Plaintiffs Keith Fischer, Michael O’Sullivan, John Moeser, Louis Pia, Thomas Barden, Constance Mangan, and Charise Jones, (collectively “Plaintiffs”), sought class action certification for their claims against Government Employees Insurance Company (“GEICO”) for failing to pay overtime wages in violation of the New York Labor Law (“NYLL”).
Plaintiffs and the Class Members are current and former non-exempt employees of GEICO in the Special Investigations Unit (“SIU”). Plaintiffs relied exclusively on a damages model developed by their proposed expert, Dr. Catherine O’Neil, to show that damages are capable of class-wide determination.
O’Neil described a model to measure lost wages by assigning an amount of time to each type of case-related activity (through a regression analysis of the Plaintiffs’ self-reported hours compared to case-related activities tracked in the GEICO SICM database), and then applying the assigned time-per-activity amounts to the case-related activity data for each Investigator to determine “actual” time worked by each.
O’Neil contended that she would then subtract a putative class member’s reported hours from the hours derived from the regression model to determine the unreported and uncompensated hours for each month.
Data Science Expert Witness
Dr. Catherine H. O’Neil earned a Ph.D. in math from Harvard and previously taught Mathematics at the Massachusetts Institute of Technology and Columbia College. She is a data scientist who founded an algorithmic auditing company.
In 2016 she wrote the book Weapons of Math Destruction: how big data increases inequality and threatens democracy. and in 2022 the book The Shame Machine: who profits in the new age of humiliation.
Discussion by the Court
Here, O’Neil contended that she can create a linear regression model to estimate the time investigators spent performing certain activities. But this model is nothing more, as she concedes, than a “thought experiment.” She has neither built the model nor applied it to the data available.
As GEICO pointed out, O’Neil had access to seven months’ worth of SICM data for 34 Plaintiffs and putative class members and Plaintiffs’ testimony as to their estimated hours.
By limiting her model to a “thought experiment” and failing to show the applicability of this model to even a subset of available data, the very data O’Neil contended she would use to build her model, O’Neil’s report has failed to show that her opinion is rooted in actual facts or data to properly assess class-wide damages; it amounts to a kind of “trust the expert” methodology. But such “ipse dixit” cannot satisfy Daubert.
Contrary to Plaintiffs’ presentation, this is not a kind of “plug and play” expert analysis, where simple math—here a basic linear regression model—is the core of the model. The inputs into that model, including the allocation of time for a task, are based on a series of assumptions, which O’Neil has not tested or explored in any meaningful way, even if it was not necessary to build the final complete version of her model.
For example, O’Neil assumed that certain investigator tasks will take a standard amount of time. However, the Court held that O’Neil has no known expertise or experience in doing the kind of investigations conducted by these kinds of employees, making merely accepting Plaintiffs’ accounts or her own uncredentialed assumptions problematic.
Held
The Court found Dr. Catherine O’Neil’s opinion unreliable for determining whether Plaintiffs have satisfied the requirements of Rule 23. As a result, the motion to certify a class was denied.
Key Takeaway
Plaintiffs have failed to show that O’Neil’s opinion “is the product of reliable principles and methods.” Rule 702‘s focus is “the scientific validity and thus the evidentiary relevance and reliability—of the principles that underlie a proposed submission. The focus, of course, must be solely on principles and methodology, not on the conclusions that they generate.”
Unsupported assertions “made without explanation or elaboration that would allow a fact finder to follow his reasoning and come to the same conclusion” are inadmissible expert opinions.
Case Details:
| Case Caption: | Fischer V. Government Employees Insurance Company |
| Docket Number: | 2:23cv2848 |
| Court Name: | United States District Court, New York Eastern |
| Order Date: | February 20, 2026 |
Leave a Reply