---
name: Effort justification
slug: effort-justification
category: Motivation
categories: [Onboarding, Retention]
evidence: mixed
page: /principles/effort-justification/
updated: 2026-10-15
---

# Effort justification

> After paying a real cost for something, people revise upward what they think it was worth — the cheaper conclusion is that the cost was wasted.

## What it is

Effort justification is retrospective revaluation. You went through something demanding to get a result; the result turns out to be ordinary; the two facts sit badly together, and the one that is still editable is your opinion of the result.

## Evidence

Three accounts compete for the mechanism, and the field has not settled on one. Festinger's dissonance account, which Aronson and Mills tested in 1959, says the discomfort of holding both cognitions is what drives the adjustment. Self-perception theory says nothing needs to hurt: you infer from having worked hard that you must have wanted the thing. Animal work favours a third reading, contrast and state-dependent valuation, in which effort produces an aversive state and whatever arrives next looks better by comparison. Inzlicht, Shenhav and Olivola put the case against dissonance plainly — effort–value effects now replicate in rats, pigeons, starlings and even invertebrates such as locusts, and “given that dissonance relies on a deliberate justification process, it is unlikely to explain the positive valuation of effort across these different species.” Their 2025 follow-up chapter adds a blunter possibility: that “the paradox may be illusory, with effort's apparent value arising from misattribution of emotions from outcomes rather than effort itself.”

The dispute matters less for interface work than the one structural fact all three accounts share: the effort has to be the thing that produced the outcome. In every paradigm where the effect appears, work is instrumental — the initiation is what gets you into the group, the assembly is what makes the box exist. Effort merely adjacent to the outcome is just cost. The review's formal model is explicit that effort magnifies the outcome in whichever direction it ran: “both positive and negative outcomes would be magnified by the amount of effort exerted, such that winning becomes sweeter (and losing more bitter) the more effort you exerted.” A long path into a mediocre result makes the result read worse, not safer. Three neighbouring principles divide the territory: the IKEA effect is narrower, covering a specific artifact the person built, where completion and perceived contribution rather than effort are the dials; the labor illusion is the observer's side, someone else's visible work raising your estimate of a service; and if users persist because of what they have already spent, that is sunk cost, a prediction about continuation rather than about liking.

Aronson and Mills (1959) is the source study, and it is small. Sixty-three college women were randomly assigned, 21 per cell, to a severe initiation (reading twelve obscene words and two explicit passages aloud to the experimenter), a mild initiation (five sex-related but not obscene words) or no initiation, then all heard the same deliberately banal tape of a group discussion and rated it and its members on 17 scales. Totals: control M = 166.7 (SD = 21.6), mild M = 171.1 (SD = 34.0), severe M = 195.3 (SD = 31.9). Severe beat control, t = 3.32, p < .01, and severe beat mild, t = 2.33, p < .05. The result usually dropped is the one that matters most here: control and mild were indistinguishable, t = .49, n.s. The authors' own conclusion is that “it is not an initiation per se that leads to increase in liking for a group. The initiation must be severe enough to constitute a genuine investment and to render it difficult to reduce dissonance by playing down the extent of the pain involved.” Small costs get discharged by deciding the cost was trivial.

Gerard and Mathewson (1966) is the canonical follow-up and the cleaner design, replacing embarrassment with electric shock and adding conditions meant to rule out arousal, relief and afterglow. The primary report could not be opened for this entry, so what follows is Oxenham's review in The Psychologist: the more severe the shock, the higher initiates rated both the discussion and its members; being told you had passed the screening test lowered those ratings rather than raising them, against the afterglow explanation; and the shocks raised liking only among women recruited to join a discussion club, not among those who believed they were simply in a psychology experiment. That last contrast is the instrumentality condition arriving as data — the same pain bought nothing when it bought nothing. The same review records that the procedure pushed at the ethical limits of its time and that few comparable studies have run since, which is why the canonical base is still two experiments from the 1950s and 1960s.

The modern record is not reassuring. Kamau (2013) re-ran the Aronson and Mills design with 70 participants plus an extrinsic reward manipulation: severe initiation did not produce more group identity than mild, and effects on ratings of the discussion were non-significant. What moved identity was reward — mild initiation plus reward beat severe initiation plus reward, while under severe initiation reward made no difference at all. Oxenham's objection is fair, that an initiation built from arithmetic subtraction tasks is not the dose the earlier studies used, and one underpowered non-replication settles nothing; but there is no meta-analysis and no large preregistered replication of severity-of-initiation to set against it. The mechanism took a separate hit. Vaidis, Sleegers and colleagues (2024) ran a constructive multilab replication of induced compliance — dissonance research's flagship design — with 4,898 participants across 39 labs in 19 countries and 14 languages, found no significant attitude difference between the high-choice and low-choice conditions, saw the pattern hold across exclusion rules, lab variability and attitude measures, and concluded that the results call into question whether induced compliance provides robust evidence for cognitive dissonance. That is a different paradigm, so it does not falsify this effect, but you cannot cite dissonance as the reason your interface works.

What survives well is the broader effort–value relationship rather than the initiation story. Inzlicht, Shenhav and Olivola (2018) assemble convergent evidence that the products of effort are valued more: sunk-cost effects, reluctance to part with money earned rather than won, the IKEA effect, amplified reward signals in striatum, caudate and subgenual anterior cingulate when rewarding feedback follows effortful performance, and the cross-species results. Their own asymmetry is worth keeping — “there is more work supporting value for the products of effort than for the value of effort itself” — and the IKEA effect in their table requires products people “successfully build or prepare themselves.” Olivola and Shafir's martyrdom effect is the one respectable argument for designing effort in, and it is about anticipated effort: 136 undergraduates imagining a tsunami-relief fundraiser offered more when it was a five-mile run (M = $23.87) than a picnic (M = $13.88), permutation p = .004, t(68.57) = 2.87, d = .70, replicated at $26.33 against $16.59 (d = .51) for a Katrina appeal and once with real money and a real cold-pressor task ($4.17 against $3.18 of a $5 budget, N = 36). Perceived meaningfulness partially mediated it and the effect was strongest for causes tied to human suffering. The same experiment warns you off: shown both events side by side, 76% ranked the picnic as more appealing, χ²(1) = 8.76, p < .004.

The honest verdict is mixed. The direction of the effort–value link reproduces across labs, species and paradigms, which is why the acknowledgement half of this principle is worth building on. The specific claim that putting people through more to get something makes them like it more rests on two tiny mid-century experiments, one of which could not be read in the original, has one published failure to replicate, lacks a meta-analysis, and sits on a mechanism whose flagship paradigm failed a 39-lab test. Nothing here licenses manufacturing effort; the evidence supports naming the effort that is already happening.

## When it holds

- Only when the effort actually produces the outcome. This is the load-bearing condition: initiation granted membership, assembly produced the box, the run raised the money. Effort bolted onto a flow that would deliver the same result without it is friction, and nothing here predicts a valuation gain from it. Gerard and Mathewson's shocks raised liking only among women recruited to join the discussion club, not among those who believed they were simply in a psychology experiment.
- Not at small doses. The mild initiation in Aronson and Mills was statistically identical to no initiation, because trivial costs get discharged by deciding the cost was trivial. Interface friction lives almost entirely in the range the original study showed does nothing.
- Not at large doses either, ethically or empirically. The dose that worked in 1959 and 1966 was embarrassment and electric shock; the one modern attempt at the severe end did not reproduce the effect, and the research itself is now considered ethically marginal.
- Only on a good outcome. The effort-as-value model magnifies losses as well as wins, so a long onboarding in front of a product that underdelivers amplifies the disappointment. The labor illusion literature shows the same reversal for visible effort followed by a weak result.
- Only when the effort was chosen. Every account that relies on self-justification requires volition, and choice is exactly the manipulation that produced nothing in the 39-lab induced-compliance replication. Effort you impose reads as a cost imposed on the user, because that is what it is.
- Reward may matter more than severity. In the only modern replication attempt, the reward manipulation moved group identity and the severity manipulation did not, and reward did nothing when initiation was severe.
- Completion, where the artifact is the thing. The IKEA-effect boundary applies whenever the user is making something: unfinished work produces no premium.
- Individual differences are real. Need for cognition, a stable trait predicting enjoyment of effortful thinking, changes how much people value mental exertion, so the same demanding flow will delight part of your base and repel the rest.
- Prospective effort is a special case. Anticipated effort raised charitable contributions, mediated by meaningfulness — evidence about donation and volunteering framing in causes with moral weight, not a licence to make a checkout harder.

## How to apply

- Show what the work bought at the moment of payoff, not how long it took. “You reviewed 420 cards; your accuracy on this deck went from 61% to 88%” ties the cost to the outcome it produced, where “you spent 6 hours here” does not.
- Put milestones at real thresholds — a skill unlocked, a dataset complete, a project shipped — and print the criterion on the marker, so the milestone carries information rather than confetti.
- State on the certificate what was done: the modules, the hours, a verification link. Make it exportable and shareable, because a credential that cannot leave your product cannot do the job a credential does.
- Publish each badge's criterion and make it non-trivial. One badge per genuinely hard thing beats a cabinet of participation awards, and a user who notices the badges are free stops reading any of them as evidence.
- Build recaps from logged activity — an end-of-course summary, an annual review, a post-migration report of what moved — and show nothing when the data is thin.
- Offer the demanding path as an option rather than the route: an advanced setup, a full tutorial, a manual review mode. Chosen effort is the only kind the mechanism is even hypothesised to act on.
- Spend the effort budget on the step that produces the result — the practice, the review, the decision — and strip steps, waits and required reading out of everything else.
- After a demanding step, lead the screen with the outcome and keep the next action small, since effort magnifies what comes next in both directions.

## What to avoid

- Adding friction to inflate perceived value. The evidence is against it from four directions: small costs produced nothing in the original study, the severe manipulation failed to replicate, imposed rather than chosen effort is the condition that produced a null across 39 labs, and effort amplifies a bad outcome into a worse one. You pay a measurable cost in drop-off to buy an effect that may not exist.
- Lengthening a flow and calling it commitment. Extra steps, forced waits and required reading that does not change the result are cost without instrumentality.
- Hazing dynamics in community design. Gated, uncomfortable initiations modelled on this research reproduce exactly the practice its own authors flagged as ethically marginal, and the modern replication attempt did not find the payoff.
- Rewarding effort you did not require, or requiring effort you do not reward. A badge for nothing teaches users to ignore badges; a hard milestone with no acknowledgement wastes the one application that is legitimate.
- Fabricated credentials. A certificate that implies accreditation you do not have, or a badge whose criterion is undisclosed and arbitrary, is a false claim about attainment, not a motivational flourish.
- Effort as an exit tax. Where EU law applies, this is a hard line, not a judgement call. Article 25(1) of the Digital Services Act prohibits providers of online platforms from designing, organising or operating their interfaces in a way that deceives or manipulates recipients of the service, or otherwise materially distorts or impairs their ability to make free and informed decisions, and Article 25(3)(c) names making the procedure for terminating a service more difficult than subscribing to it as a practice the Commission may issue guidelines on. Friction on the way out is regulated; friction on the way in is merely ineffective.

## Measuring it

- Worked: retention, repeat engagement and renewal among users who received the acknowledgement versus matched users who did the same work without it, measured after the milestone rather than at it. Hold the work constant and A/B the milestone, certificate or summary; an experiment that varies how hard the task is measures your drop-off curve, not this principle.
- Worked: how often the artifact is used away from your product — certificates downloaded, shared, verified. Credential value shows up outside the session.
- Backfired: completion rate and time to first value for the flow containing the effort. If completion falls, you have manufactured the one condition under which nothing works.
- Backfired: satisfaction and support contacts among users who completed the demanding path. Effort before a weak outcome reads worse than no effort before the same outcome.
- Backfired: badge criteria being gamed. If activity rises on the exact metric a badge counts and nowhere else, the badge is now the goal and the attainment it was certifying has gone.

## Live examples

- Hers runs a staged consultation before any treatment plan appears, so what arrives at the end reads as the product of the answers given rather than a catalogue page.
- Replika asks what the user wants the companion to help with before it exists, so the thing they meet afterwards is one they specified.
- bunq has the user tick together the accounts and features their plan will contain, so the setup is configured rather than handed over.

## References

- Aronson & Mills (1959). The Effect of Severity of Initiation on Liking for a Group. Journal of Abnormal and Social Psychology. — https://doi.org/10.1037/h0047195
- Gerard & Mathewson (1966). The Effects of Severity of Initiation on Liking for a Group: A Replication. Journal of Experimental Social Psychology. — https://doi.org/10.1016/0022-1031(66)90084-9
- Oxenham (2015). The Initiation Ceremony Experiments. The Psychologist. — https://www.bps.org.uk/psychologist/initiation-ceremony-experiments
- Kamau (2013). What Does Being Initiated Severely Into a Group Do? The Role of Rewards. International Journal of Psychology. — https://doi.org/10.1080/00207594.2012.663957
- Vaidis, Sleegers, van Leeuwen et al. (2024). A Multilab Replication of the Induced-Compliance Paradigm of Cognitive Dissonance. Advances in Methods and Practices in Psychological Science. — https://doi.org/10.1177/25152459231213375
- Inzlicht, Shenhav & Olivola (2018). The Effort Paradox: Effort Is Both Costly and Valued. Trends in Cognitive Sciences. — https://doi.org/10.1016/j.tics.2018.01.007
- Inzlicht, Campbell & Saunders (2025). Effort Paradox Redux: Rethinking How Effort Shapes Social Behavior. Advances in Experimental Social Psychology. — https://doi.org/10.1016/bs.aesp.2025.04.002
- Olivola & Shafir (2013). The Martyrdom Effect: When Pain and Effort Increase Prosocial Contributions. Journal of Behavioral Decision Making. — https://doi.org/10.1002/bdm.767
- Regulation (EU) 2022/2065 (Digital Services Act), Article 25 — online interface design and organisation. — https://eur-lex.europa.eu/eli/reg/2022/2065/oj/eng

## Similar principles

- IKEA effect — /agents/ikea-effect.md
- Labor illusion — /agents/labor-illusion.md
- Sunk cost fallacy — /agents/sunk-cost-fallacy.md
- Commitment and consistency — /agents/commitment-and-consistency.md
- Self-determination theory — /agents/self-determination-theory.md
- Endowed progress effect — /agents/endowed-progress.md
- Goal gradient effect — /agents/goal-gradient-effect.md
- Streak effect — /agents/streak-effect.md