Start from
Occasions
0
not started
Rule held
—
of the sequence
Confidence
—
in the next one
Earned
—
units so far
The preference reversal
The sequence drag or arrow keys to replay
From a distance, the larger reward is obviously worth waiting for. When it arrives, the small one is suddenly more attractive. The rewards did not change. Your distance from them did. Treat the choice as a rule for a series of future occasions and the future self gets a vote.
The small reward is available now; the large reward arrives later. An exponential discounter loses value at a constant proportion and keeps preferring the large reward. A hyperbolic discounter has a fat tail: the far future still matters, but the near future takes over when it is close enough.
The crossing is not a change of mind about the rewards. It is a change in the point from which they are valued. Move the curve selector, then move the horizon and confidence controls below. The page is asking whether a commitment is an extra preference or a different way to count the same preferences.
Ainslie's puzzle is not simply that people like small rewards. Everyone likes rewards. The puzzle is that the same person can prefer the larger reward when both options are remote and prefer the smaller one when the choice becomes immediate. A hyperbolic curve models that reversal without changing the payoffs or adding a mysterious faculty called willpower. Nearness changes the relative weight of the options.
That leaves a familiar but incomplete description: the person made a plan and then failed to follow it. The bundled comparison asks a different question. What if the choice is not one small reward against one large reward, but the whole series of occasions on which the rule could hold? A larger reward at every future occasion can outweigh the first attractive lapse, but only if the person expects the series to remain intact.
A commitment changes what counts as the choice. It does not change the shape of the curve.
A bright line is easier to treat as a precedent than a vague one. Abstaining from one named behaviour can be recognised as a test case; eating one meal or spending one sum can be explained away as an exception. The model represents that difference as clarity. When an occasion is not seen as part of the rule, it is judged as a single choice and the present reward wins. A rule can therefore be perfectly understood and still fail because its instances do not look like instances.
This is not a theory of addiction, a clinical measure of self-control, or a claim that every plan should be treated as binding. It has one decision, one discount curve and one confidence state. It leaves out changing rewards, social commitments, emotion, fatigue, habit, and the many reasons a person may rationally revise a plan. Its narrower result is useful: if the future pattern is part of the value being compared, then confidence in that pattern is not an afterthought. It is one of the stakes.
The slider is the argument. Lower confidence until the bundle stops paying, make the rule vague, and then remove the bundle entirely. The model does not decide whether commitment is admirable. It shows why a repeated policy can be worth more than the first occasion on which it is tested.