By Rachel Dworkin, Archivist
In the last year, I’ve gotten more than a few requests from researchers to confirm something AI told them. In each case, much of what AI told them was wrong. I’ve also had people ask me for things ChatGPT assured them were in them were in our archives when they very much were not. While AI certainly has its uses, it’s not the best tool for genealogical or historical research.
It’s not surprising that people are turning to AI to help them in their research. The industry is pushing it as the hottest new technology. It has “intelligence” in its name which certainly gives the impression that it knows stuff. The problem is that most commercial AI is not a database. Rather, they are large language models (LLM). Their creators fed them large amounts of text scoured from books, newspapers, magazines, and the internet. Every time an AI chatbot answers a question, it is not retrieving information, but predicting what the likeliest answer is based on whichever texts it has been trained on. Because many historical and genealogical records have not been digitized, they are not part of commercial LLM training data. AI chatbots don’t have the information needed to predict the right answer.
As a human person, when I can’t answer a researcher’s question, I tell them “I don’t know” or “I don’t have that information” and then try to point them in a more useful direction if possible. Commercial AIs like ChatGPT, Claude, Gemini, and Grok don’t do that. They are programed to always give you an answer, even if it isn’t accurate. For example, when I asked ChatGPT how John W. Jones, local abolitionist and Underground Railroad conductor, died, it gave me the correct date (December 26, 1900) but claimed it was in a cave diving incident. AI chatbots are also trained to keep the user happy at all costs. When I pointed out that Jones’ cause of death was wrong, ChatGPT apologized, gushed about how right I was, and offered another, equally wrong, cause of death.
Me, a human person, doing research
So, if AI isn’t the best tool, what is? Actual historical records. These days, a good number of them are even available on-line. Here is a list of my favorite FREE on-line resources:
New York Heritage (https://nyheritage.org/) –FREE digital collections of material from across New York State like yearbooks, city directories, photographs, etc.
NYS Historic Newspapers (http://nyshistoricnewspapers.org) – FREE digital collection of 19th and 20th century newspapers from across New York State.
Internet Archive (https://archive.org/) – FREE digital collections of historic books, magazines, film, audio recordings, and archival material on a wide range of topics.
ProQuest (http://www.proquest.com) – digital collections of newspapers and scholarly publications available for FREE to Chemung County residents with your library card.
NOVELny (http://novelnewyork.org) – digital collections of scholarly publications, newspapers, and magazines available for FREE to New York residents with your local library card.
If you can’t find it on-line, you might have to visit a physical repository in person. For official records in Chemung County, that means visiting the Chemung County Clerk’s office for land records; the Chemung County Surrogate’s Court for wills and estates; the Chemung County Office of Vital Records for birth/death records; and various town clerks for marriage records. You might also want to visit the Steele Memorial Library or the Elmira College archives. And, of course, the Chemung County Historical Society. Our research library & archives are open from 1pm to 5pm, Monday through Friday. Admission is FREE for members and students, with a $5 fee for non-members.
Doing your own research, whether in-person or on-line, is harder and more time consuming than asking AI, but it is far more rewarding. Not only is it more likely to get you the correct answer to your question, it will often get you answers to questions you didn’t even know you should ask! Sometimes it won’t though. Sometimes you will have to learn to live with “I don’t know.”
