Every time I ask why they don't use something
You quoted something that wasn't my quote somehow. But basically I agree with
@novox77.
then why don't they use the driver monitoring camera in this score?
The current parameters they track with the camera are likely not consistent enough, and don't work well enough in varied conditions to be fair or predictive to use in the score, unfortunately. There's also the issue about them not having the comprehensive dataset from those cameras to fit to the PCF.
All of this drama and discussion about the Safety Score and it using the "right" parameters and having the fitting data properly "cleansed." There are many issues with the Score, but the biggest issue with it in my view as currently implemented is that it allows "cheating." I think that every now and again someone is going to make a mistake and slam on the brakes - this is fine (it's recoverable!), but it would be great if they couldn't just reset their way out of it. For now, everyone can reset, and it is what it is; I assume many people with 100 scores have done it, but I hope that most people have a minimal number of resets, and yes, it limits their mileage (but this is apparently not going to be a selection factor so it's not a penalty). The game is the game and resets are allowed for now. But would be great to see this loophole closed going forward, especially when it comes to remaining in the FSD Beta evaluation group, since it most definitely skews the accuracy of the Safety Score, and allows unsafe driving to occur without penalty.
It's important to know when people have FCWs, whether they be "phantom" or not, and when they have had AP disengagements from going over 90mph (or whatever the reason). FCWs definitely are often an indicator of aggressive driving. It makes sense that they be counted, even if some of them are false. It's fine; we're looking for correlations here, not perfect predictive capability. Lots of FCWs are correlated with running up on people from behind who are turning (even if you see them), drivers predicting other driver behavior and assuming people will move from their path before they reach the other vehicle's current location, driving in traffic where other drivers are slamming on their brakes, driving on curving streets with many parked vehicles, driving on roads with blind hills, etc.
It's a very imperfect measure. But it doesn't have to be perfect.