Binding Your Future Self: The Science of Commitment Devices (And Why 55% Fail)

In Homer's Odyssey, Odysseus orders his crew to tie him to the mast so he can hear the Sirens' song without steering the ship onto the rocks. He recognized his future self would be unable to resist, so his present self removed the option.

Twenty-eight centuries later, behavioral economists formalized this idea. In 1956, Robert Strotz published the first mathematical model of time-inconsistent preferences. In 1978, Thomas Schelling proposed modeling a person as containing multiple selves in a "continual contest for control" -- a straight self (the planner) and a wayward self (the one who eats the cake at midnight).

And in 1997, David Laibson gave us the math: hyperbolic discounting. We discount the near future steeply but the far future gently. That's why you genuinely plan to start the diet on Monday, but when Monday arrives, next Monday seems like a better idea. Each postponement feels tiny. The accumulated cost is massive.

The question became: can we build modern equivalents of Odysseus's ropes?

The Evidence For Commitment Devices

In 2006, Ashraf, Karlan and Yin tested the idea in the Philippines. They offered 1,777 bank clients a product called SEED -- a savings account where clients could restrict their own withdrawals until reaching a goal or date they chose. No extra interest. Just a lock on the door that only opened when they said it should.

28% of those offered SEED signed up. After 12 months, their savings were 81-82% higher than the control group. People voluntarily chose to restrict their own freedom -- and it worked.

In 2010, Gine, Karlan and Zinman tested commitment for smoking cessation. They offered 2,000 smokers in the Philippines a CARES account: deposit money for six months, take a urine test, get your money back if you pass. Fail and the money goes to charity. Only 11% took the deal, but those who did were 3 percentage points more likely to quit -- and the effect persisted in surprise tests at 12 months.

In 2015, Kaur, Kremer and Mullainathan ran a year-long experiment with data entry workers in India. Workers voluntarily chose dominated contracts -- contracts that penalized low output but gave no extra rewards for high output. They literally chose to make their own pay worse if they slacked off. Why? Because they recognized they needed the discipline.

This is exactly what the theory predicts. Sophisticated agents -- people who understand their own weakness -- are willing to pay for commitment.

Where It Falls Apart

Here's where the story gets complicated. And honest.

In 2020, Anett John published a field experiment called "When Commitment Fails." She offered low-income individuals in the Philippines a commitment savings product where they chose their own savings plans and penalties.

The result: 55% of clients defaulted and lost money.

They didn't choose contracts that were too hard. They chose contracts that were too weak -- penalties strong enough to hurt when they failed, but not strong enough to actually change their behavior. The theoretical framework explains why: these were partially naive agents. They knew they had self-control problems but underestimated how bad those problems were. So they chose commitment devices calibrated for a better version of themselves that didn't exist.

And here's the mathematical punchline from O'Donoghue and Rabin's work: commitment adoption actually INCREASES with naivete. The people most likely to sign up for commitment are the worst at calibrating it. Even a tiny departure from perfect self-knowledge can flip the welfare effect from positive to negative.

The Acceptance Paradox

Halpern and colleagues published one of the largest studies on commitment contracts in the New England Journal of Medicine in 2015. They offered 2,538 CVS employees either reward programs (~$800 for quitting smoking) or deposit contracts ($150 of their own money at risk, plus $650 reward).

Result: 90% of people accepted the reward programs. Only 13.7% accepted the deposit contracts.

But among those who accepted deposits, quit rates were about 13 percentage points higher than the reward-only group.

The most effective commitment devices are adopted by the fewest people. And the commitment devices most people will actually use tend to be too weak to work.

This is the core design tension of all commitment devices: one that's easy to escape is useless. One that's hard to escape may trap people in bad outcomes.

The Sustainability Problem

There's another uncomfortable pattern. Across study after study, the effects of commitment devices diminish or vanish when the contract ends.

Volpp and colleagues (2008, JAMA) found that deposit contracts produced 14 pounds of weight loss versus 3.9 in controls. But weight was rapidly regained after incentives ended. The SEED savings effect attenuated at 2.5 years. No platform -- StickK, Beeminder, or others -- has published convincing long-term follow-up data.

This raises a hard question: do commitment devices change behavior, or do they just shift its timing?

What Actually Works (According to the Evidence)

The research points to a few things that are more robustly supported:

Implementation intentions have the strongest evidence.

Gollwitzer and Sheeran's meta-analysis of 94 studies (8,000+ participants) found d = 0.65 for "if-then" planning. Later meta-analyses with objective measures found smaller effects (d = 0.14-0.37), but even the conservative estimates beat goal intentions alone. The format is simple: "If [situation], then I will [behavior]." It creates automatic associations that bypass the decision point where present bias hijacks you.

External structure beats self-imposed structure.

Ariely and Wertenbroch (2002) found that externally imposed evenly-spaced deadlines outperformed self-imposed ones. People know they procrastinate but don't calibrate their own constraints well. (Note: a recent replication found negligible effects of self-imposed deadlines, raising questions about the original study.)

Mixed commitment (financial + social) appears most beneficial.

A 2021 analysis of StickK data found that combining financial stakes with social accountability produced better outcomes than either alone.

The Recovery Angle

If you're recovering from burnout and trying to build sustainable work habits, the commitment device research has a specific message for you.

Don't set up elaborate punishment systems for yourself. John's study showed that 55% of people who try that approach end up worse off. And psychological reactance theory (Brehm, 1966) tells us that even self-imposed constraints can trigger a rebellious backlash -- you start fighting your own rules.

Instead:

Use implementation intentions.

"If it's 10am and I haven't started my most important task, then I will close my email and work for 25 minutes." Simple. Specific. No punishment.

Accept external deadlines over self-imposed ones.

A coaching call, a co-working session, an accountability partner. The research consistently shows external structure works better than internal willpower.

Be honest about your sophistication level.

If you've repeatedly set goals and missed them, you're probably partially naive about your own present bias. That means any commitment device you design for yourself will probably be too weak. Build in more margin than feels necessary.

Focus on structural change, not willpower.

Move the phone to another room. Block the websites. Cancel the subscriptions. The research on defaults shows that structure changes behavior where information and intentions don't.

Your future self isn't a better version of you. Your future self is the same you, in a different situation. Change the situation.

Key Studies

- Laibson (1997) -- Hyperbolic discounting model, QJE

- O'Donoghue & Rabin (1999, 2001) -- Sophistication vs naivete, AER/QJE

- Ashraf, Karlan & Yin (2006) -- SEED commitment savings: 81% higher savings, QJE

- Gine, Karlan & Zinman (2010) -- CARES smoking cessation, AEJ Applied

- Gollwitzer & Sheeran (2006) -- Implementation intentions meta-analysis: d = 0.65

- Halpern et al. (2015) -- Deposit vs reward contracts for smoking, NEJM

- Kaur, Kremer & Mullainathan (2015) -- Workers choose dominated contracts, JPE

- John (2020) -- When Commitment Fails: 55% defaulted, Management Science