Will it be possible to opt out of this?
This seems poorly thought out. The reason I use FC is to solidify information I would otherwise forget. I am much more likely to forget the low yield material of past content than lets say "what are the major functions of insulin?". A much better idea would be to bring back the "never see this again" option which somehow magically disappeared during the transition.
Which piece specifically are you concerned about? The concept of priority levels or marking the topics as learned in order to unlock the related questions for review? Happy to chat through either. In the meantime, good catch on the latter point; I should have been a bit more clear regarding the purpose of this, rather than implying it was a catch all solution. Put simply, the concepts of priority and yield are relative functions of your goal board exam score. All things being equal, a student who starts studying 2 years out from her Step 1 with a desired score of 230 only needs to master a certain percentage of the total content, while a student who starts at the same time but with a desired score of 260 would need to master a much greater percentage of that content (no duh, right?). For the former student, then, "high yield" equals the absolute minimum percentage of content needed to achieve that goal score. For the latter student, "high yield" also equals the absolute minimum percentage of content to achieve that goal score, however that minimum percentage is much greater. No student, regardless of goal score, needs to master everything. However, the amount that does need to be mastered is largely dependent on that goal score. As such, introducing priority and yield as filtering functions allows a recommendation engine to always maximize for efficiency. Having said that, let me dig in a bit deeper (or feel free to bail out here).
One of the fundamental challenges of any spaced repetition system is in both keeping students on top of the content they've studied and knowing when to present them with new study content, while at the same time minimizing the amount of time they're expected to spend doing this. This issue is compounded when the volume of content that must be learned and retained is extremely high, as is the case with medical education. Put simply: When there is an exceptionally high volume of content, keeping a student on top of all past learned stuff is difficult, without also having to make the student commit an egregious amount of time to continually refreshing everything. (Thats enough for the boring stuff.)
This is why introducing the concept of yield is a thoughtful way for any spaced repetition system to begin prioritizing and sequencing all of the available content. Essentially, yield helps by giving the system a great starting off point, allowing it to more quickly and efficiently work through the content in order to gauge a student's weaknesses and strengths, and by doing so calibrate future recommendations. This is especially important for a few specific cases: For example, for students who come into a spaced repetition system midway, having accumulated a moderately large volume of information, but not having used the system to continually refresh the information. In such a common case, the system struggles to make sense of all of that past content because it has no history of data to rely on, and as such is forced to present it all as if it was equally important. However, the simple fact is that not all concepts are of equal importance. So, by using the concept of yield, it allows us to better narrow down the 'past' stack in order to get an accurate assessment of each student's progress and knowledge. In addition, this allows the algorithm to reduce the required number of daily questions that each user is expected to complete to something that is much more manageable, yet is still sufficient for getting a student where s/he is trying to go (i.e. to a goal exam score).
All of that being said, a properly calibrated recommendation engine is driven first and foremost by contextual inputs, namely the many variables that contribute to the uniqueness of each individual student. Some students start using a spaced repetition system on day one, while others will join after a year of studying. Some students will be content with a Step 1 score of 230, while others will be disappointed with anything below a 260 (Gunners gonna gun). Some students will need to review a concept 5 or 6 times before they can commit it to memory, while others only need 1 or 2 recalls. All of this (and inputs like them) need to drive the system. Over time, as we learn more (meaning as the student does more), the recommendation gets more finely tuned. This means that for those more intense students, yield will cease to be a filtering factor, much in the same way that the number of recommended daily questions might increase).
I know that's a wall of text, so let me summarize: This is new, but I promise you it wasn't just haphazardly tossed in. Anyone working with spaced repetition or recommendation algorithms is used to having to work through some pretty cognitively complex problems (which I actually think is why we're attracted to the medical education space). Regardless, this certainly isn't a do or die situation. We won't all-out block any student from accessing any content; rather, the system will always calibrate itself to maximize efficiency, specific to each individual student. That said, thoughtful feedback like this is
invaluable in that it allows me to better identify concerns that I can address in future communication and product updates. We're all on the same team at the end of the day, so the more candid you are, the better a product I can build for you.... Cheers.
