Sorry but I have to disagree! The Analytical Writing Section is scored by two readers, and the first reader is a computer program called Intellimetrix. If these two scores differ by more than one point, a third (human) reader is called in to settle the discrepancy.
The computer is not brought in to score a given prompt until answers to that prompt have been read and scored by human readers many times. The details of how the computer program works are proprietary and it seems that the company that developed them is very concerned that they not get out, so all we outsiders know is some general ideas of how it works.
The program parses each sentence, identifying key elements of sentence structure, and measures how sophisticated the vocabulary used is. Specifically, it searches for:
(1) “surface features” such as the number of words and commas, and the average lengths of words and sentences as well as the frequency of specific types of words, like verbs and articles, and
(2) “content features,” which includes specific words or phrases and their frequency.
So this is all very general, and the company that makes these things really wants to keep it that way. For more, here’s a website about the Bayesian Essay Testing Scoring sYstem (BETSY).