Everybody needs insurance, in one form or another (at least in the US).
I wonder if a company like Google that can possibly determine non-obvious risk factors, has a major advantage here.
E.G., imagine if people who searched for the term "dui attorney" were 40x more likely to be involved in a vehicular homicide as a defendant. Google could refuse to insure those searchers, and, as such, significantly cut everyone else's premiums (giving them a major competitive advantage).
I have no special knowledge about the distribution of insurance payouts, but I would guess it follows a power distribution. If so, removing the top 10% of payout insurees could HUGELY decrease insurance payouts.
Does anyone know if the major costs to auto/home/health insurers are payouts?
Excluding healthcare, do people care if folks who exhibit risky behavior pay higher premiums? In the US, young men (actuarian-proven to be higher risk) pay higher auto insurance, and everyone seems fine with it.
I think the more you reduce your pricing based on being able to exclude buyers using secret criteria, the more you run the risk of failing to exclude the riskier population, and under-pricing your product as a result.
If the conditions for purchase are stated in the contract, then in a sense, you know exactly what you are getting, and you can price more accurately.
If pricing is based on something like secret tracking or user surveillance, these are necessarily factors the buyer hasn't explicitly agreed to disclose to you.
I'm super curious about the limitations of information an insurance company can use to profile you. Do you have to agree to every factor explicitly? Or could Google throw in something like "...and factors based on your interaction with Google products"
I could see this having large adverse selection effects. All the "low risk" people buy cheap insurance from Google, and the incumbent insurers are left with a pool of only high-risk customers and have to dramatically increase their rates accordingly.
Even if that works well for 99% of people, it could have seriously negative effects for some. After all, a 99% accurate ML algorithm still leaves 1% of "innocent" customers with no way of getting affordable insurance.
Ding ding ding! NSA surveillance is just the tip of the iceberg. I've been saying it's only a matter of time until the data mining companies (https://news.ycombinator.com/item?id=7853118) get into the insurance game. If you want or need to save money, you'll have to conform to the average boring citizen - the more you stick out from the herd (literally "deviation"), the more risky you will be to insure (auto/health/home). If you have the resources you won't see a problem - surely there's people living much riskier lives and why should you subsidize them. If you cannot pay increased premiums, these companies' actuarial tables will become mandatory conditions of your life. This is the standard carrot-stick (Huxley-Orwell) class-dependent gradient.
And in the crypto facist US it will be hard for most people to even see what's wrong with such a massive chilling effect on individual behavior and thought. After all, the "market decided" (ignoring both the extreme government intervention driving the market as well as its path-dependence) and "you can choose not to patronize them" (until there's no other game left in town for your de facto mandatory policies). The government (NSA) may be the only entity that can lawfully send men-with-guns after you, but they aren't the only entity that can kick you out of your home and prevent your access to food.
9 comments
[ 3.3 ms ] story [ 31.7 ms ] threadI wonder if a company like Google that can possibly determine non-obvious risk factors, has a major advantage here.
E.G., imagine if people who searched for the term "dui attorney" were 40x more likely to be involved in a vehicular homicide as a defendant. Google could refuse to insure those searchers, and, as such, significantly cut everyone else's premiums (giving them a major competitive advantage).
I have no special knowledge about the distribution of insurance payouts, but I would guess it follows a power distribution. If so, removing the top 10% of payout insurees could HUGELY decrease insurance payouts.
Does anyone know if the major costs to auto/home/health insurers are payouts?
Excluding healthcare, do people care if folks who exhibit risky behavior pay higher premiums? In the US, young men (actuarian-proven to be higher risk) pay higher auto insurance, and everyone seems fine with it.
If the conditions for purchase are stated in the contract, then in a sense, you know exactly what you are getting, and you can price more accurately.
If pricing is based on something like secret tracking or user surveillance, these are necessarily factors the buyer hasn't explicitly agreed to disclose to you.
Now, I would bet that they let you OPT-IN to sharing that data, in order to qualify for a discount, which is effectively changing the price curve.
Even if that works well for 99% of people, it could have seriously negative effects for some. After all, a 99% accurate ML algorithm still leaves 1% of "innocent" customers with no way of getting affordable insurance.
If there's one thing Google will not stand for, it's highly trained customer support reps.
In that context, insurance is an interesting automation problem which is totally solvable.
http://www.bloomberg.com/news/2013-02-19/paulson-leads-funds...
And in the crypto facist US it will be hard for most people to even see what's wrong with such a massive chilling effect on individual behavior and thought. After all, the "market decided" (ignoring both the extreme government intervention driving the market as well as its path-dependence) and "you can choose not to patronize them" (until there's no other game left in town for your de facto mandatory policies). The government (NSA) may be the only entity that can lawfully send men-with-guns after you, but they aren't the only entity that can kick you out of your home and prevent your access to food.