Anne Frank betrayer possibly found. How did AI work in this case?

Mods, feel free to move to another category if this Q isn’t factual enough

After 5 years of work, a cold case investigation has come up with a name that they think is backed by strong circumstantial evidence:
“Arnold van den Bergh was a prominent Jewish businessman with a wife and kids in Amsterdam. After the invasion, he served on the Jewish council, a body the Nazis set up, nefariously, to carry out their policies within the Jewish community. In exchange for doing the Nazis’ bidding, members might be spared the gas chambers.” Supposedly he gave addresses of homes where Jews might be hiding to the Nazis in exchange for his own safety. Long story here: Investigating who betrayed Anne Frank and her family to the Nazis - 60 Minutes - CBS News
Shorter versions here: Anne Frank may have been betrayed by Jewish notary | Anne Frank | The Guardian and https://www.timesofisrael.com/anne-frank-may-have-been-betrayed-by-jewish-notary-new-book-claims/

The Anne Frank museum has a favorable response: it “was not involved in the cold case investigation, but it did share its archives and museum with the team, as well as its own 2016 investigation into the arrest of the people in hiding.
The Anne Frank House is impressed by the work that the cold case team has carried out. The investigation was carefully set up and performed, and the book The Betrayal of Anne Frank reports on it in an engaging and readable way.” Statement: Anne Frank House and cold case investigation | Anne Frank House
Interestingly, Arnold van den Bergh was not on the list of suspects that the Anne Frank House has had on its website for years: Was Anne Frank betrayed? | Anne Frank House

I’m curious as to how AI was more effective than standard detective work in this case? From the long CBS article (1st link above) “The team fed every morsel they could—letters, maps, photos, even whole books—into the artificial intelligence database, developed specifically for the project. Then they let machine learning do its thing.”

One clue the AI may have picked up on, was a note sent anonymously to Anne Frank’s father after the war, saying that van den Bergh was the one who betrayed him.

AI simply allowed the analysis to consider all the evidence about every possible suspect. Where were they at the time, what happened to them, are they mentioned in any official documents, did they get any special treatment and when, etc. AI simply allowed the searchers to quickly cross-examine any and all documents for traces of each suspect candidate.

One clue that van den Bergh is likely responsible was the special treatment he got compared to most other Jewish victims. He even got himself and family declared “not Jewish” for a short while. Then was not sent to the camps. Theory is he gave the list of known Jewish hideouts (not people) to the Germans, in return for being overlooked.

The story I seem to remember reading was that one of the Franks had accidentally knocked something off a table one evening and caused a clatter, and that an office worker who was working late heard it and reported that he thought there was someone living in the attic.

Is that off the table now, or was that just apocryphal in the first place?

That’s some fine AI work, right there. There’s no way you’d ever be able to make anything of that, without AI!

I’m surprised they didn’t need a psychic.

ETA: perhaps it was a psychic who wrote the note?

AI is used very broadly these days. I expect “all” it did was do a fuzzy search through the available evidence. It’s not like they taught a computer to solve crimes, they programmed a search algorithm to take a bunch of different factors into consideration and return the best clusters of evidence from their custom database.

They considered the people associated with the front building, but none were plausible for assorted reasons. they certainly went through the list of “usual suspects” and instead found that Bergh was the most likely.

There is a book about this investigation by Ms. Sullivan - perhaps it has a lot more detail.

In an interview, she mentioned that Mr. Frank did not provide the note to the first investigation in the 1950’s, but then presented it to the 1963 investigation. Presumption was that in the 1950’s he did not want the bad publicity of fingers pointing to another Jew as the culprit - fearing it might excite the antisemitic crowd to say “see, they all turn on each other like rats”.

Her best explanation for Bergh’s behaviour was desperation - he was part of the council that determined who went to the camps. Near the end of the war everyone was marked, and he did what he could to save himself and especially his family.

From the very brief description in the first link, I’m assuming this was some sort of model built on top of a graph database. For those who don’t want to follow links, graph databases are pretty similar to the old detective tropes where evidence is pinned to a board with string between them to show relationships.

Once your data is in such a format, there are many different possible techniques used to gain insights. Some simple examples:

https://neo4j.com/docs/graph-data-science/current/algorithms/ml-models/linkprediction/

They were interviewing the lady who wrote the book. I presume the book release is the reason for all the news stories in recent days.

Without knowing more about how they did it, this was my thought too. The AI could tie together a lot of disparate data about a lot of different people to help narrow the suspect list. The card that mentioned van den Bergh as the betrayer but that by itself may not have been convincing. There may have been other anonymous accusations of other people too. But the AI database helped to link van den Bergh to the case with other facts, such as his proximity to the Franks, his work with Nazis on the council, his special treatment by the Nazis for a time, etc. The AI probably created similar files van den Bergh and many other suspects and investigators could look at all the information in one place, draw some inferences, and come up the person they thought was the most likely suspect.

Even the lead investigator admits this is a lot of circumstantial evidence and would not be sufficient for a conviction but it’s perhaps the best we can do unless more evidence comes up.

Ironically, what that AI is doing is what’s fondly remembered as “Good old-fashioned police work” Fondly…unless you were a cop working with the huge task forces required to do it.

It’s remarkably similar to the batcomputer in the old 1966 Batman TV show.

Perhaps you missed the part where the AI had to cross reference the name with a second list. Does a name automatically give you an address? Hmm??

The NY Time had a skeptical article last night, paywalled so I’ll quote:

Scholars Doubt New Theory on Anne Frank’s Betrayal

A new book that names a Jewish notary as a suspect made headlines. Now that World War II and Holocaust experts have had time to review its claims, many doubt the methods and conclusion.
. . .
Pieter van Twisk, a Dutch media producer, was sure that modern crime-solving technologies, like artificial intelligence, big-data analysis and DNA testing could arrive at better conclusions than previous investigations. . . .
David Barnouw, the author of the 2003 book “Who Betrayed Anne Frank?,” said he had also considered van den Bergh as a suspect but ruled him out because, beyond the note, there was no evidence that he had played a role.

Barnouw was among several Dutch historians who voiced concerns about the findings of van Twisk’s team. . . .

Emile Schrijver, the director of Amsterdam’s Jewish Cultural Quarter, said he got an advance copy of the book late last week. “The evidence is far too thin to accuse someone,” he said. “This is an enormous accusation that they made using a load of assumptions, but it’s really based on nothing more than a snippet of information.”

By Tuesday, after experts on Anne Frank, World War II and the Holocaust had a chance to consider the findings, many said they doubted the book’s theory.

“They came up with new information that needs to be investigated further, but there’s absolutely no basis for a conclusion,” said Ronald Leopold, the Anne Frank House’s executive director. He added that the museum would not be presenting the findings as fact, but perhaps as one of several theories . . . .
But there is no evidence that the council had any such list [of Jews in hiding], said Laurien Vastenhout, a researcher at the NIOD Institute for War, Holocaust and Genocide Studies, who is an expert on the Amsterdam Jewish Council’s history. “Why would the people in hiding provide the Jewish Council with their addresses?” Vastenhout said."

For those with NYT access, link is https://www.nytimes.com/2022/01/18/books/anne-frank-betrayal-arnold-van-den-bergh.html

I appreciate that the AI may have done something.

However the sentence in question “One clue the AI may have picked up on, was a note sent anonymously to Anne Frank’s father after the war, saying that van den Bergh was the one who betrayed him” was - no doubt unintentionally - rather funny because on its own it seems weaker than weak sauce.

It might have started as weak sauce. But, this is AI - high tech stuff. So I’m guessing it didn’t just report the name that was written down, but was also so sophisticated it googled the name to come up with a possible address, too. So advanced.

The accounts I’ve read said the police went directly to the bookcase and opened the hidden door. They seemed to know exactly where to look.

I would expect that someone randomly giving out addresses (of Jewish families) wouldn’t have that information. It would have required searching the building.

My suspicion is that the whole “AI” angle is basically just a publicity stunt. Don’t forget this whole story is about selling a book. I don’t doubt they used AI for something but I strongly suspect it’s just an attempt to add a “wow” factor and a talking point to their press releases and marketing material.

He was on the cities “Jewish Council”. The nazis required the city to form such a council.
Now you can’t actually have a council if you send the members to the gas chamber, so wouldn’t it be that most council members and their families were relatively safe ???

I very much doubt it. The Nazis would just wait till you had betrayed everyone else, then send you and your family off to the gas chambers. The “final” in “final” solution was not intended to have any exceptions, I don’t think.

In an article and interview about this book - it was mentioned that after the council was dissolved and he was in danger of being deported to the death camps - he managed to get himself and family designated no longer Jewish. This suggests some real connections or extremely favourable treatment from the local Nazi hierarchy. When the rival who had taken over his business (Jews were not allowed to have certain businesses) complained, apparently his Jewish status was reinstated, but then he and his family never did get deported.

So yes, special treatment. Other ex-members of the council did get deported near those final days. After D-Day and the writing on the wall for the Nazi regime, presumably nobody was safe.