Rico

Why Rico works the way it does.

Every part of Rico follows a finding from learning research. Here's each one, what Rico does about it, and where it comes from.

  1. Memory fades on a curve you can predict.

    Forgetting is fast at first, then slows. Each successful recall makes the next drop slower.

    In Rico Every card is scheduled by FSRS-7, an open memory model that tracks how well you know each card and brings it back just as it starts to slip. On the open SRS benchmark it predicts recall better than earlier versions.

    open-spaced-repetition, fsrs-rs and srs-benchmark. Rico's version is checked against 120 reference sequences.

  2. Recalling beats rereading.

    Pulling an answer from memory strengthens it far more than looking at it again, and spreading those recalls over days does the most good.

    In Rico Lessons ask a question after every short step, and cards come back over days and weeks.

    Rowland 2014, a meta-analysis of the testing effect. Szpunar, Khan & Schacter 2013, on tests during lectures. Rawson & Dunlosky, on successive relearning.

  3. A guess first helps it stick.

    Trying to answer before you're taught, even wrongly, makes the lesson that follows stick better, as long as the question asks what's really taught. Curiosity peaks when you're about half sure.

    In Rico Lessons open with a warm-up question that doesn't count, then teach.

    St. Hilaire et al. 2023, on prequestions. Kang et al. 2009, on curiosity.

  4. Mix it up when you practice.

    Practice that mixes topics beats practice that does one topic at a time: in one classroom trial, by a large margin a month later. The first time you meet something, though, related ideas together help.

    In Rico Reviews are mixed so neighbouring topics are kept apart. New cards arrive in small blocks from one course.

    Rohrer et al. 2020, a randomized trial of interleaved math practice. Carvalho & Goldstone 2014.

  5. Learn from worked examples, then do it yourself.

    Studying a worked example, then filling in more of the steps each time, beats solving from scratch. Finding the mistake in a worked example helps too.

    In Rico Step-by-step cards fade the worked steps as you improve and bring them all back after a miss. Spot-the-error cards ask you to find the slip.

    Renkl et al. 2002. Atkinson, Renkl & Merrill 2003. Große & Renkl 2007. Adams et al. 2014.

  6. Say why it's wrong.

    Feedback that explains why an answer is wrong teaches far more than right or wrong alone, and short feedback works best.

    In Rico Every wrong option carries a short reason, and common mistakes get their own message.

    Van der Kleij et al. 2015, a meta-analysis of computer-based feedback. Shute 2008, on formative feedback.

  7. Produce it, don't just pick it.

    Typing or saying an answer strengthens memory more than choosing one. When there are options, three good ones beat four or five.

    In Rico Many cards ask you to type before any options appear. Choice cards use three or four shuffled options.

    Kang, McDermott & Roediger 2007. Haladyna, Downing & Rodriguez 2002. Rodriguez 2005.

  8. The same idea, in new clothes.

    Practising an idea in varied forms helps it transfer to questions you haven't seen.

    In Rico A card can have related versions. Close ones come first; different ones only after you've remembered it on a later day.

    Likourezos, Kalyuga & Sweller 2019. Braithwaite & Goldstone 2015.

  9. Short, and nothing extra.

    Attention drops off after about six minutes of video, and decorative pictures that don't teach anything make learning worse.

    In Rico Lessons run 3 to 10 minutes, videos have questions pinned inside them, and pictures are there to teach.

    Guo, Kim & Rubin 2014, on MOOC video engagement. Mayer, on multimedia learning. Rey 2012, on seductive details.

  10. You know less than you think, and learn more than you think.

    People overrate how well they know something and underrate how much more study will help. Checking yourself against the actual ideas makes your judgment far more accurate.

    In Rico Problem cards ask you to tick off the ideas your answer included instead of a bare “Got it.” Rico shows how its read of you compares with how you actually did.

    Dunlosky & Rawson 2012. Lipko et al. 2009. Kornell & Bjork 2009. Miller & Geraci 2011.

  11. Seeing progress keeps you going.

    Keeping track of progress toward a goal makes you more likely to reach it. Feedback that only warns tends to backfire, and a broken streak can make people quit.

    In Rico Rico shows where you stand and where you're headed. A drop always comes with one thing to do next. The streak is a small caption, with no points or badges.

    Harkin et al. 2016. Kluger & DeNisi 1996. Bandura 1977. Silverman & Barasch 2023.

  12. Show the uncertainty honestly.

    A forecast is clearer as one number with soft bands around it than as a hard-edged range, which people read as all or nothing.

    In Rico The exam forecast is one number, rounded down and never 100%, with shaded bands that narrow as Rico learns how you study.

    Spiegelhalter et al. 2011, on visualizing uncertainty. Correll & Gleicher, on error bars. Wilson 1927 and Brown, Cai & DasGupta 2001, on intervals.

And we test it ourselves.

Rico's planning (exam modes, how it brings you back after a break, the forecast) is checked against simulated learners, including ones who forget faster than the model expects. When a rule didn't hold up in simulation, it changed.