A random-sample Congress might work, but there's two major problems: -The expert problem: Running a government is not easy. There's a whole slew of problems (including the organization and nuances of the bureaucracy, for instance) that require extensive knowledge to make informed decisions about. "Professional" politicians alleviate this somewhat, since they can (ideally) devote all their time to learning about these issues, over the course of several terms if necessary. (That's at least one good reason why freshmen members of Congress usually aren't committee chairs.) Without the benefit of experience, the de jure decision makers would have to rely (even more than they do now) on lobbyists, meaning government would be even more in the hands of those with the most money. Longer terms or more stringent selection would help with this, but then that runs into the second problem. -The civil rights problem: Random selection may work for composing juries, but running a government is a full-time job that would take several years to become acquainted with (see above). Not only are people going to be rather unhappy to be pulled from their lives to do something they may not be suited for, but from a purely economic standpoint, you're removing productive members of society from their places. (Imagine if, say, Steve Jobs was chosen - suddenly a major corporation has lost the leader it has been taking a significant amount of its direction from.) Exemptions based on various circumstances might help, but that would at the same time result in a lower quality of legislator.

To be fair, this is part of how the original Athenian democracy worked, which functioned well enough. Perhaps in an entirely new government, where society would then grow based around the expectation of being randomly selected as a legislator, it might work, but I can't see this functioning in the U.S. system without major concurrent overhauls.

There are a few other cases I can think of offhand where careful reading is important. Decoding poorly written instructions and this card game in particular came to mind immediately, but on reflection I think the best example in my experience would be coding, and in particular debugging. I'm by no means an expert (or necessarily a representative sample), but at least half of the bugs I spend 10+ minutes chasing around end up being caused by a minor brain-dead error in a single expression.

To me, the simplest solution that comes to mind is to grade on a curve at the end of the course, based on the quality of the work (or some other subjective measure - I'm not familiar with university art courses, but I assume there's some kind of widely-accepted grading methodology). That is to say - tell the students before the course they will be graded on quality or quantity depending on the group, but grade each student on a curve relative to their own section for their permanent grade once the course has finished. This would obviously still require lying to the students (at the very least by omission, arguably), however.