Truth Through Repetition

In articles like this one about edX, people say things like:

The edX project will include not only engineering courses, in which computer grading is relatively simple, but also humanities courses, in which essays might be graded through crowd-sourcing, or assessed with natural-language software.

And I'm thinking that this is not really entirely 100% true. Grading both math and humanities seems to be equally hard. Most online systems are pretty good at determining whether a student has acquired a "right answer" to a question that has a right answer. It is just as good at identifying a student's ability to complete a routine exercise as it is at assessing whether a student understands its vs. it's or there, their, and they're or put together a sentence that conforms to the rules of standard written English or whatever. I'm not aware of any online math homework systems that will tell students that they keep making mistakes because they are bad with negative signs or don't know the first thing about the distributive rule. Nor do I know of any that will look at the diagram that the student drew and recognize that the student used the wrong triangle when trying to set up the related rates problem with similar triangles. Grading is only simple when the questions are simple.