New revelations about science fraudster Francesca Gino

And how Lawrence Lessig used AI to convince himself Gino is innocent

Back in 2023 I wrote about why Francesca Gino was almost certainly guilty of producing fraudulent data. That article was a response to her old buddy Lawrence Lessig attempting to defend her in a post on Medium.

Recently I found this delightful quote in The New Yorker from Lessig:

If you recall, in mid 2023 Data Colada published evidence of anomalous and improperly manipulated data that was associated with four major papers where Gino was PI. Before making their findings public, out of courtesy Data Colada had let Harvard conduct its own investigation. Harvard’s internal investigation resulted in a 1,200 page report which concluded that Gino, “committed research misconduct intentionally, knowingly, or recklessly.” After Data Colada published their findings, Harvard placed Gino on leave, citing the outcome of their investigation.

In response, Gino sued both Harvard and Data Colada for defamation. The case against Data Colada was eventually thrown out. Although Data Colada got hit with massive legal fees, the scientific community rallied around them with generous donations to their legal defense. Portions of Gino’s case against Harvard were eventually dismissed, but other parts are still being litigated. Gino later amended her lawsuit against Harvard to sue for discrimination as well.

On August 18, 2025, Harvard counter-sued Gino, arguing that she created a fake dataset in an attempt to exonerate herself and make Harvard's investigation look bad.

Back in 2023, Gino began to publicly claim that Harvard's investigation had been improperly conducted and ignored key files, and she published a spreadsheet file that she claimed exonerated herself in one of the cases. Harvard's countersuit argues that the file in question “was fabricated by Professor Gino and placed on her laptop” and that the University “suffered reputational damage and economic losses” as a result of Gino's allegedly false statements.

Harvard calls the fake file the “cover-up file”. A recent post by Data Colada breaks down the forensics that Harvard used. The original file was replaced with a new file with different data, and there is clear evidence that the the date was digitally tampered with to make it look like it was ten years old. The tampered date is a digital smoking gun that Gino (or possibly one of her lawyers) created the cover up file as fake evidence to try to exonerate Gino.

The countersuit also has some juicy evidence that makes the case for Gino being directly involved in the fraud even stronger.

Thus far, Gino has tried to blame multiple different RAs she had over the years for all the fabricated data.

But look at this timeline Harvard constructed using digital forensics:

Gino’s RA sent the original (unaltered) study data to Gino on Saturday, and then on Sunday Gino sent out the fabricated dataset. Then nine days later the RA sent the original data again! Julian Ackert, Harvard’s forensic analyst, writes “If [the RA] had participated in a modification of the raw data [that Gino sent coauthors] . . . it is implausible that she would continue using the raw data”.

Now, is it possible that Gino’s case is receiving more media attention and a harsher penalty because she’s a woman? Conceivably. But, I think the increased media attention may have to do with some of the jaw-dropping facts about this case:

• Her second book is entitled Rebel Talent: Why It Pays to Break the Rules at Work and in Life.

• Gino was a superstar researcher, commanding a $1,000,000-a-year salary from Harvard, and charging tens of thousands to speak at corporate events.

• Some of the suspicious data comes from a paper that had already been retracted because of fabricated data that was found in a different experiment in the same paper (the paper reports results from three separate experiments). It appears that two researchers (Dan Ariely and Francesca Gino) independently committed two separate acts of fraud in the same paper. What are the chances? What does this say about the frequency of fraud within behavioral science?

It may be true that Gino has received a harsher penalty than other fraudsters caught at Harvard, like George Serafeim. However, as Andrew Gelman points out, it doesn’t make sense to default to the most lax way anyone has been treated by Harvard in the past. I have long argued that firing researchers who commit multiple counts of fraud should become the new norm (with due process following a “preponderance of evidence” standard) .

How much does any of this matter?

This is a blog that normally talks about stuff that matters. This post was a little different, with information presented mostly for entertainment value.

However, I do think it’s important to make clear why Harvard’s firing of Gino was justified. Unfortunately, when strong evidence emerges that a scientist committed fraud, them ultimately having to face serious consequences is rare. Among the six notorious Alzheimer’s fraudsters I discussed previously, only two have faced serious consequences (Eliezer Masliah and Berislav Zlokovic). The other four are still gainfully employed as faculty and able to run labs.

Postscript - Lessig has gone off the deep end with AI in a bizarre attempt to defend Gino

Coincidentally, Kelsey Piper just published an article about what Gino’s friend Lawrence Lessig has been up to the last year. Bizarrely, Lessig has been producing a podcast with the help of AI that attempts to exonerate Gino. (Lessig, a Harvard Law School faculty, has been doing pro bono legal work for Gino and says he convinced billionaire Bill Ackman to help fund Gino’s legal defense).

To help with the research for his podcast, Lessig used ChatGPT, Claude, and Grok.

Then recently, he fed his podcast transcripts into an AI and asked it to create a website based on his podcast, which he published at https://theginocase.info.

In a recent Substack article, Lessig says he was “blown away” by the AI’s output and how well it constructed a defense of Gino.

This example shows the ways that AI can lead people in epistemically bad directions. A detailed understanding of how Harvard conducted their investigation and Harvard’s case against Gino requires reading a 1,200 page report as well as many more recent evidentiary filings. But if you tell AI to “read these documents and construct a defense of Gino” it will do just that - cherry picking out the pieces that make Harvard’s investigation look shoddy, and ignoring all the strong pieces of independent evidence that all point towards Gino being guilty.

Here’s Kelsey’s full piece, if you want to read it:

Yes, you can trick AI into exonerating you