I'm personally opposed to the idea of computers assessing writing for a number of reasons... but I don't think it is helpful to play the "we're essentially human and computers can't get that" argument. Mostly because if that argument still holds water, it probably won't for very long. It's an argument that's bound to end up on the losing side of technological history.
Let's face it, if a human activity is based on any sort of pattern or set of conventions, sooner or later there will be an algorithm that can anticipate and reproduce it. Those who discuss the nuances of writing and the beautiful complexity of human thought that writing encompasses are playing a waiting game. Technology is only getting better, and that argument is not only doomed to historical failure, it completely misses the real issue at hand here.
Humans (and especially students) are masters at adapting to their environments. Language use is no different. It doesn't take very long for most students to work out what a teacher wants from them, what they have to do to gain approval, and what they can get away with. In short, school writing, like any kind of language use, is a social activity that is always in a process of adaptation. Users of language are always "adapting to others' adaptations". I can remember plenty of times that I wrote to teachers or professors in order to project a certain image of myself, by referring to a random comment they may have made in passing, trying to establish a common point of interest, political perspective, or world-view... In short, writing is about making a connection with a reader.
I'm sure this sounds like I'm slipping into the "but we're so wonderfully human" argument... I'm not.
As we human-folk are always in a process of adapting to our environments, I'd like to consider what it would mean to ask students to write their school essays to a computer. What kind of writing can we expect from a student who is (very wisely) adapting to the expectations of a machine that is judging her work? Gone is the opportunity to 'impress' your reader, to coax the reader into or out of a particular way of thinking, of projecting an image to the reader, of surprising the reader.... I think that we might expect this student to write more efficiently, in a more standard format, to take fewer risks, and to generally do no more or no less than complete the required task.
In a fabulous short essay called Why I won't be Using Rubrics to Respond to Students Writing, Maja Wilson demonstrates how teachers have the capacity to become machine-like in their grading and how there is constant pressure for us to do so.
One might suggest that the more standardized and quantified our responses to our students become, the more feasible it becomes that a computer could do our work. The task is, then, not a matter of exposing the shortcomings of current technology but in addressing the ways that our jobs as teachers have already been reduced to work that could be done by a machine. The issue has nothing to do with computers taking the place of humans but of questioning the context in which such a thing could even be conceived.
I remember co-teaching in a high school English class. The full-time teacher and I decided to put together a lesson on juxtaposition. We asked students to bring in a photograph and to select a song. During the class, students swapped songs and pictures. While listening to the songs, we asked them to write their thoughts- "what does this mean?" / "how does the song relate to the picture?" and so on. Before we started, we got the obligatory "how long should it be?" / "what exactly do you want us to write?" / "is this for a grade?"
What struck me, both before and during the activity, was that we were envisioning the students writing for the purpose of catharsis or expression. We were hoping for students to communicate a message through the song/picture juxtaposition and for those listening to not only 'get' the message, but to produce a creative expression of how the juxtapositions impacted them.
What we got were short sentences that were obviously written only in order to complete the assignment.
My point is not that the lesson flopped (though it clearly did). My point is that writing... school writing, is something that we have been in a process of reducing to its lowest common denominator for years. As we make writing into a 'school' activity, we subject our students' ideas, thoughts, and expressions to a numerical grid. Now that computer programs are being written which are just as efficient, consistent, and accurate (perhaps) at quantifying written thoughts, teachers are calling on the "all too human" card. We (not everyone mind you, but public education as an institution) have been acting as imperfect computers for years. It strikes me as ironic that we are now in an uproar over the very natural progression to computers acting as imperfect humans.
I hope that the fact that computers can (or soon will) be able to quantify writing as effectively as humans might serve as a wake-up call. My hope is that this exposes not the imperfections of computers but the long evolving process of reducing human experience, activity, and expression to quantified measurements. The current debate strikes me as little more than humans acting as imperfect computers complaining about computers acting as imperfect humans. All the while, we are leaving unquestioned the dehumanizing effects of an education system that has given birth to the idea that the work of teachers could be reduced to the calculations of automated machines.