those who had learned first liked the chosen song as much as the other, while those who learned afterward preferred the other song. Asked which SmartRadio should choose their next song, 13 of 30 who had learned first chose the same system again, against 6 of 31 who had learned afterward; 15 of the 30 said they would buy it, against 5 of the 31.

In the third experiment, 49 women took the role of recipient or observer in a gift exchange with a supposed third participant who had read their questionnaire answers. The gift list held a stuffed animal and a certificate for a free tank of fuel; the giver was said to have chosen the stuffed animal, which pretesting had found to be the less desirable item. Half learned of the choice before rating the two items and half afterward. Recipients who learned first preferred the stuffed animal to the fuel, 5.92 against 4.46, and rated the giver as more considerate than recipients who learned afterward, 6.58 against 5.68. Observers, who were not getting the gift, showed neither difference. The word considerate was used because pretesting suggested many participants did not know the word benevolent.

The illusion appears wherever an outcome you adjusted to was delivered by something that could be thanked. The rows are illustrations, not findings from the cited studies.

The illusion lives in a missing step, the recognition that satisfaction can be self-made, so the counter is to put that step back before crediting anyone.

• Ask the counterfactual. Would you have come to like the other option too? If the honest answer is probably, your liking is evidence about you and not about whoever chose.
• Judge the agent by its process, not by your feelings. A recommendation, a decision or a choice made on your behalf should be assessed by what it was based on at the time. Your contentment afterward is a different fact.
• Remember immune neglect. People reliably underestimate how well they will adjust to outcomes they did not want. Expect to adjust, and discount the adjustment when handing out credit.
• Look for the observers. In the 2000 experiments, people watching the exchange showed no illusion. When unsure, ask someone who did not receive the outcome what they make of the agent.
• Keep the gratitude, lose the inference. Being pleased with what you got is good for you; concluding that the giver, the system or fate must have known is the part that misleads, and it can be set aside without spoiling the pleasure.

The illusion of external agency is a mistake about where satisfaction came from. Its neighbors are other ways of reworking an outcome after the fact.

• It is not choice-supportive bias, in which a choice you made yourself is remembered as better than it was. That bias flatters your own decision; the illusion of external agency hands the credit to someone else.
• It is not the illusion of control, the belief that your own actions influence outcomes that are actually chance. The illusion of control overstates your agency; the illusion of external agency overstates somebody else's, and underrates the part your own mind played.
• It is not impact bias, the tendency to overestimate how strongly and how long an event will affect you. Impact bias and immune neglect are forecasting errors made before the event; the illusion of external agency is the explanation reached for afterward, when the forecast turned out wrong and nobody noticed why.