“Documents obtained by The Daily Tarheal revealed the use of Artificial Intelligence in the UNC admissions process. The Durham-based company Measurement Incorporated provides the technology for the essay review.”
Article says the AI was trained on essay scores given by humans to look for similar characteristics in the writing. To me, this suggests there could be a “formula” to follow to assure a high score. It is not clear to me if the score reflects quality of writing in the essay or quailty of personal characteristics revealed by the essay. The article says all essays are also read by humans (at least as of 2023) but seems like the goal of using AI would be to reduce the human time spent.
Even without any AI involvement, there could be a formula for writing an essay that gets scored highly. An infamous example was the writing section of the 2004-2015 SAT, where in the later part that period, some test prep companies apparently figured out or reverse engineered the scoring rubric and then taught their customers how to score highly on that section.
But colleges can change their essay scoring rubrics more frequently than the SAT, so there is less ability and utility to reverse engineering a specific college’s essay scoring rubric.
I note an alternative (although also non-mutually-exclusive) possibility is that there are “formulas” for writing BAD essays, meaning various practices that could get flagged as problematic for various reasons.
There are more details in this companion article:
As they explain there:
According to the 2022-23 admissions reader guidelines, the computer-generated essay score reflects the quality of word choice, sentence structure, sentence variability, vocabulary, grammar, punctuation, length and more in an applicant’s essay.
The reader guidelines explain that the essay score may give the reviewer an idea of the writing quality, but reading the entirety of the essays will give the reviewer “a more accurate sense of a student’s writing abilities.”
This is my interpretation, but it seems to me like a BAD score given those criteria could be a negative, but a GOOD score would just move you on to a human reviewer evaluating the actual content of the essay.
Of course you could try to reverse engineer this sort of thing to make sure your essay didn’t get a bad score on these criteria. But that would be less a “formula” for writing a good essay than just a more robust version of a traditional spelling/grammar checker. Which can help you avoid bad things, but they can’t actually tell you what content to use.
It’s ironic, and more than a bit hypocritical, that students are expected to not use AI, ChatGPT, etc. in essay writing, but here’s a university using that very same technology to evaluate students applying to their institution.
AI will look at grammar, sentence structure, vocabulary etc and give a score (1-4). This is always followed by human readers who are able to spend time looking at content and fit, etc. Every essay is read by humans.
Right now my feeling about this is neutral. Isn’t it possible now to check your grammar?