AI Skepticism - History Wrongs

An early mention of artificial intelligence in the The Evening Star, Sept. 22nd, 1961. This was part of a classified ad from RCA Data Systems Division.

(Cross posted on the Teaching with Primary Sources Teachers Network)

I keep a list of ‘history wrongs’ that include stories like this:

• Black students were cast as slaves in New York teacher's mock 'auctions,' state finds

• D.C. Third Graders Reportedly Told To Reenact Graphic Scenes From The Holocaust

• Teacher Placed on Leave Over ‘Mock Slave Auction’ in Classroom

I am so so so very skeptical of role playing activities in general because they are often handled without the deep reflection and consideration needed. More often than not, I see instances  that place students in roles to be traumatized or traumatize others - which these above mentioned activities most certainly do. 

So, last week when TeacherServer (feel free to look it up on your own) was shared on bluesky, I decided to kick the tires and see what resulted. Two of the first options were Historical Event Perspective Generator and Historical Figure Interview Simulator. Oof. I suspected my investigation was not going to go well - at all. For the Historical Event Perspective Generator I asked for 7th grade, Civil War, 1860-1865, Perspective of an enslaved person. Here is an excerpt of what it produced:

The historical inaccuracy jumped out at me and took me out of the whole thing. The rest of it was *fine* but in my estimation wholly unnecessary. What is this providing that a primary source doesn’t? How about a vetted secondary source? Why would we use a silly amount of computing resources to produce what already exists in a more factually accurate form?

Then I moved onto the Historical Figure Interview Simulator and went with the most potentially troubling person - Hitler. I’ll spare you the result - as I find it wildly inappropriate ‘creative’ exercise. I will share that I’ve tried to report my concerns with this output through the site and both times I’ve received this response.

When you come across these AI based tools, please be extremely skeptical of their use in history. While there may be some valuable tools in there, simulators like those above have the potential for harm. Additionally, the outputs in this particular tool are not all that interesting, but rather a march through factual’ish information in a narrative or interview fashion. Far more accurate and potentially instructive sources exist throughout the primary source landscape.