Professors and executives test AI twins trained on their work

FORMAT B - Trend

In short: More professors and business leaders are testing “AI twins” that can answer questions using a person’s past work, but most uses are still limited and closely controlled.

What's going on

Some universities are piloting AI “twins”, which are chatbots trained on a specific person’s materials, like lectures, emails, documents, and FAQs. The goal is simple, let students or staff ask routine questions at any time, like having an always-available office hours helper (like a teaching assistant that never sleeps).

EdScoop reported that several U.S. universities are using professor “twins” built with Anthropic’s Claude through a vendor called Praxis AI. Institutions mentioned include Clemson University, Alabama State University, and the American Indian Higher Education Consortium, which includes 38 tribal colleges. In those classes, Praxis AI’s CEO said about 75% of students used the tool, compared with 14% for older, basic chatbots, and some classes saw average grades rise by about one letter.

Separately, a company called Spheria markets “digital AI twins for teachers” and says individual faculty from Harvard Business School, the University of Virginia, and the University of Catalonia have used its product. This is not a Harvard-wide program, it is individual teachers using a tool. Researchers are also experimenting with “twins” of scientists that can talk to each other and then report possible collaboration ideas back to the humans.

What to watch

The big question is trust and responsibility. Research from Harvard Business School found people often rate a message as less helpful when they think it came from AI, even if the content is strong. For now, most “AI twins” are being used for low-stakes tasks like answering FAQs, not replacing people in important meetings.

Source: NYTimes