Examining AI-generated historical narratives and their reception through the example of history POVs on TikTok
This paper investigates the TikTok "history POV" trend, where AI-generated first-person historical scenes are analyzed through a two-stage empirical study to reveal that while emotionally charged contemporary topics dominate, the choice of subject significantly influences audience reception, with Holocaust content specifically attracting disproportionately higher rates of hate speech and disinformation compared to other historical events like the Black Death.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine TikTok as a massive, bustling digital town square where young people gather to share stories. Recently, a new type of "storyteller" has arrived, armed with a magic wand called Generative AI. Instead of drawing pictures by hand or filming real actors, these creators use AI to instantly conjure up short, immersive videos that say, "POV: You wake up in the year 1348 as a peasant during the Black Death," or "POV: You are a soldier in World War II."
The paper by Nina Brolich and Anna Neovesky is like a pair of binoculars and a magnifying glass, used to look closely at this new phenomenon. They wanted to answer three main questions: What stories are being told? How do people react to them? And how can we study this without getting lost in the noise?
Here is a breakdown of their findings, using simple analogies:
1. The "Menu" of History: What's Being Served?
The researchers looked at over 5,500 of these AI videos (a "menu" of content) to see what flavors were most popular.
- The Taste: The most popular videos aren't about quiet, boring history lessons. They are like spicy, high-octane thrillers. The topics are almost always dramatic: wars, disasters, executions, and the "final days" of famous figures.
- The Time Travel: While you might expect videos about ancient cavemen or medieval knights, the "menu" is actually dominated by recent history (the 20th and 21st centuries). It's as if the town square is obsessed with the most recent, emotionally charged events rather than the distant past.
- The Recipe: These videos are made quickly (4–8 hours). The creators use AI to write a script, generate images, and animate them. They claim they aren't trying to be history textbooks; they just want to spark curiosity. However, the researchers noticed that the "labels" on these videos (the captions) often contain historical errors, like calling the "Black Death" the "Black Plague" or placing a caveman in 40 BC (a time when humans were already long gone).
2. The "Town Square" Reaction: How People Respond
To understand how the audience reacts, the researchers picked two very different "dishes" to compare:
- Dish A: The Black Death (a deadly plague from the Middle Ages).
- Dish B: The Holocaust (the genocide of Jews during WWII).
They acted like social scientists observing a crowd, reading thousands of comments to see what people were saying. They used a "sorting machine" (a computer program called DistilBERT) to categorize the comments into buckets like "Questions," "Jokes," "Facts," or "Hate."
The Findings:
- The Black Death Crowd: People mostly treated these videos like entertainment. The comments were filled with jokes, "wow" reactions, and general curiosity.
- The Holocaust Crowd: The reaction was completely different. Because this topic is deeply sensitive and painful, the comments were much more serious.
- More "Fact-Checkers": People were more likely to try to explain the history or correct the video.
- More "Toxic Spills": Tragically, this topic also attracted a disproportionate amount of hate speech and lies (disinformation). While the Black Death videos got a few rude comments, the Holocaust videos were flooded with them.
- The "AI" Factor: Interestingly, people noticed the AI aspect in both cases. About half the videos had the hashtag #AI, and people frequently commented on the fact that the images were generated by a machine, sometimes joking about it or questioning its authenticity.
3. The "Flashlight" Problem: How They Studied It
The researchers had to figure out how to study this fast-moving trend without getting overwhelmed. They tried two methods:
- Method 1: The "Manual Scoop" (Pilot Study): They manually found 210 videos and looked at them one by one. This is like walking through a market and picking up a few apples to inspect. It gave them a good initial idea but was slow and small.
- Method 2: The "High-Tech Net" (API Study): They used TikTok's official research tool (an API) to pull in 5,500 videos at once. This is like using a giant net to scoop up the whole market.
- The Catch: The net had holes. It couldn't catch videos from people under 18, and it couldn't grab the actual video files (just the text descriptions and stats). It also missed some data due to technical glitches.
- The Result: Despite the holes, the "High-Tech Net" confirmed what the "Manual Scoop" found, but with much more detail. It proved that while manual observation is good for a quick look, you need the big data tools to see the full pattern.
The Bottom Line
This paper tells us that AI is changing how young people experience history on TikTok. It turns history into a fast-paced, emotional, first-person movie. While this can be engaging, it often prioritizes shock value over accuracy.
Most importantly, the study shows that the topic matters. When the content is about a tragedy like the Holocaust, the internet reacts with a mix of serious correction and toxic hate. When it's about a plague in the Middle Ages, it's mostly treated as a fun, spooky game. The researchers conclude that while we can study these trends effectively using big data tools, we must be careful because the "videos" themselves are invisible to the data tools, meaning we can't easily analyze how the AI is visually distorting history, only what people are saying about it.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.