From my recollection in general what happens is that there's a certain ratio in which low performers and high performers interact. If the ratio is stacked towards high performers in general the low performers end up doing significantly better than they would've if they were in a group consisting of people entirely like them.
This would suggest that societally there's some ratio that's optimal such that we intentionally put in "low performers" as a mechanism to raise their performance. As long as this is done very carefully it simultaneously increases the performance of low performers and maintains the potential of high performers.
Conversely if you have too many low performers and put in a high performer, not only is their potential not met, but it's generally detrimental.
In other words there's a sort of "force" pulling everyone to the mean. The "dilution" that results from adding high or low performers to an otherwise low or high performing group depends on how high or low they are and the group itself. With data this can be optimized such that low performers exceed their potential. Given the rate of change for low performers vastly exceeds high performers inherently, this is the societally optimal outcome.
That being said K-12 education isn't like higher ed is that it's lower level (intellectually).
