The Japanese Red Cross Society’s Tokyo chapter has canceled an exhibition on the 1923 Great Kanto Earthquake that featured survivors’ “testimonies” created using generative artificial intelligence, which were criticized as being misleading.

“We needed to give more careful consideration,” a chapter representative said. “We are deeply reflecting on (what occurred).”

The organizers had planned to display “new” testimonies from 100 people and portraits of 20 individuals that were produced based on written accounts, such as survivors’ testimonies, as well as an oil painting depicting relief activities during the earthquake.

Critics online questioned the validity of the AI-generated texts, saying they are works of fiction rather than actual accounts.

One person even said the exhibition was fabricating history.

The event, which was scheduled to run from Aug. 26 to Sept. 7 at two locations in Tokyo, was canceled on Aug. 24.

The magnitude-7.9 earthquake struck the Kanto region on Sept. 1, 1923, claiming 105,385 lives.

20230907-Testimony(p)-L
A woman carries a placard with the names of relatives she is searching for near the Imperial Palace after the Great Kanto Earthquake in 1923. (Asahi Shimbun file photo)

The organizers told The Asahi Shimbun that they had hoped the exhibition, produced with support from the latest technology, would make it easier for visitors to visualize the realities of the earthquake and learn from what transpired.

They said they had planned to add a disclaimer stating the displays were AI-based.

No experts were on hand to oversee the work producing testimonies and portraits with AI.

Tatsuhiko Yamamoto, a professor of constitutional law at Keio University, said the problem was using the word “testimonies,” which is normally associated with facts. 

He said the organizers probably wanted to gain sympathy from audiences by using AI, which is capable of drafting sentences that actual survivors would have spoken.

“But the sentences AI created are nothing but fiction,” said Yamamoto, who is familiar with the relationship between AI and legal ethics.

Takuya Onodera, an associate professor of modern German history at the Tokyo University of Foreign Studies, said the event’s objective to prevent history from falling into oblivion should not be dismissed.

But AI-generated accounts should have been fact-checked and placed in a historical context under the supervision of experts, he said.

Onodera said there are many testimonies from the time of the disaster that are not generally known and that those materials would have produced a satisfying exhibition.