Investigator’s Corner

[Dave English]

Hello:

Looks like you’ve spent a lot of time building a framework here. Let’s see if we can bring it to life.

I recently had a fun piece of data analysis published, and am interested in other people’s ideas for extending the research.

“Reliability of Eyewitness Reports to a Major Aviation Accident.” The (free) PDF is online at http://commons.erau.edu/ijaaa/vol1/iss4/9/

It’s not news that eyewitness reports have low reliability and validity. But maybe it is news that you can demonstrate that for a wide-body airliner falling out of the sky on a clear day.

Thoughts? Ideas?

Formal Abstract
There is a paucity of studies on the reliability of eyewitness reports to aviation crashes. We examine witness statements to a widely observed major airline accident to determine if reported accident investigator distrust of details in eyewitness reports is supported by empirical evidence. The extensive archival witness record (N > 300) of a wide-body airliner crash in clear daylight conditions is subjected to statistical analysis to test eyewitness reliability. Even with over 200 witnesses within a three square kilometre (1.6 square mile) area answering a binary observation question, the variance is sometimes high enough to preclude forming statistically significant conclusions about a basic large-scale event. There is no statistical difference between the distributions of witnesses that reported diametrically opposed observations; and no discernible geographic pattern, preferred orientation or optimum viewing distance. These results support the current reported practice of placing low value to eyewitness accounts of aircraft crashes.

[Ian Maione]

It would be quite interesting to be able to quantify in more detail to what degree (or if) pilots perceptions are any more accurate than those without any type of aviation experience (i.e. would a subset of observers who have some degree of aviation expertise show a smaller variance in observed characteristics of an incident, compared to a subset of observers who do not). You mentioned some anecdotal cases where pilots were unable to correctly identify certain aspects of an incident, but such observers are much rarer most of the time. I wonder if incidents that occur near airports would be more likely to have more such witnesses to form a better sample.

The other factor I wonder about with respect to the AA587 incident is whether there would be any correlation between variability of reports and the amount of time between the event and when the interview was conducted. One intuitively would think that the longer the time, the greater the amount of potential contamination from various sources. That you would imagine would be that much more the case for a case like AA587, where the amount of media attention and the high profile nature of the event probably couldn’t be higher.

[Author]

Posts