Empathy Gap: Why You Cannot Feel a State You Are Not In

A visceral state does not sit beside your preferences; it rewrites them while it lasts. Hunger makes food the most interesting thing in the room, fear makes an audience feel like a threat, and craving makes a rule you set an hour ago sound negotiable. When the state passes, the preferences it produced go with it, and what is left is a memory of the facts without the force. George Loewenstein's term for the resulting mistake is a hot-cold empathy gap: a cold self predicting a hot self, or a hot self predicting a cold one. The phrase "empathy gap" is also used loosely to mean a shortage of concern between groups; this page is about the prediction failure across states, which is the version the research measures.

The gap runs in both directions and produces opposite errors. From cold to hot, people underestimate the pull of the state: they commit to a 6 a.m. run, agree to speak, plan a frugal week, and are surprised by themselves when the moment arrives. From hot to cold, people overestimate how long the state will last, which is why shopping while hungry loads the basket for a hunger that will be over by the time the food is cooked.

The same gap operates between people. Someone who is not in pain, not afraid and not craving anything tends to read another person's lapse as a character trait rather than as what a state does, which is one reason Loewenstein's 2005 review argues that clinicians can undertreat pain and that healthy people underweight risks they have never felt. Van Boven and Loewenstein showed in 2003 that the reverse also happens: people project the state they are in. When participants described how hikers lost without food or water felt, 30 percent of those who had just exercised mentioned only thirst, against 9 percent of those who had not yet exercised.

One limit matters for reading the demo above. Estimating what other people did in a published experiment is a cold judgment about a hot situation, which is the interpersonal half of the gap; it is not a measurement of your own behavior in a hot state.

The experiment the demo uses is the second of two reported by Leaf Van Boven, George Loewenstein and David Dunning in Organizational Behavior and Human Decision Processes in 2005, under the title "The illusion of courage in social predictions." Students in a large class were offered 5 dollars to dance alone to a song in front of everyone. Some faced the offer for real; others answered the same question hypothetically. Of the students who faced the real choice, 8.33 percent agreed to dance. Of the students answering hypothetically, 31 percent predicted that they themselves would dance and 35.66 percent predicted that other students would. Asked for the smallest payment they would accept, the students facing the real performance named 52.88 dollars, a geometric mean computed from logged amounts, while those answering hypothetically named 21.07 dollars for themselves on the same measure.

The paper's first experiment used the same design with a different task, miming a scene in front of the class for 5 dollars. There, 15.12 percent of the students facing a real performance agreed, while those facing a hypothetical performance predicted 31.03 percent for themselves and 43.33 percent for others. In both experiments the people in the cold condition were not merely wrong about strangers; they were wrong about themselves in the same direction, and their self-predictions statistically accounted for their predictions about others.

The hunger case comes from a separate experiment. Read and van Leeuwen, publishing in the same journal in 1998, asked office workers to choose a snack to be delivered a week later, approaching some of them late in the afternoon when they were hungry and others just after lunch when they were not. The hungry choosers picked more unhealthy snacks for their future selves, and when the week was up and the choice could be remade on the spot, far more unhealthy snacks were taken than had been ordered in advance. This page reports the direction of both effects; the overall choice shares are on the projection bias page.

The thirst experiment is Van Boven and Loewenstein's, in Personality and Social Psychology Bulletin in 2003. Forty-seven people at a campus gym answered a questionnaire about three hikers lost without food or water, either immediately before or immediately after at least 20 minutes of vigorous exercise. Those who had just exercised rated themselves thirstier and were more likely to say thirst, rather than hunger, would be the worse problem for the hikers.

The gap appears wherever a decision is made in one state and carried out in another. The rows below are illustrations, not findings from the studies cited here.

Nothing here makes a cold self able to feel a hot state. The useful moves either move the decision or bind it.

• Decide in the state, or at least sample it. Choose running shoes after a run and groceries after a meal.
• Bind the decision in advance. A standing order, a pre-booked class and a list in your pocket work because they do not require the hot self to agree.
• Write the if-then half. "If I am still at my desk at seven and starving, I will eat the thing I already bought" is a plan the hot self can follow. A resolution without that half is only the cold half.
• Ask someone who is in it. For judgments about other people, the person in the state is a better source than your own simulation.
• Expect the overshoot too. A decision taken at the peak of a state tends to be built for a state that will not last the week, so put a delay between the feeling and anything irreversible.
• Do not read a lapse as a diagnosis. The research describes a prediction error almost everyone makes.

Several biases concern the gap between a prediction about a feeling and the feeling itself.

• It is not projection bias, which is assuming your future tastes will match your current ones in general, not only across visceral states. The empathy gap is the version driven by a bodily state, and it is the mechanism most often cited for the broader pattern.
• It is not the impact bias, which is overestimating how long and how intensely a future event will affect your mood. That is a forecast about the size and duration of a feeling; the empathy gap is about failing to simulate a state well enough to predict what you would choose while in it.
• It is not the restraint bias, which is overestimating your own ability to resist a temptation and so walking into it. The restraint bias is one consequence of a cold-to-hot gap rather than a separate account of it.