Commitment Devices: How to Bind Your Future Self to Your Goals

You already know what you should do tomorrow morning. The problem is that tomorrow morning, a different version of you shows up—one who would rather stay in bed, skip the gym, or check email instead of writing. Economists call this gap between your intentions and your actions present bias: the tendency to weight immediate rewards far more heavily than future ones, even when the future ones matter more (O'Donoghue & Rabin, 1999). A commitment device is a deliberate move by the far-sighted version of you to constrain the impulsive version before it takes over. This article looks at what commitment devices are, the real evidence on whether they work, and how to use one without fooling yourself.

The planner and the doer

The foundational idea is older than the apps that now sell it. Thaler and Shefrin (1981) modeled a single person as two agents in tension: a far-sighted "planner" who cares about long-run welfare, and a myopic "doer" who lives in the present and wants gratification now. Self-control, in this framing, is not willpower in the moment—it is the planner using rules, incentives, and external constraints to limit what the doer can do. That is exactly what a commitment device is: a voluntary restriction you impose on your future choices, accepted now precisely because you expect to want to defect later.

The classic image is Ulysses ordering his crew to tie him to the mast so he could hear the Sirens without steering toward them—and to refuse any later order to be released. A modern Ulysses contract works the same way. It has two parts: a binding restriction and a mechanism that keeps you from quietly undoing it when temptation hits. Ashraf, Karlan, and Yin (2006) built a real one into a Philippine bank product, a savings account that locked away deposits until a self-chosen date or amount was reached. Clients who were offered it and opened accounts saved substantially more than a comparison group—evidence that people will pay, in liquidity and flexibility, for a constraint on their own future behavior.

Hard commitments, soft commitments, and deposit contracts

In their review of the field, Bryan, Karlan, and Nelson (2010) draw a useful line between two kinds of commitment. Hard commitments impose real, external penalties: money you forfeit, an account you cannot touch, a bet you lose. Soft commitments work through psychological costs—the sting of breaking a promise you made publicly, or of missing a streak you have been building. Both can help, and the same device often blends them.

The most studied hard commitment is the deposit contract. You put your own money on the line and get it back only if you hit a verifiable target; miss, and the money is gone—sometimes to charity, sometimes to a cause you actively dislike. The design leans on loss aversion: an outcome framed as losing $100 you already handed over stings more than failing to earn a $100 bonus. Deposit contracts pair well with public accountability, because a neutral third party has to verify the outcome, which is also where a tool like TrackInPublic fits—turning "I'll try" into a checkable, visible record.

Does it actually work? The smoking evidence

The strongest single piece of evidence comes from a randomized trial in the Philippines. Giné, Karlan, and Zinman (2010) offered smokers a product called CARES: deposit your own money into a locked account for six months, then take a urine test for nicotine byproducts. Pass, and you get your money back. Fail, and the balance goes to charity. About 11 percent of smokers offered the product took it up. Crucially, those randomly offered CARES were roughly 3 percentage points more likely to pass the test than the control group—and the effect persisted in surprise tests at 12 months, after the money was no longer at stake. A voluntary, self-imposed financial trap produced lasting behavior change.

A commitment device is the sober self writing a contract the drunk self cannot tear up.Adapted from Thaler & Shefrin's planner–doer model (1981)

Deadlines are a subtler version of the same lever. Ariely and Wertenbroch (2002) let university students choose their own deadlines for three papers, with real grade penalties for lateness. Students could have set all three deadlines at the end of term to keep maximum flexibility. Many did not—they voluntarily spaced their deadlines earlier, accepting the risk of penalties as a self-control mechanism. Those self-imposed deadlines improved performance relative to a single end-of-term deadline. But students did not set them optimally: evenly-spaced deadlines imposed by the instructor beat the students' own choices. The lesson is double-edged. People know they have a self-control problem and will act on it, but they are not always good at designing the fix.

Where the online tools come in—and their limits

stickK, co-founded by Dean Karlan (a co-author on much of the research above), turned the deposit contract into a website: you name a goal, set the stakes, appoint a referee to verify your progress, and choose where forfeited money goes—including "anti-charities" whose mission you oppose, to sharpen the loss. It is a faithful commercial translation of the academic idea. Modern habit and accountability apps blend the hard and soft commitments Bryan, Karlan, and Nelson (2010) describe: money at risk, plus a public audience and a visible streak.

A few honest caveats before you sign anything:

• Take-up is low. In the CARES trial only about 11 percent of smokers accepted the offer (Giné, Karlan & Zinman, 2010). Commitment devices help the people who choose them—they are not a mass solution.
• Verification is everything. If nobody checks, a deposit contract collapses into a promise. Choose goals a referee can actually confirm.
• You can set the stakes wrong. Students systematically chose worse deadlines than an outsider would (Ariely & Wertenbroch, 2002); consider borrowing a schedule rather than inventing one.
• Beware the tidy numbers. Popular claims that habits form in "21 days" have no research basis; the often-cited alternative comes from Lally et al. (2010), who found a median of about 66 days to reach automaticity—but with a range from 18 to 254 days. Treat any single magic number with suspicion.
• A device you can quietly cancel isn't a device. The binding part only works if you cannot release yourself the moment the doer takes the wheel.

Used well, a commitment device does not require you to become a more disciplined person. It requires one good decision, made while you are clear-headed, that removes some of your future options on purpose. The planner writes the contract; the doer lives inside it. The evidence—across savings, smoking, and deadlines—suggests that when the stakes are real and the outcome is checkable, that trade is often worth making.