Why users stop attempting, and what to change so they start again.
This skill owns the frame, the stakes and the attribution around failing — the things that decide whether a correctly-targeted difficulty is tolerated or fatal. It does not own the difficulty target itself. If your content is genuinely mis-calibrated, no amount of reframing rescues it; fix the calibration and come back.
Most of what follows is copy, defaults and ordering rather than new features. That is the point — the cheapest interventions in this set are sentences.
Diagnose first
Before picking a fix, work out which input is missing and whether the block is emotional or structural.
- [[references/exposure-and-need-are-both-required]] — skill is exposure × need; either at zero makes the result zero, so more content never rescues an unmotivated user.
- [[references/anxiety-spends-the-same-budget-as-the-task]] — anxiety eats the working memory the task needs, so nervous users score below their real ability and then believe the score.
- [[references/the-blocker-is-willingness-to-be-visibly-bad]] — "I can't do this" is usually avoidance wearing a capacity story; build cheaper failure, not better explanation.
Make the goal scorable
An unmeasurable promise makes every state the user occupies a state of failure.
- [[references/decompose-a-goal-nobody-can-score]] — split the untestable binary into components and can-do bands, and tell users where they may stop.
- [[references/flag-the-stretch-goal-as-a-stretch]] — marking the ambitious outcome in advance keeps its pull and deletes its failure.
- [[references/set-success-at-functional-adequacy]] — define success as "it worked", not "it was correct", and sanction the degraded path by name.
Unblock the attempt
The scary core action is the one that decides whether anything else in your product matters.
- [[references/declassify-the-perceived-prerequisite]] — find the one capability users think disqualifies them and rule it non-blocking in a sentence.
- [[references/lower-the-cost-of-the-first-public-attempt]] — pre-exposure, filler to buy thinking time, and an undo written from inside the changed state.
- [[references/ship-an-easier-variant-for-people-who-failed-before]] — a completable low-exception version so a returner can accumulate evidence against their own verdict.
Frame the difficulty and the failure
Same effort, opposite meaning, depending on what you said beforehand and what you say when it goes wrong.
- [[references/pre-announce-how-bad-it-will-feel]] — specific, unflattering expectations set before the first attempt turn a bad week into confirmation they are on track.
- [[references/attribution-decides-whether-they-continue]] — ship the mechanism with the correction; an unexplained failure becomes evidence about the user.
- [[references/reframe-the-difficulty-rather-than-removing-it]] — a puzzle frame makes the user the agent; an evaluation frame makes them the subject. Check the frame before you sand anything down.
- [[references/suppress-the-noise-so-the-blocking-error-is-audible]] — rank errors by whether they block the goal, and tell users out loud what they may skip.
Sibling skills
- [[../duo-retention/SKILL]] — once they are attempting, what brings them back tomorrow. Churn diagnosis starts there and routes here when the cause is "I'm bad at this".
- [[../duo-gamification/SKILL]] — XP, progression and the difficulty ramp itself. This skill frames the difficulty; that one sets it.
- [[../duo-voice/SKILL]] — the actual wording for error copy, onboarding and celebration. Most fixes here land as copy changes there.
- [[../duo-experimentation/SKILL]] — how to test whether a reframe moved anything, and which guardrail catches it if the reframe cost you accuracy.
Sources
Distilled from blog.duolingo.com — 32 posts spanning 2020-02-27 to 2026-05-06, each cited in dated form inside the node that uses it; see scripts/sources.json for the full bibliography.