We value the present far more than the future, and we do it inconsistently, so the patient plan we make today is the one our future self happily breaks. The reliable fix is rarely more willpower; it is design.
Start with the basic fact: a reward in the future is worth less to us than the same reward now, and we mark it down the longer we have to wait. That is discounting, and there is nothing irrational about some of it, since the future is uncertain and money now has uses. The textbook model (Samuelson, 1937) assumes we apply one steady exponential rate, the same percentage markdown per unit of delay, which would at least keep our choices consistent over time. Samuelson himself doubted it described real people, and he was right.
What people actually do is discount hyperbolically: very steeply across the near term and much more gently across the far term (Ainslie, 1975). That shape has a strange consequence called dynamic inconsistency, or preference reversal. Picture two offers. Most people, asked to choose between £100 today and £110 next week, take the £100 now. Asked to choose between £100 in a year and £110 in a year and a week, the same people happily wait the extra week for the extra ten pounds. It is the identical one-week wait in both cases, yet the answer flips, purely because one version puts a reward within reach today. Laibson (1997) captured this with the quasi-hyperbolic model: a sharp present-bias discount (a factor called beta, less than one) applied to anything that is not immediate, on top of ordinary exponential decay between future dates. The signature is a cliff between now and soon, then a gentle slope.
This present bias is the engine behind a whole family of self-control failures (O'Donoghue & Rabin, 1999). When an activity has an immediate cost and a delayed payoff (exercise, saving, filing the tax return), we procrastinate. When it has an immediate reward and a delayed cost (the extra drink, the impulse buy), we overdo it. The authors add a crucial wrinkle: some people are naive, blind to their own future weakness, while others are sophisticated, aware that their future self will misbehave. Sophistication is what makes pre-commitment possible, locking a choice in advance so the tempted future self cannot undo it. That is why commitment devices and automatic defaults beat willpower: they settle the decision while you are still the patient version of yourself (Laibson, 1997).
The most famous strand here is delay of gratification, the marshmallow studies. Mischel, Shoda and Rodríguez (1989) found that four-year-olds who could wait longer for a bigger treat grew into more academically and socially competent adolescents, and identified the mental tricks (looking away, distracting themselves) that helped them hold out. The lab phenomenon, that young children vary in this and that attention strategies help, is solid. The grander claim built on top of it is where the trouble starts.
The biggest recent correction is actually to that very marshmallow story. A larger conceptual replication (Watts et al., 2018) found the link between early waiting and later achievement was only about half the size originally reported, and shrank by roughly two-thirds once the researchers controlled for the child's family background, early cognitive ability, and home environment. Most of the behavioural predictions did not hold up at all. The honest reading is that delay of gratification is real, but it works more as a marker of a child's circumstances, the stability and resources around them, than as a fixed willpower trait that stamps their future. The popular "one biscuit now decides your life" version does not survive.
There is also a live argument about whether present bias is even a single, stable thing. Hyperbolic curves fit the data, but fitting a curve is not the same as finding a mechanism, and measured discount rates swing wildly depending on how you ask, on the size of the reward, and on whether it is a gain or a loss. Using a cleverer method called 'convex time budgets', Andreoni and Sprenger (2012) measured much lower discount rates and found no dynamic inconsistency at all, arguing that a lot of what looks like present bias is really the curvature of value plus ordinary uncertainty about the future. And the brain story is unsettled too: an early account proposed two competing systems, an impulsive one and a patient one, while later work found a single value signal that already discounts hyperbolically (Kable & Glimcher, 2007). Why does this matter? Because the two stories point to different fixes. If present bias is a real impulsive streak, you fight it with commitment devices and defaults. If it is mostly doubt that the future reward will actually arrive, the stronger move is to make that reward more certain and credible, and to be slower to call people irrational for not waiting on a promise the world might not keep. So the real question underneath is whether waiting less is a mistake to be corrected or a reasonable answer to a world that does not always pay out.
Most of the evidence comes from small, hypothetical money choices made in a lab, and the numbers vary enormously with the method, so any single "discount rate" should be treated with suspicion. Samples skew Western and educated. And the central question, whether present bias is a stable trait you carry or a state that shifts with context, is still open.
Is present bias one real parameter or a side effect of value curvature and uncertainty? What is actually happening in the brain when the near reward wins? And how durable are the fixes: commitment devices clearly help in the short run, but interventions that try to make the future self feel closer have only modest effects.
The single most useful assumption is that, in the moment, the present will usually win, so build for that rather than betting on resolve.
Present bias is already in your pricing. Subscriptions and instalments work because spreading a cost into the future makes it feel smaller today, and "buy now, pay later" pushes the pain out of sight entirely. Free trials and one-click purchases front-load the reward and back-load the cost, which is exactly the shape that gets people to act now and regret later. Used well, this helps customers act on intentions they already hold; used carelessly, it tips into harm and invites regulation, so it is worth being deliberate about which side you are on. When you want people to do something good for their later selves (start saving, sign up for the healthier plan), do not rely on persuasion; make it the default, make it automatic, and offer a way to commit in advance.
Policies whose costs land now and whose benefits arrive years later (climate action, pensions, infrastructure) are fighting present bias directly, which is part of why they are chronically hard to sell. Two moves help: make the future benefit feel near and concrete rather than distant and abstract, and lean on automatic mechanisms (auto-enrolment, escalators that rise by default) that do not depend on anyone summoning willpower in the moment.
Present bias is the core case for defaults, automatic enrolment, and commitment devices. Save More Tomorrow, which let employees pre-commit a slice of future pay rises to their pension, raised participants' saving rates from 3.5% to 13.6% (Thaler & Benartzi, 2004), something no amount of urging had achieved. Design for the present-biased citizen, not an idealised planner. And be careful about treating self-control as a measure of character: the deflation of the marshmallow result suggests that a person's apparent patience reflects their circumstances as much as their grit, so stability and resources may do more than lectures about willpower.
Three habits. First, expect the present to win and design around it: set the good behaviour as the default, automate it, and commit in advance while you are still calm. Second, watch for your own preference reversals, the gap between the patient plan you make for next month and what you will actually want when next month is today. Third, resist reading willpower as destiny or character, in yourself or others, because it tracks circumstances far more than the famous story admits.
.png)