Restraint Bias: Why You Overrate Your Power to Resist Temptation

Restraint bias is a mistake about the self made from a safe distance: asked how well you would resist a craving, you answer in a state where the craving is absent, and from there restraint looks easy. Loran Nordgren, Frenk van Harreveld and Joop van der Pligt named the bias in 2009 and showed its consequence: people who overrate their impulse control put themselves in the way of temptation, expecting to withstand it, and then do not.

The forecast fails because impulses are visceral states, hard to imagine when you are not in them. George Loewenstein set out this account in 1996: hunger, craving, fatigue and pain reshape what a person wants and can resist, yet someone in a cool state underestimates how much a hot state will change them. Nordgren and colleagues' first two studies made the point on a university campus: people who had just finished an easy two-minute memory task rated their power to overcome fatigue higher than those who had done a tiring twenty-minute one; people leaving a cafeteria, fed, rated their control over hunger higher than people arriving hungry.

The third study closed the loop from belief to exposure to failure. Fifty-three smokers were told, at random, that a sham test showed their impulse control was high or low. They then watched a film while trying not to smoke, choosing in advance how close to temptation to sit. Those told they had high control chose higher temptation and were about three times as likely to smoke before the film ended.

The fourth study took the effect out of the laboratory. Fifty-five people who had quit smoking for at least three weeks were asked about their impulse control and followed for four months. Those with more inflated beliefs were less careful about avoiding the situations that trigger smoking, and it was that carelessness, not the beliefs themselves, that predicted relapse. The restraint bias is therefore a decision rule, "I can handle it," that moves people toward the thing they are trying to avoid, and it is strongest when the impulse is quiet.

Nordgren, van Harreveld and van der Pligt published "The restraint bias: How the illusion of self-restraint promotes impulsive behavior" in Psychological Science in 2009. The paper reported four studies, each testing a link in one chain: beliefs about impulse control are inflated when the impulse is absent, inflated beliefs lead to more exposure, and exposure leads to more impulsive behavior.

The first study, with 72 students, manipulated mental fatigue. Participants did either an easy two-minute memory task or a strenuous twenty-minute one, then rated their capacity to overcome fatigue and said how they would schedule their coursework. The rested group rated its capacity higher and planned to leave more work to the final week. The second study, with 79 students and staff at a university cafeteria, approached people on their way in, hungry, or on their way out, fed. They ranked seven snacks, rated their control over hunger, and chose one to take home against a small payment for bringing it back uneaten a week later. The satiated group reported more control, chose snacks higher on their own preference list, and were somewhat less likely to return them intact.

The third study is the one the demo card quotes. Fifty-three smokers completed a sham test of impulse control and were told at random that their control was high or low. They then chose a level of temptation at which to watch the film Coffee and Cigarettes without smoking, with higher payment for higher temptation, from a cigarette in another room to an unlit cigarette in the mouth. Smokers given high-control feedback chose higher temptation and were about three times as likely to smoke during the film. The fourth study followed 55 people from a smoking cessation program who had been smoke-free for at least three weeks. Those with more inflated beliefs about their impulse control were less likely to avoid high-risk situations, and it was that avoidance, not the beliefs themselves, that predicted whether they were smoking again four months later.

The paper has limits the authors state: the samples are small, the laboratory studies used students, the cafeteria study compared people entering with people leaving rather than assigning them at random, so other factors cannot be ruled out, and the smoking studies rest on self-reported relapse. Two of the authors' own figures are marginal at p = .06: the threefold smoking difference in the third study and the snack return rates in the second. The exposure difference beneath the smoking result is not. The mechanism sits inside Loewenstein's well-supported account of hot and cold states. The demo asks for your forecast and your choice of exposure rather than reproducing the sham feedback.

The pattern appears wherever a decision about future temptation is made while the temptation is absent. The rows are illustrations, not findings.

Because the bias is in the forecast, the counter is to stop relying on the forecast.

• Treat "I can handle it" as a warning. The sentence is produced by a calm state about a state that will not be calm.
• Manage exposure, not resistance. The people in Nordgren's studies who did best were those who kept their distance from temptation, not those who resisted it up close.
• Decide in the hot state's favor. When planning, assume the version of you who meets the temptation will be hungrier, more tired and more tempted than you feel now, and plan for that person.
• Use base rates rather than self-ratings. A measured rate carries no flattery. When your estimate is far above what most people manage, that gap is information.
• Rehearse the moment. Implementation intentions, specific if-then plans for the situation in which the impulse will arrive, move the decision out of the moment, where restraint is weakest, into the planning, where it is strong.

Restraint bias belongs to a family of errors about one's own future states and abilities, and its neighbors differ in what is being misjudged.

• It is not the hot-cold empathy gap, which is the general difficulty of imagining a visceral state you are not in. The empathy gap is the mechanism; restraint bias is one of its consequences, the specific overestimate of self-control and the exposure to temptation that follows.
• It is not the overconfidence effect, the broad tendency for confidence to exceed accuracy across kinds of judgment. Restraint bias is about one domain, impulse control, and it has a specific behavioral cost, approaching temptation, that general overconfidence does not imply.
• It is not the planning fallacy, which is underestimating how long a task will take. The two can combine, but the planning fallacy concerns time and the restraint bias concerns resistance.