Hyperbolic discounting
Rewards lose value fast at first, so today beats tomorrow and preferences reverse.
People discount future rewards steeply over the near term and much more gently further out. Because the curve is steeper than exponential discounting, a person can prefer a larger later reward when both are distant and then flip to the smaller sooner one as the near date arrives. That reversal, not impatience itself, is the signature of the effect.
Two ways of valuing a future reward
How it shows up in software
Every product asking for effort now and paying out later is fighting this curve: budgeting apps, language learning, backups, security hygiene. Products that win either shorten the payoff delay or let the user make a binding commitment while they are still far from the temptation. On the other side, one-click purchase and buy-now-pay-later sit on the same curve pointing the other way.
Using it well
- Move the reward closer. A visible result in the first session beats a promise of value in week six.
- Offer commitment devices set in advance: a scheduled savings transfer, a study reminder, a spending cap chosen while calm.
- Break long-horizon goals into sub-goals with real completion, so the payoff schedule matches how people actually discount.
- When you ask for a delayed action, make the future step automatic rather than relying on the user's future self to remember.
Where it turns manipulative
- Buy-now-pay-later placed at the point of impulse exploits the same curve it claims to smooth, and regulators in the UK and US have moved to bring it under consumer credit rules.
- Free trials that convert silently at the moment the user has stopped paying attention take the cost from the version of the person least able to object.
- Loot boxes and timed offers compress the decision window so the near-term pull wins by default.
Where you have seen it
Duolingo
Lessons are short and end with immediate feedback and points rather than a delayed proficiency score.
Monzo
Users can set spending category limits and round-up savings rules ahead of time rather than at the moment of spending.
Apple Screen Time
App limits are configured in advance and then interrupt usage when the limit is hit.
What the research says
- Thaler, 1981Well evidenced
Implied discount rates fell sharply as the delay lengthened: people demanded a much higher annualised return to wait one month than to wait ten years, which exponential discounting cannot produce.
- Ainslie, 1975Well evidenced
A review of animal and human self-control work showing systematic preference reversals as a smaller sooner reward approaches, the empirical basis for hyperbolic curves.
- DellaVigna and Malmendier, 2006Mixed evidence
Gym members on monthly contracts paid substantially more per visit than the pay-per-visit price would have cost them, and delayed cancelling after they stopped attending.
Present bias is well supported. The specific functional form is not settled: Andreoni and Sprenger, 2012, found that when experiments pay with convex budgets and equalise transaction costs, much of the apparent present bias over money shrinks. The effect is sturdier for consumption and effort than for money in a lab.
Grades are a judgement about the evidence, not about how useful the idea is. Plenty of contested effects are still worth knowing, as long as you do not cite them as settled.