Re: Tesla never said it's driverless
"What Tesla has done here, deliberately reducing cognitive load below safe sustainable levels, would have the researchers that pioneered the fields of cognitive ergonomics and studied human factors as they relate to machine control spinning in their graves."
This thesis is wrong. "Deliberate" is your subjective assertion, which you cannot substantiate anyway. However, they have done opposite of the factual aspect of the claim. The car I bought 4 years ago had a VASTLY different Autopilot in terms of capability and cognitive load, than what is the case now.
The autopilot software is updated over the air several times a year, when the car is parked and connected to wifi. Four years back, Autopilot lacked a lot of the finesse and detail it has now. Artificially induced cognitive load was basically nonexistent then. There were warnings to hold the wheel, but that's it. It used to accelerate and brake like an inexperienced teenage driver - lacking the smoothness of gradual human control. It simply didn't engage in a lot more situations. It was clear it was an early technology, but still better than its then peers.
Today, even v1 Autopilot is a vastly different creature. The deceased Mr Huang had a car with v2 Autopilot that's even better. All Model 3s have v2, only the older Model Ss and some Xs have v1.
The difference is on two fronts. One, it's a LOT more smoother and 'real'. It brakes and accelerates gradually. With hands on wheel as immediate backup, I let it work in pouring rain - it handles the circumstances extremely well, managing astonishing lanekeeping combined with noticeably more gradual acceleration and braking than in dry conditions. It recognizes stopped cars and objects, an early criticism It detects and adjusts to speed limit signs. These are the smoothness and 'realness' factors.
The other change is in cognitive load. It is much higher now in terms of how often and how much you need to keep telling the car that you're actually responding to it. It has become - as lamented heavily on Tesla forums - a nanny role now. If you do not respond to the car's 'demonstrate attention' notices thrice, it disables autopilot and forces you to stop the car, wait and restart to re-enable it.
There are therefore two themes here - one is that the car has dramatically more cognitive load applied artificially, on the very basis you claim. The other is that car's Autopilot system has become vastly more capable than it was 3-4 years ago. It's nowhere near Musk's exaggerations, but it works spectacularly well both in stop-start city and highway grand tourer cruising modes.
I do not think the problem is a lack of cognitive load. Rather, L2+ autonomy is fundamentally a sweet spot of the perfect storm of being competent enough to convince the user that it works great, but it fundamentally does not. Making it either less or more capable would improve things. I do not see even an increased cognitive load compensating for the fact that the system works so well that it engenders complacency, but still lacks the competence to be actually trusted.