Two Principles That Actually Work
Decades of cognitive science research have produced two findings that are robust enough to build a product around: retrieval practice and spaced repetition.
Retrieval practice (also called the testing effect) shows that actively recalling information from memory produces stronger long-term retention than re-reading or passive review. The act of struggling to retrieve something — even failing and then seeing the correct answer — encodes it more deeply than passive exposure.
Spaced repetition is the scheduling algorithm that determines when to show each card again. Material reviewed at increasing intervals (1 day, 3 days, 7 days, 14 days) is retained far longer than material reviewed in a single cramming session. This is mathematically modeled by the SM-2 algorithm, which most serious flashcard systems implement in some form.
What Most Flashcard Apps Get Wrong
The majority of consumer flashcard apps optimize for engagement metrics — streaks, daily card counts, XP systems — rather than learning outcomes. These features are not inherently bad, but when they conflict with the science, the science should win.
The most common mistake: showing cards too frequently. If a user gets a card right three times in a row, showing it again tomorrow feels satisfying but doesn't aid retention. The card should disappear for a week or more. Most apps don't do this because it feels like the app is 'taking away' content.
The second mistake: not testing recall in both directions. Seeing a term and recalling the definition is different from seeing a definition and recalling the term. Both directions should be tested.
How We Translated This into Recall iQ
Recall iQ was built from these principles forward, not from a feature list backward.
The AI generation layer (Claude API backend) produces cards in question-answer format, not definition-term format, because questions force active retrieval framing. A card that says 'What is the capital of France?' produces better recall than one that says 'France: Paris.'
The adaptive exam sessions track per-card performance and adjust future scheduling using a modified SM-2 implementation. Cards you consistently get right are shown less frequently. Cards you struggle with are shown more often and in varied contexts.
The quota and subscription system is built around sessions rather than card counts — because a 20-minute focused session produces better outcomes than flipping through 200 cards mindlessly.
The Research We Relied On
The foundational papers are Roediger & Karpicke (2006) on retrieval practice, Ebbinghaus's original forgetting curve work, and Cepeda et al. (2006) on distributed practice. If you want to go deeper, Bjork's work on desirable difficulties is worth the read.
Recall iQ is available on web and Android. If you're building an EdTech product and want to talk through the architecture, get in touch at /contact.