This is purely a conjectural question though someone asked me and, just as a thought exercise, I thought I’d ask here where all the gurus hang out.
The example question was: for an example note with two clozes, C1 and C2, what considerations would be present for Anki to be able to handle those cards as being related?
In other words, for retention/review history of C1 to determine or influence the interval for C2 (specifically if you know C1 down pat it might increase the interval for C2, as an easy example, or vice versa). Why this would be an untenable idea, how exactly it would be difficult to implement, etc.
To be clear: I’m not suggesting that this should be the case. I have my own thoughts, and suspicions, but there’s a lot of shared and communal knowledge on here and I wanted to bounce some ideas off the big brains.
We’ll need to look at the review history for this just like the FSRS optimiser does. Potentially, what should happen is every time you review a sibling card we should assume a “partial review” happening across the sibling cards.
I think @expertium tried a similar idea where the S would increase every time you did a sibling of some card.
The limitations are both technical and theoretical, but mostly technical. Theoretically, it’s hard to determine how much difficulty and stability of card A should be changed when card B is reviewed. But I’m sure me and LMSherlock could come up with something reasonable.
The technical limitation is that in Anki, when you are reviewing a card, FSRS and SM-2 cannot access the review history of any other card. In fact, they can’t even access the full history of that very card that you are reviewing. This is why it’s important that FSRS and SM-2 have what’s called a Markov property - you don’t need the full history of the card each time you calculate the next interval.
