By Eli Ashcroft and Ian Zotter-Barlow
According to Dean of Admission Milyon Trulove, Reed College currently has no formal policy on the use of generative AI in admissions essays. “[Reed doesn’t] have a policy regarding AI use in the admission process,” Dean Trulove said, “[but] we expect that everyone will write their college essays independently of AI software.”
According to Greg Anderson, a professor in Reed’s Computer Science Department, artificial intelligence, “typically refers to the use of neural networks, which are mathematical objects inspired by the human brain.” Generative AI specifically, Anderson explained, “is based on large language models (LLMs), which are neural networks that have been trained by taking a block of text from the internet, feeding part of the text as input, and asking the network to predict the next word in the text.”
Early in the introduction of generative AI, some institutions responded by investing in AI detection tools. However, there’s no definitive way to actually tell the difference between AI generated text and human work. According to Anderson, “it is currently more or less impossible to automatically detect text generated by AI. By design, AI-generated text is statistically very similar to human-generated text, so automated tools relying on statistical analysis are going to have trouble distinguishing the two.”
In fact, artificial intelligence research company OpenAI discontinued a classifying tool which was supposed to detect whether or not text was written by artificial intelligence. After originally developing the tool, OpenAI discontinued it on July 20, 2023, citing a low accuracy rate. According to OpenAI’s website, the classifier, on a “challenge set” of English texts, correctly identified only 26% of the AI-generated content and incorrectly labeled human-generated content as artificial 9% of the time.
Beyond Reed, admissions officers across the country are also weighing how to adapt to the widespread availability of the new tools. One big reason for this is ChatGPT is free — or at least a version of it is. Not only that, but it’s intuitive and accessible: if you want it to do something, you just go to a website and write what you want it to do. That accessibility means a lot of people are using it. ChatGPT in particular has proven its popularity — from writing scientific papers to curating an art exhibition at Duke University’s Nasher Museum, large language models are being used for basically anything humans can imagine – including college applications.
Some colleges, like the Sandra Day O’Connor College of Law at Arizona State University, have introduced application policies that prohibit or limit applicants’ use of AI. Other schools, like Reed, are more ambivalent. Reed’s reasoning, much like that of the handful of Ivy-league colleges, like Harvard, that have commented on AI, is that it just isn’t that good. Trulove said, “At its very best, any AI essay used in the admission process will fall right in the middle of the pack in terms of quality. I consider a candidate’s use of AI less an issue of personal morals or motivation but rather an exercise in bad decision-making. Why would anyone leave their personal story and opportunity to impact the admission committee up to a digital presence?”