On this page
Asking an AI to become C. V. Raman and answer your questions is a fun way into a hard topic. You get a story, a voice and a reason to care about light scattering. The catch is that the AI is acting. It will tell you dates and details with the same confidence whether they are right or wrong. This post gives you a script to run the interview and a routine to check every claim, with a worked example.
Is an AI role-play of a historical scientist accurate?
Not reliably. In Why Language Models Hallucinate (Kalai and colleagues, 2025), the authors write that large language models "sometimes guess when uncertain, producing plausible yet incorrect statements instead of admitting uncertainty." They argue this happens because training and evaluation procedures "reward guessing over acknowledging uncertainty."
A role-play can make this harder to spot: you have asked the AI to stay in character, which may make it less likely to step out and say "I'm not sure." A made-up detail can sound just like a real memory.
The exercise is still useful: the interview gives you questions and a story, and the checking is where you learn.
How do you set up a scientist interview with AI?
Copy this script and change the name:
"Please play C. V. Raman in an interview with a Class 11 student. Stay in character, but if you are not sure of a date, name or number, say so in brackets."
"Tell me about your childhood and where you grew up."
"What did you discover, and on what date?"
"Who worked with you, and where did you publish it?"
"Explain your discovery the way you would to a student who knows basic optics."
"End by listing every factual claim you made as a numbered list."
The last line matters most. It turns a flowing story into a checklist you can verify.
How do you fact-check what the AI scientist said?
Use the method professional fact-checkers use. In a 2019 study by Sam Wineburg and Sarah McGrew, 45 experienced internet users (10 PhD historians, 10 professional fact-checkers and 25 Stanford undergraduates) evaluated live websites. The fact-checkers read laterally: they left a site after a quick scan and opened new browser tabs to judge whether the original site was credible. Compared with the other groups, they reached better-supported conclusions in a fraction of the time.
For your interview, that means:
Do not ask the same AI whether it was right. Open a different, reliable source.
Go to the primary record first: the Nobel Prize page, the original paper, or an official government page.
Mark each claim as confirmed, wrong, or not found. "Not found" is not the same as true.
Fix your notes, not just your memory.
A worked example: checking an AI's C. V. Raman
Here is what you would check from a typical Raman interview, and where to check it.
Birth. The Nobel Prize facts page for Raman gives his birth as 7 November 1888 in Tiruchirappalli, and his death as 21 November 1970 in Bangalore.
The prize. The same page says he won the 1930 Nobel Prize in Physics "for his work on the scattering of light and for the discovery of the effect named after him." It lists his affiliation at the time of the award as Calcutta University.
The discovery date. The Government of India's National Science Day page says Raman discovered the effect on 28 February 1928, and that India marks that date as National Science Day.
The paper and the co-author. The Crossref record for the Nature letter shows "A New Type of Secondary Radiation" by C. V. Raman and K. S. Krishnan, in Nature volume 121, March 1928.
Point 4 is an easy one to get wrong. A first-person story invites a solo hero, but the letter has two authors. If your AI's version left out K. S. Krishnan, or claimed the paper came out on 28 February, mark it wrong and write the correction down.
What should you do with the mistakes?
Keep a short "AI got it wrong" list at the back of your notebook. Each entry is the claim, the correction and the source.
Then go one level deeper than the history. Raman's prize was for the scattering of light, so make sure you can explain what scattering is. If that is the gap, the Fermi Concept Library has a lesson for every concept in physics, chemistry, biology, mathematics and computing.
If you want an AI that teaches rather than performs, Fermi's AI tutor works out what you already know, builds a plan for the session and checks understanding as it goes. Either way, the habit is the same: enjoy the interview, then check the record.
