Contribution
A review that reframes undersaving. Earlier accounts looked at how people value present and future rewards. This one looks at how people see the self who will receive the future reward. It draws Parfit’s view of personal identity as degrees of psychological connectedness into a behavioural model, proposes three components of felt continuity with the future self (similarity, vividness, positivity), and reviews studies, mostly the author’s, that measure or manipulate each and find more patient choices.
Key insight
Prudence is a kind of other-regard whose object may or may not feel like oneself. “If people tend to consider the future self as a stranger … then they may rationally have no more reason to save their money for their future selves than to give that money to a stranger” (Theoretical underpinnings). The model’s prediction is graded: “the degree to which an individual feels disconnected from his or her future self should correlate with the degree to which that individual discounts future rewards” (ibid.).
Assumptions
- Parfit’s reductionism, psychologically read. The self is “a collection of distinct identities that overlap with each other over time”, connected by degrees (Theoretical underpinnings). What matters is not whether a person changes “but how much a given individual feels he or she will be the same person over time”.
- Helping follows similarity. Since people help similar others more (Eisenberg and Miller; a microfinance field study), one “donates” to a future self by saving (The composition of continuity).
- Discounting tasks measure concern for the future self. The outcome in most studies is the number of larger-later rewards chosen on Kirby and Maraković’s questionnaire, or hypothetical retirement allocation.
Key results
- The measure (Similarity, Measuring the relationship). Participants choose, from pairs of Euler circles ranging from no overlap to almost complete overlap, the pair that “best represents how similar they feel to their selves in ten years time”. It is “based on existing measure of self-other similarity”, citing Aron, Aron and Smollan 1992. It had “strong test-retest reliability” and correlated with a “me/not me” task matching present and future self-descriptions. No figures are given.
- Similarity predicts patience and assets (same section; Ersner- Hershfield, Garton et al. 2009). In undergraduates, with hypothetical and incentive-compatible choices, more similarity went with more larger-later choices. In working adults aged 18–86 in the San Francisco Bay Area, it went with more accumulated assets, controlling for age and education. No coefficients are reported.
- Manipulated similarity (same section). Bartels and Urminsky: a passage saying identity is stable over time raised overlap with the future self and patience; similarity carried unique variance with uncertainty about the future, affective appraisal of outcomes and present bias controlled. Bartels and Rips found comparable or stronger results with similarity defined as likes, values, ideals and interests.
- The future self as another person, neurally (Neurobiological basis; Ersner-Hershfield, Wimmer and Knutson 2009). In fMRI, judging trait words for one’s current self activated rostral anterior cingulate and medial prefrontal cortex more than for one’s future self, which resembled judging another person. The larger a participant’s current- versus-future difference in rACC, the steeper their discounting a week later: r = 0.47, P < 0.05 (Figure 2 caption).
- Vividness (Vividness of the future self).
- Undergraduates who saw an age-progressed avatar of themselves in a virtual mirror allocated more to a hypothetical retirement account than those who saw their current-age avatar.
- Seeing one’s own aged avatar, against another participant’s aged avatar, after a delay, produced more patient choices: “relevance matters”.
- On a web slider, a future-self face (happier as more was saved) raised allocation over a current-self face; neutral faces did too, in working adults.
- Positivity (Positivity of the future self; Hershfield and Galinsky,
unpublished in 2011).
- Earlier, directly asking “how much they liked and cared for their future selves” gave a null: “Neither construct, however, mapped robustly onto intertemporal choice.” Both answers were positively skewed with little variance.
- Instead: national attitudes toward the elderly across 26 countries tracked gross national saving, with controls; taking an elderly man’s perspective, recalling respecting an elderly person, and a word search pairing “elderly” with “esteemed” each raised larger-later choices against controls.
- Open questions the author raises (Remaining questions).
- Similarity to and liking of the future self showed “a significant correlation, albeit a relatively weak one”.
- With age, the future self “may not be confined to just the self, but may also include one’s children and grandchildren. In this way, the future self may continue existing after one has died in the form of one’s offspring” (Changes in future self-continuity with age).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Felt similarity to one’s future self predicts patient intertemporal choice and accumulated savings | moderate: correlational lab and field studies and one manipulation by others; no effect sizes given here | Similarity |
| C2 | The future self is represented neurally more like another person than like the current self, and more so in steep discounters | weak to moderate: one fMRI study, r = .47 at P < .05 in a region of interest | Neurobiological basis |
| C3 | Making the future self vivid raises saving | moderate: four experiments, hypothetical allocations | Vividness |
| C4 | Respect for older people, as a proxy for positivity toward the future self, raises saving | weak to moderate: a 26-country correlation and three experiments, unpublished when reviewed | Positivity |
| C5 | Directly reported liking and caring for the future self do not predict choice | weak: a null from measures near ceiling | Positivity |
| C6 | Future self-continuity is one construct with three components | conjecture: the author’s model, with untested links (Figure 4) | Remaining questions |
Concepts
- future self-continuity: how connected a person feels to the self they will be; graded, not all-or-none.
- future self-similarity: felt likeness to one’s self in ten years, measured with overlapping circles.
- vividness: how easily and realistically the future self is pictured.
- positivity: whether the future self is viewed favourably.
- temporal discounting: caring less about outcomes the later they come.
- planner and doer: the economist’s dual-self model, set aside in favour of Parfit’s overlapping selves.
Connections
- Parfit (THEORY-464). Hershfield takes Parfit’s connectedness as the thing felt. He also says “what matters, for long-term planning, however, is that one person has but one identity, and it is with this link that the assets of the present and future selves are tied together” (Theoretical underpinnings). So he keeps numerical identity as what ties the assets and puts felt connectedness in charge of motivation. That is THEORY-464’s point, that prudential concern follows a relation that comes apart from identity, found as a fact about behaviour.
- Aron’s inclusion of other in the self (LIT-1706; NOTE-1276). The circles are Aron’s, relabelled from closeness to similarity. Branand, Mashek and Aron report the IOS as validated for closeness; here the adaptation is validated against overlap of self-descriptions.
- Social discounting. The model cites Jones and Rachlin’s finding that people discount rewards to others by social distance as they discount by delay. Self and other, now and later, sit on one scale.
- Wade-Benzoni et al. 2010 (NOTE-1259, read the same day). They hold that “one cannot create a legacy … by affecting merely one’s own future self”. Hershfield’s closing suggestion runs the other way: at life’s end the future self may continue in one’s offspring.
- Tarsney (NOTE-1247). That note asks that any test of concern tracking felt continuity separate it from discounting for the risk that one’s effects are undone. Bartels and Urminsky’s control for uncertainty about the future does this within a lifetime.
Bearing on the record
-
The triviality objection (CLAIM-492). Has future-self continuity been separated from caring about one’s future self? Partly, and only at second hand. These are inferences from the review’s reports.
- For separation. In Ersner-Hershfield et al. 2009 the circles predicted patient choice and directly asked liking and caring did not. Bartels and Urminsky found similarity predicted choice with uncertainty, affective appraisal and present bias controlled. The neural index, which asks nothing about caring, predicted discounting. And the circles are validated against overlap of self-descriptions, a cognitive criterion, not against closeness.
- Against. The caring null comes from answers “positively skewed” with little variance, so it shows a ceiling, not irrelevance. Similarity and liking were correlated. Positivity, the component nearest caring, is itself proposed as part of continuity and, through respect for the elderly, raised saving. No study reported here pits a well-varied caring measure against the circles.
- So this is the record’s best case yet of a measure of the self’s extent predicting concern-like behaviour where reported caring did not. But it is a case within one person over time, and the separation is not established.
-
The owner’s thesis (CLAIM-491). The review states the thesis’s mirror image. The thesis says concern for others is self-interest on a larger self. Hershfield says concern for one’s own future is other-regard toward a self that may feel like a stranger. Both put self and other on one scale of overlap. If that scale is real, “no action is selfless” and “no prudence is purely self-interested” are the same claim read from opposite ends, and the thesis’s content lies in the scale and its criterion, not in the word “self”. That is an inference.
-
Batson’s objection (CLAIM-494). Indirect. Hershfield borrows the similarity–helping link from the empathy literature to explain saving. It does not test empathy without overlap.
-
The legacy question (QUESTION-301). The question cites this review for prudence having the thesis’s shape. That holds. The review adds an untested suggestion that the future self extends past death into one’s offspring, the bridge from prudence to legacy that the question needs.
-
A dated note on CLAIM-492 is proposed:
Hershfield (NOTE-1259) reviews the strongest case the record holds of a measured extent of the self predicting concern-like behaviour where reported caring did not. Felt similarity to one’s self in ten years, chosen as overlapping circles and validated against overlap of self-descriptions, predicted patience and savings; asking how much people liked and cared for their future selves did not. But the caring answers were near ceiling, similarity and liking correlated, and the case is within one person over time. The separation the objection asks for is suggested here, not established.
-
No THEORY filed.
-
No ML instruction.
Limitations
- A review without numbers. Apart from r = .47 in a figure caption, no effect sizes, sample sizes or test statistics are given. The studies were not read.
- Mostly the author’s own work, several studies “in press” or unpublished in 2011.
- Saving, not sacrifice. Outcomes are hypothetical allocations and questionnaire choices between smaller-sooner and larger-later sums.
- Read from the accepted manuscript, without page numbers or figure images, and not compared with the published text.
- Errors and typos.
- Figure 2’s caption says “Figure adapted from Ref. 72”; ref. 72 is Mitchell et al.; the study is ref. 73.
- “based on existing measure” (missing “an”).
- “Maracovic” and “Marakovic” for Maraković’s name in the text.
- In the references: “Social Secruity and Heal Care”, “acros time”, “chumpto champ”, “identifcation”, “Communicaton”, “comparions”, “favortism”.
Open questions
- With caring measured so that it varies (for instance, about a future self described as worse off), do the circles still predict patience when caring is held fixed? That is CLAIM-492’s test, run where it is easiest.
- Does felt continuity with one’s children predict saving or sacrifice for them, as the review suggests? That would carry the measure across death.