by Michelle Kukan (History Major)

Michelle Kukan’s research project, “Jane Addams Papers Project: Humanities and Artificial Intelligence” tested out two generative AI tools to see whether they could assist students workers in transcribing documents. About one-third of the documents in the Jane Addams Digital Edition are handwritten, and some of the authors, including Jane Addams, are difficult-to-read. Michelle investigated two questions: 1) Would AI be efficient enough to replace student workers? 2) Would AI be accurate enough for students to use it as a tool to help them transcribe documents. Michelle worked with Transkribus, an trainable tool designed to read historical documents, and with ChatGPT.

She uploaded and trained the Transkribus model using both typed and handwritten documents, had it transcribe the texts, and then corrected them, to further train the model. She found the training process time consuming and that due to our images (which were scanned from microfilm) and the poor handwriting, that the results were not as good as an average student transcriber. For typed and printed materials the results were far better. With ChatGPT4 no training was necessary and again, especially with typed materials it did a good job and was twice as fast as un-aided transcription, even when taking time to proofread the results. Michelle found that while AI transcription was not yet at the level needed for the project, it was a useful tool, especially for longer typed documents. Students could use the saved time focusing on the handwritten texts and conducting historical research. Her research was featured as part of Scholars Day at Ramapo College in 2025.