Discussed a week or so ago in the TUI Zante thread.
It's worth reading the whole paper, for anyone who hasn't - it's not that long.
Those calculated probabilities are based on a 2-hour US domestic flight, and factor in the probability that there is one or more infected passenger(s) on the flight. They also assume that all passengers (whether infected or not) are wearing masks.
It's a useful study, but it won't necessarily apply (a) in other parts of the world where the infection rate is higher/lower, and/or (b) to a longer- or shorter-duration flight.