At Cerego, we understand that knowledge is vital for learning, expertise, and creativity. Our mission is to help learners become knowledgable, and for that knowledge to last, and remain useful over the long term. Without engagement, all knowledge naturally fades from memory over time. The rate of forgetting, which scientist Hermann Ebbinghaus defined as the forgetting curve, is related to its retention.
How it works
Knowledge for which you have low retention fades quickly, and is difficult to recall if you need it some days or weeks later. On the other hand, knowledge for which you have built high retention fades more slowly, so that you can recall and use that information when you need it, weeks, months, or even years down the line.
When we talk of building expertise and ultimately mastery over a subject, we’re really talking about retention. It’s possible to remember almost anything if you saw it very recently, but unless you build retention that knowledge will quickly fade again and be unavailable when you really need it. That’s why goals and progress in Cerego are centered on building retention, not test scores, resulting in lasting, useful knowledge.
So, given that retention is so important, how do we measure it? How do we know we’re doing a good job building it, and how much time and effort does it take to build up?
The Cerego adaptive learning platform combines research from cognitive science with machine learning techniques and a rich database of real-world learning data. This allows us to constantly test and improve the review schedule for every learner, and yield insights into learning over long periods of time that are hard to uncover in a research lab. Importantly, we can track how every individual memory in Cerego fades and stabilizes over time, and how that is affected by reviewing in Cerego.
A look at the data
To understand the effect of reviewing in Cerego, we examined data from a sample of 7,334 learners in Cerego:
- 3,582 following an optional online course
- 3,752 enrolled in a college-level economics or statistics course
- Together, over a period of several months, these learners recorded 16.6 million interactions with Cerego, in which they either learned a new concept or reviewed an old one.
How reviewing in Cerego builds up stable memories
The chart above shows how users’ ability to remember material changes over the six weeks since they last saw it. Naturally, the longer since you last saw something, the harder it is to remember it—but the number of times you reviewed the material before that has a big impact on whether it fades quickly from memory, or stabilizes. Each line shows the average change in accuracy after a certain number of reviews on Cerego. The gray line, for example, shows that after studying the material for the first time, the ability to remember it quickly drops from around 80% in the first couple of days to 63% three weeks later, and to below 60% after six weeks—simply studying something new doesn’t mean it sticks.
"Naturally, the longer since you last saw something, the harder it is to remember it—but the number of times you reviewed the material before that has a big impact on whether it fades quickly from memory, or stabilizes."
On the other hand, coming back to something you’ve learned and trying to retrieve it is the best way to build lasting memory for it: An important principle in cognitive science known as the testing effect. It’s important to remember that the review doesn’t have to be accurate—even if you get it wrong, making an effort to recall something and being reminded of the right answer afterwards is a valuable learning experience that boosts your retention for the material almost as much as the reviews you manage to recall correctly.
This is clear from the data above. Just completing a single review on Cerego (about 30 seconds per item, a few hours after first studying it) has a big effect on long-term retention for that material—even if you get it wrong!
An immediate 5% improvement in the first day or two grows to a 10% difference a few weeks later, as the memory for that material is more robust, and fades much more slowly. And completing 4 reviews (something that usually takes around 2 minutes per item over 3-5 days) builds a stable mastery of the material that gives 90% accuracy even weeks later.
Why is this important?
Well, in Cerego an instructor can set a retention goal for learners to achieve. The simplest goal that Cerego recommends is to reach Level 1 retention: this takes around 3-4 reviews in Cerego to reach, equating to around 2 minutes of work split over about 4 days. As we’ve seen, those 3-4 reviews aren’t just busywork—they get a learner from 60% to 90% accuracy six weeks later!
The evidence is clear: following your schedule in Cerego, even just for the first few days, builds mastery for material that sticks around for weeks and months afterwards.