The Heat

In the cool of the morning, the evening’s hunger seems easy to accommodate. We can name it, put it in the diary, and provide what appears to be enough. By evening, the provision looks ungenerous, and the person who made it seems to have understood very little. The familiar divided self has acquired another difficulty: a state can be foreseeable without its force being adequately imagined. Loewenstein’s account of visceral influences gives that difficulty a place in decision theory. Hunger, thirst, pain and emotion can alter what goods and actions are worth to us, while a person outside the state may underestimate its influence. The morning’s mistake can therefore begin before any contest of wills. It begins in the description of the person for whom the morning is making arrangements.¹

The distance can be crossed badly in either direction. A person who is hungry now may choose food for a later, satiated self as though some of today’s appetite will remain; a person who has eaten may make the reverse mistake about a future hungry self. This is the family resemblance between the hot–cold empathy gap and projection bias. The former concerns difficulty appreciating another visceral state, including one’s own at another time. Projection bias more broadly carries the current valuation into a forecast of a different state; it can concern the force of hunger, the fading of novelty, or an acquired taste that has yet to grow. The person may understand that change is coming and still make too little allowance for it. Knowing the direction of a journey gives no assurance that one has understood its distance.²

A small invented example isolates the mistake. Suppose an extra portion, relative to going without it, would yield eight units of benefit when hungry and two when satiated; its cost is four units in either state. All consequences occur at the same future meal, and the forecaster knows that she will be hungry then. Her correct net benefit is therefore four. Now suppose that, while satiated, she gives equal weight to the benefit in that future state and the benefit in her current state: half of eight plus half of two makes five. After the cost, she forecasts only one. At a price of six, the same error changes her decision: the actual net benefit would be two, the forecast minus one, and she declines. No reward has become immediate. No uncertainty about the time of the meal has intervened. A misdescription of future enjoyment is sufficient to lose a choice she would value.³

Present bias and projection can consequently produce similar broken plans by different routes. In the account developed in The Present and The Self, the weight attached to a dated consequence can change as it becomes immediate. In the projection model, even ordinary exponential discounting can accompany an inaccurate forecast of what the consequence will feel like. The distinctions matter when we try to help the planner. A person who accurately predicts a later preference reversal may seek a commitment against it. A person who wrongly expects tomorrow to resemble today may see no need for a commitment, or may bind the future self to a plan founded on a false forecast. A broken resolution by itself identifies neither mechanism. New information, changed circumstances and an honestly revised judgment can also separate a plan from what follows it.⁴

The mistake reaches beyond tonight’s dinner when a passing enthusiasm buys something that will last. In the authors’ durable-goods model, daily enjoyment fluctuates independently around a stable average; the purchase cannot be enjoyed on the buying day. An unusually enthusiastic day consequently supplies no information that later days will be unusually enjoyable, yet a projector treats some of that enthusiasm as enduring. A subdued day can generate the opposite error. With only one chance to buy, the direction of the mistake is ambiguous. With repeated chances and an irreversible purchase, however, declining today leaves another opportunity open, while buying on one sufficiently enthusiastic day settles the matter. That asymmetry can turn fluctuating mispredictions into excess purchasing. Its force depends on the assumptions: persistent tastes, resale, returns and genuine information acquired in the shop can change what today’s feeling should imply for the future.⁵

A better forecast needs evidence from the state it is supposed to represent. For a recurring choice, that may mean keeping a record of what was wanted and experienced when the relevant state actually arrived, then consulting it before making another commitment. Merely waiting changes the forecaster’s state; it does not establish which state best predicts the consequences. A calm person booking a demanding journey may underprovide for fatigue just as an excited buyer may overprovide for an enthusiasm that will fade. Nor does a previous reversal prove that every later plan deserves suspicion. The useful comparison holds the object, timing and circumstances as steady as possible and asks whether today’s condition is repeatedly distorting the forecast. These are ways of testing the error described here, with room for forecasts that improve and preferences that truly change.

The cold self has no automatic title to govern the warm one. Hunger can make food more valuable for an excellent reason, and a calm refusal to acknowledge that need can be the error that a later choice corrects. Where a choice also threatens enduring purposes, assessing it requires those purposes and consequences to be stated; the word hot cannot do the moral work. The morning’s duty, if it undertakes to plan the evening, is to represent the evening with enough accuracy to deserve that authority. We fail when the feeling available to imagination quietly becomes the measure of a feeling yet to come. To choose well across time is partly to learn a form of hospitality: to leave room in the plan for the person who will have to live it.