[SystemSafety] Autonomous Vehicle Safety

[Prof. Dr. Peter Bernard Ladkin]

Well, I 've now read the paper (it is pretty short, just over 10pp long + references). 
Disappointingly, it mentions neither of its predecessors in aerospace, Butler/Finelli and 
Littlewood/Strigini, both almost a quarter of a century previous.

On 2025-12-17 16:34 , Phil Koopman wrote:
> Going by memory, the free RAND report has the same or similar material as the paywall article.
>
> This is a dated paper, and since then there has been much exploration of the limits to such a 
> calculation, which I will enumerate as threats to validity to a conclusion that such vehicles are 
> acceptably safe:

It is a collection of mathematical calculations, and the conclusions are: for the simple (crude) 
questions considered (fatalities resp. injuries per miles driven) it is not possible to show to any 
reasonable confidence level that autonomous vehicles are safer than human-driven vehicles, even 
assuming a realistic (for 2016) US baseline rate for human-driven vehicles.

Their conclusions are, as with Butler/Finelli and Littlewood/Strigini, that if you want to show 
autonomous vehicles are more safe than human-driven, you can't do it this way at this time (or at 
any reasonably-near point in the future).

I don't see anything dated about either the questions or the calculations or thereby the answers. 
One could observe that it is frequentist rather than Bayesian, but similar calculations are going to 
hold in a Bayesian framework.

The authors themselves point out the limits to such calculations, which is that they don't tell you 
much about the actual safety of autonomous vehicles. For the simple reason that it will take an 
absurd number of driving miles to reach a reliable conclusion to a reasonable confidence level using 
such calculations.

PBL

Prof. Dr. Peter Bernard Ladkin
Causalis Limited/Causalis IngenieurGmbH, Bielefeld, Germany
Tel: +49 (0)521 3 29 31 00