> Keep in mind the whole point of insurance is shared, proportional risk. The fundamental problem is that the more information is brought to bear on the idea o

> Keep in mind the whole point of insurance is shared, proportional risk. The fundamental problem is that the more information is brought to bear on the idea of "risk", the less shared it is. This is especially stark in health insurance: do you have family history of serious diseases? Do you carry a particular gene? Do you engage in risky behavior like smoking, adventure sports, promiscuous sex? It can price some people out of the insurance market altogether, and make it less of a "shared" pool. This is precisely the problem the ACA is trying (with debatable success) to tackle. Auto insurance is arguably more directly tied to statistically risky behavior rather than factors beyond your control (like genetics), but the potential is still there. We as a society are not ready for total information awareness on the part of our insurer. If Google offered insurance directly, could they deny a claim based on an email you sent before or after an accident? Raise your rates if you search for fast cars online? What's the difference between actually risky behaviors, and behaviors that are merely correlated with higher risk? The insurance markets are opaque enough without adding black box algorithms working on nearly limitless data inputs.