ML SRS Model & Research Study

A Smarter Skritter, Personalized for You

Our new machine-learning review system learns how you learn. It adapts review timing based on your strengths and weaknesses, prioritizes the cards that matter most, and helps prevent overwhelming review backlogs.

The goal: learn more, review less, and stay motivated.

Personalized review schedules
Smarter card prioritization
Reduced due card overload
Privacy-protected research
Opt-out anytime

A note from researcher Xufei Liu

Hi everyone!

If you’re on this page, you’re probably interested in how the new algorithm works and how it might improve your Skritter experience.

First, imagine that Skritter can serve as your personal teacher, learning your strengths and weaknesses over time so that each review session helps you become a better learner. Maybe definitions come easily to you, but writing takes more repetition; or maybe reading is your strong suit, but tones need extra attention. When I first started using Skritter as a heritage speaker, I had my own profile of imbalances just like many other students. I almost wished there was a button I could click to tell Skritter I was a heritage speaker and have it personalize my study experience, since it was obvious that reading and writing were much more difficult for me.

That is where the idea for this project came from. We wanted to study how memory and learning work, and how adaptive algorithms can increase learning speed and memory strength.

The new proposed algorithm tracks your strengths and weaknesses over time and adapts the cards you see. Today, Skritter uses the same review rules for every student, regardless of background or learning speed. If you get a card correct 3 times in a row, you’ll wait a week before it appears again. However, under the new system, your past study sessions help Skritter learn about your learning. For example, if reading comes more easily than writing, then writing cards may appear more frequently. Getting a reading card right 3 times in a row might mean the next review interval occurs in 2 weeks instead of 1. Behind the scenes, it builds a personalized profile for each student. As a result, your reviews should feel more tailored to you. The reviews shown may also be a little more challenging than the old system because, by pushing you just a bit, we hope to make your studying more effective and make you a better learner.

In addition, I used to have this problem where I would take a weekend off, and my queue would jump to several hundred cards (as of right now, there are 3,000 cards in my queue…). To combat this, the new algorithm flags cards as “active priority” for review. If the algorithm detects that you are having trouble finishing your queue, it prioritizes the most important cards to show you. Instead of pushing you to finish all three thousand cards to clear the queue, it will show you a smaller, higher-priority set: enough to help you maintain your base of knowledge without burning out.

So if you plan on opting in, here’s the TLDR:

Personalization: Get better recommendations based on your strengths and weaknesses, so the cards you see match what you actually need to practice.
Active Priority: When your queue gets overwhelming, Skritter can focus on the cards that matter most first instead of forcing you through thousands of reviews for catch-up.
Security: Data will remain anonymized. We do not see private, identifying user details.
Data use. We only use study-related information, such as the cards you study and the scores you give those cards. This data will be used to improve the system for future personalization, and the results from this project may be published in the aggregate (for example, students review x% fewer cards and intervals between reviews are x% longer).
Opting out: You can do so at any time. Your future study activity will no longer be used for this project, and you’ll be reverted to the old algorithm.

Xufei Liu

PhD Student | The Wharton School

University of Pennsylvania

Enabling smarter reviews

Tap the hamburger menu from the Home screen
Tap Settings
Go to the Smarter Reviews research study under Review settings options
Tap to opt in and confirm

After joining the research study, your review queue will begin the optimization process. You'll get an alert on the Home screen informing you that the queue has been updated. You'll also see an extra icon on your due cards moving forward.