Virginia Tech will use AI to handle the second read of essays.
The new model replaces the previous system, in which each essay was initially scored by two human reviewers from the admissions team. Now, each essay will be reviewed by one human admissions team member and one AI reviewer. The AI model is designed to confirm, not replace, the human reader’s score.
No word as to whether Va Tech will use AI to detect if students used AI to write their essay. I assume Va Tech AOs don’t want students to use AI to write their essays. To students: we can use AI, but you can’t.
Also…The EA deadline will move from November 15 to Nov 1.
More here:
3 Likes
Not sure how schools manage to read all those essays, but I do know that when AI is used in job recruitment it prescreens out many exceptional candidates.
9 Likes
Many schools use temporary readers during the review season. These jobs pay $17-$30 per hour or so and are really competitive to get.
2 Likes
You’d be good at it, I’m sure. Reader jobs get posted on the Higher Ed jobs site as well as Indeed! I know UC Berkeley has some reader jobs posted now.
I’m definitely remembering this for my upcoming retirement.
4 Likes
AI tends to ‘lean toward the mean’ --and I wonder if it will make the school lose out on some exceptional candidates that don’t fit the traditional profile. Also having just one human reader for each will potentially introduce more unintentional bias into the reading. I guess it’s not something I would trust AI to do just yet -especially since we KNOW that AI has all the worst bias of the internet world built in.
3 Likes
I would expect the human readers go thru regular anti-bias training, but agree that bias can always insert itself into the process, whether human or AI.
It’s interesting that the linked announcement from Va Tech specifically talked about their in-house developed AI. A third reader (human) will get involved if there’s a large discrepancy between the human and AI reads (they also did this previously if there was a discrepancy between the scores of the two human reviewers.)
It’s all so interesting! I hope they ultimately publish their results on this.
How was the AI model developed?
The AI reviewer uses internal university resources, including a large language model that was trained, rigorously tested for accuracy and fairness, and validated by Virginia Tech researchers.
Who are the Virginia Tech researchers?
The research team is being led by Louis Hickman, a faculty member in the Department of Psychology. His research background focuses on the intersection of technology and work with an emphasis on applications of machine learning and artificial intelligence. He is working in close coordination with Juan Espinoza, vice provost of enrollment management, and Rob Hopkins II, assistant vice provost for analytics and AI in the Office of Analytics and Institutional Effectiveness.
1 Like
Thanks for the clarification - I didn’t see about the in-house developed AI. Maybe that will be better.
I mean I 100% understand why they are doing it. I’m just not sure AI is ready for such big responsibility! I mean these acceptances/denials change life paths (maybe for the better, and maybe for worse - but definitely changes life directions!)
1 Like