STEM at GFA: Grace Owen’s Project
Oliver Williams ‘27
Last year, I was reading the write-ups GFA did for symposium, and to be honest, I wasn’t learning much. While trying to find interesting presentations to go to, I was getting steamrolled by titles that I didn’t understand at all. I had no clue what (and I quote) “Drosophila Melanogaster As a Potential Model For Host-Microbe Interactions in IBD: Comparing the Efficacy of Different Probiotic Species on Dextran Sodium Sulfate-Induced Colitis” (GFA symposium schedule, title by Viola Cullen) means. That struck me as a shame- I’m sure Vi did a lot of really interesting work, and I’m really sure that the project was interesting, but frankly, I had no clue what that work actually was. While reading the titles, I wished the people doing the projects could be standing next to me, explaining the projects. Because of that, I am quite excited to launch the first-ever Symposium write-up, a column interviewing students about their projects, to hopefully shed some more light on the incredible work that’s being done.
While I was trying to figure out who to interview for the first write-up, I heard from Mr. Lowenstein that Grace Owen would be presenting her work in her advanced inquiry in computer science. Due to this, I decided she would be my first interview for this column. My first question was, of course, what actually is your project? Grace is building a spam detection website that uses a neural network to do the detection. You input an email or text, and the website outputs 2 things: a percentage chance that the message is spam, and a short paragraph explaining why it thinks whether the message is spam or not. The way it does this is by parsing the message for keywords that show urgency, sow doubt, or discuss money. Then that gets inputted into the neural network, and that network decides whether or not the message is spam. It does this by comparing the message as a whole and the parsed text to its training data of spam and non-spam messages, and then it outputs a percentage chance that the message is spam. Grace is currently working on training a Llama model (an AI model owned by Meta) to get the website to output a paragraph explaining why it thinks the message is spam or safe, and is working on the prompt engineering to get this feature into development.
I thought this project was really cool and asked Grace where she got the idea. Her response was much more personal than I would’ve guessed - her great aunt got scammed out of a very large sum of money by one of these spam emails. She had early-stage dementia, and because she gave the money away “willingly”, it was impossible to get it back. Grace said she felt that older generations had been somewhat left behind by technological advances and that she felt that a tool like this could be very helpful in that regard.
Grace says she has quite enjoyed her work this year in the CS department, and is excited to show it off in May! If you thought this was interesting, come see Grace’s presentation at symposium!
