How risky are automobile driver assist systems (for example Tesla Full Self Driving)?

If one defines a relevant metric like deaths per passenger mile, and if one applies the metric while avoiding giving undue weight to things like “magnitude of headline splashing”, then much insight can be gained about relative safety or riskiness of various transportation choices.   And there are many reminders that lots of people are not very good at assessing risks when left to their own devices about it. 

So for example on every relevant metric, for intercity trips, travel by air or train or bus is safer than travel by automobile.  The riskiness of automobile travel is not even close — riskier by something like times-5 or maybe times-10.  The statistics-gathering process for this is mature and no trained people seriously doubt the process or the conclusions that are drawn from it.

But when an airplane crashes, this is all over the worldwide news and makes worldwide headlines.  The automobile deaths that would need to get added up to be compared appropriately to an airplane crash only make local news if at all.  So lots of people get the wrong answer on the question of whether it is safer to drive your own car than to travel by air or train or bus.

We can then try to apply a relevant metric to relative riskiness of various ways of operating a car:

    • driving a car while using old-fashioned cruise control that merely maintains a constant speed;
    • driving a car that has ABS (an anti-lock braking system);
    • driving a car while using modern “adaptive” cruise control that also keeps track of the behavior of the car in front of you;
    • driving a car while using “lane assist” which is a feature that warns you if you are deviating from lane boundaries;
    • driving a car using “blind spot warning” to protect against risky lane changes;
    • driving a car using foreward collision monitoring;
    • driving a car using hill-start assistance;
    • driving a car using Mercedes-Benz’s Drive Pilot;
    • driving a car using Volvo’s Pilot Assist;
    • driving a car using Tesla’s Autopilot;
    • driving a car using Tesla’s Full Self Driving;
    • driving a car using Nissan’s ProPILOT Assist;
    • driving a car using Ford’s BlueCruise;
    • driving a car using Ford’s CoPilot 360;  and
    • driving a car using General Motors’ Super Cruise.

A big challenge on evaluating riskiness for the “autopilot” features (the last eight or so items on this list) is that each of them is a moving target.  Software changes happen through over-the-air updates that the car owner does not necessarily notice.  A second challenge is that there just aren’t very many Volvo cars equipped with Pilot Assist (because it is a luxury optional feature) on the road, compared with (say) the number of Tesla cars that have Tesla Autopilot (which is all Tesla cars).  Because many of these systems came into production service only in very recent years, there is necessarily not as much history available as one would like to have.

But studies by third-party regulators like the National Highway Transportation Safety Administration suggest that no matter which one of the driver assistance systems is being studied, it always turns out that “driver assist plus alert driver” is safer than “driver alone”.  Often the measured safety difference is “not even close”, with the “driver alone” riskiness turning out to be three times worse or five times worse.

Note by the way that absolutely none of this analysis is linked in any way to whether the car is an ICE (internal combustion engine) car or an EV (electric vehicle).  Nearly all Ford BlueCruise and CoPilot 360 cars are ICE cars, for example.  Nearly all Volvo Pilot Assist cars are ICE cars.

I am fully convinced, from some tens of thousands of miles of Tesla cross-country driving, that “me plus Tesla Autopilot” is notably safer than “me alone driving”.  I am further fully convinced that “me plus Tesla Full Self Driving” is notably safer than “me plus Tesla Autopilot“.

But absolutely every time that any Tesla car ends up in a fatal auto accident, it is splashed all over the nation-wide news.  No such nation-wide attention is given to similar fatalities for Ford or Volvo or Mercedes-Benz cars that were using autopilot systems.  This leads to a situation where lots of people get the wrong answer when trying to assess the relative riskiness of various automobile autopilot systems.

A similar observation may be made about car fires.  Every time an electric vehicle has a battery fire, the event is reported in headlines nation-wide.  When an ICE vehicle burns up, this goes unreported or is only mentioned in local news.  Many people get the idea that an EV is much more dangerous than an ICE car in this respect.  Again a relevant metric permits dispassionate comparisons.  See for example Do electric cars pose a greater fire risk than petrol or diesel vehicles?  (The Guardian, November 20, 2023) which says that the probability of being caught in an EV fire appears overall to be much lower than for ICE cars.

Leave a Reply

Your email address will not be published. Required fields are marked *