In a second statement after I pressed Lyft harder on the metadata issue, Mr. Brous said, “Lyft reviews and investigates damage claims using a variety of signals to determine if photos are legitimate, which can include metadata. If an image does not have metadata included, we do not automatically assume it’s fraudulent and additional signals are reviewed.” (He also noted that Lyft was rolling out a new fraud review tool this summer.)
Even without the metadata requirement, the company seemingly ignored that the driver did not meet its published instructions for driver damage reports, which require “at least three clear, well-lit photos of the damage from different angles.” There are three photos, and the lighting is good enough — if ruining my appetite is a reasonable criterion. But all three photos are from almost exactly the same angle, just cropped differently. And they were taken in a way that shows very few features of the car aside from what appears to be a rear bucket seat and a few design details.
So you asked A.I. to analyze the photos and shared those results with Lyft. The chatbot determined the vehicle was most likely a Jeep Grand Cherokee and noted that key BMW design details were absent from the instruments and door panel. My human comparison of the photos with online images of BMW X3 back seats reached the same conclusion, although it’s impossible to be sure.
Monica Osborne, a spokeswoman for Lyft, finally called me to admit that the company had made an error. “Based on our initial assessment at the time, we did think it was a BMW,” she said. But after “a more thorough review, we definitely agree that the vehicle does not match the specifications of the BMW.”
She said the company has now provided you with an additional $100 credit, recognizing your “extremely negative experience end to end.” Lyft drivers are independent contractors, not employees, but Ms. Osborne said the company would have “deactivated” the driver if he had not previously been deactivated for another reason.