While I think every player should at least have some basic knowledge of every position as they play more of the game, I agree that having players get randomly filled into roles they didn't want carries more negative consequences than Riot originally thought about. As was initially presented, the three factors Riot had posted for an ideal queue were a) timeliness (the shorter the queue time, the better), b) accuracy (i.e. getting the position you wanted), and c) fairness (you queue up with and against players with about the same amount of skill as you). In this respect, autofill is the devs' way of trading off a bit of accuracy in exchange for better timeliness. However, the consequence of that, beyond just the dissatisfaction of players not getting the role they wanted, is that it often lowers the quality of games, as players are filled out into roles they're unfamiliar with, and therefore much less skilled at, or given roles they dislike, which is likely to generate frustration and non-cooperation. Effectively, autofill generates quicker games, but those games are lower in quality, in addition to often being even more frustrating than they can generally be.
I also think part of the issue here is that Riot's way of going about shortening queues kind of approaches the problem at the wrong angle: the main reason why queues are so naturally long in most elos is because supporting is by far the least popular position, as we've all known for years now, despite being essential to the game. Beyond supports just naturally being less popular than more kill-intensive roles across all games, supporting in League can be particularly unpleasant due to how restrictive its gameplay is, and how supports tend to be easy targets for abuse and bullying. Instead of focusing major design work on making the support role more desirable (minor support work on items every major update doesn't really count), Riot's mostly tried to resolve the issue by forcing people to support, which not only fails to make the role more enjoyable, but ends up reinforcing the stigma that supporting is a chore in League, rather than a fun role in its own right.
Eventually, I think the only viable way to make queues fully functional will be to improve the quality of less popular roles until we reach about an equal split across all roles. A perfectly equal split might not be likely, but there needs to be at least something done to make supporting not nearly as unpopular as it currently is. In theory, the support role should be the one receiving the most love, the most privileges and the most attention simply to combat the natural stigma that comes with supporting, but in practice it is not only the least popular role in the game, and so by a wide margin, but also one that has been treated even by the developers as a second-class role. When the game needed vision to function, the burden was originally placed squarely on supports, who had the least gold income to work with in the first place. When support players complained about being constantly behind on both gold and experience, due to being forced to not farm and constantly roam alone to ward or counter-ward (despite being the game's least independent role), Riot's response so far has simply been to make that an official part of supporting, with mechanics being implemented right now based on that, instead of bringing up their progression to match that of other roles. Riot reached out to players a little while back and asked for pointers on how to improve supporting, but no major conclusion has emerged from that, and instead the underlying trend has been to allow mages to become viable "supports", which on its own is detrimental to the support class. There has been an ongoing problem with not only the way supports have been treated in the game, but with the mentality that has led to a lot of its design, and I think that if Riot truly wants their queues to be quicker and more satisfying, they need to stop thinking of supports as a slave role, and start bringing them up to much higher standards of enjoyment.