Our Case Studies
- Case #1 : Authorship Cheating: Helping to Confirm a Student’s Authorship
- Case #2 : Authorship Determination: Helping to Identify Source of Ananymous Hate Speech
- Case #3 : Authorship Determination: Helping with Reputation Management Issues
- White Paper: Linguistic Fingerprinting to Flag AI Authorship Cheating
CASE #3 : AUTHORSHIP DETERMINATION: HELPING WITH REPUTATION MANAGEMENT ISSUES
Jim was an arbitrator, family law. For the last 10 years he had been arbitrating divorce cases. He had seen many cases, many of which were not pretty. The kids were always the first to suffer. The finances – next on the list. It was so personal, that the formerly married couples would spend much of their savings in denying the other access to those funds. It just didn’t make sense. But this case was pretty high up there.
Before him was a case where a highly respectable doctor and her husband were going through a messy divorce. The judge hoped that Jim could make some sense, and save the 8 year old daughter from being exposed to a harsh legal proceeding.
As usual, the crux of the case was division of the financial assets, where on top was, what was until recently, a highly lucrative medical practice. But this is where things got interesting. After a string of negative reviews on webMD and similar web sites, the practice took a nose dive. The doctor claimed that the husband authored the negative reviews to damage her practice and cause her financial harm. She wanted full custody and was seeking financial restitution in the form of a large alimony check as well as financial damages.
The husband claimed that actually it was the wife who authored the reviews in order to place him in a negative light, for her benefit at the custody hearing.
The doctor submitted a forensic linguistic report in support of her brief. The report was 60 pages in length. It detailed the process they went through. It included analyzing 7,000 words from documents authored by the husband, and 27,000 words from documents authored by the doctor. The expert was thorough. They also analyzed documents authored by an unknown person, someone who regularly submitted complaints to the medical board, but who, it was established, was never a patient of the doctor.
The expert also analyzed 23 reviews, from patients who submitted negative reviews to the web site, on similar medical practices. The expert was looking whether the patients complained about similar things, was there something unique that recurred across the reviews, or were there consistencies across the reviews.
The report included feature charts and numeric information comparing the distribution of overlap of features across the different authors. The features included word patterns; word repetitiveness; grammar used; syntax; grammar rules followed; repetitiveness of typographical errors; punctuation used. The report concluded that indeed it was the husband that authored the offending reviews. Additional support was provided by an external source, a forensic computer analysis, that traced the IP address of several of the offending reviews to a computer owned by the brother of the husband.
The bill submitted for a claim for expenses included nearly 30 hours of analysis, including attorney, paralegal and expert report, amounting to over $ 15,000.
Jim, who had no prior knowledge of the field of forensic linguistics, sought out advice on how to evaluate the report, and what kind of credence to allocate to it. The expert he worked with, a founder of Flint AI, received all of the documents used in the report. The Flint report was sitting on Jim’s desk. The conclusion was identical. The report included much of the same detailed reasoning that was included in the 60 page expert report. Time and cost – less than a day and would have been under $50.
When confronted with the report and the evidence, the husband confessed.