Admittedly, the data is now stale. A new comprehensive study is needed, and one that takes into account such things as crossing double yellow lines, Smart Summon being "nothing but a gimmick"(c)Elon, phantom braking, and other various problems that are routinely reported by FSD drivers right here within this forum
I'm not sure how one should even go about a cumulative score for a driver assist feature.
The biggest issue is Performance versus Safety.
The problem being is that safety actually does down as performance increases. The reason for the inverse relationship is the human behind the wheel can't possibly be expected to be engaged when a system works 99.9% of the time. When things always work people get used to it always working, and an unconscious trust forms.
There is probably a fine line between functionality, performance, and safety. Some point where there is enough functionality, and performance that drivers use it while also not causing so much trust in a system that they fail to pay attention.
CR/IIHS,etc all put a lot of faith on driving monitoring. But, that can only really combat purposeful inattention. The "I'm going use the driver aid as an excuse to text" kind of people.
Probably the most meaningful thing Supercruise does is they geofence it to only areas where its designed to work. Tesla doesn't do this so it performs wildly differently depending on where you're at.
But, that ability to always use it is useful to owners.
I personally can't wait to wipe my hands free of L2. They simply can't be made to perform well, and to also be safe.
One thing CR ignores is the hands free offered by supercruise does reduce safety because of increased take over time. Now I'm a huge advocate for hands free, but even I acknowledge that it reduces safety in that regard.
