Fascinating! Having put the childcare plan in particular to an insurance expert, they give the below challenges and I'm interested in any pushback you may have on these points?
1. There are v few ‘mandatory insurance’ requirements in this country, for good reason. Insurers don’t find them profitable so it’s difficult to find insurers to offer the insurance - at an affordable price anyway.
2. Insurers would hate a menu of government regulations saying ‘child losing one finger = £10,000’ and so on.
3. For the most part there’d be contributory negligence because children do stupid things.
4. Any ‘fine’ for a child carer would be accompanied by a stinging premium increase from the insurer - thereby driving up child care fees for that carer, thereby nullifying savings due to fewer staff.
5. Unsafe practice only comes to light after an accident happens.
Thanks so much for your comments – broad take is that we agree among the principal implementation issues, and see our thoughts below.
TLDR: We agree the general point for childcare is that, in fact, serious accidents (of the kind that would be covered in compulsory insurance, for childcare and nuclear power) are extremely rare, and much of our response comes from this point. Though Ofsted doesn't keep numbers! (https://www.nurseryworld.co.uk/news/article/exclusive-concern-over-lack-of-data-on-nursery-deaths-and-injuries), because all deaths/injuries are often reported on national news, we'd be surprised if they were more than low single-digits a year, for ~4 million children 0-4 years old.
1. Re. Profitability: The requirements for childcare feel analogous to workplace insurance, despite fines and downsides presumably being similar (even businesses under £2m in turnover are liable for up to £450k in fines https://worknest.com/health-safety-fines/) - so given insurance is currently provided for even very small businesses it seems likely childcare would work also?
2. Yes, we agree that this might be costly to assess. For workplace insurers, there is also uncertainty over the payout for a specific injury because it is awarded by courts/tribunals, but they do range significantly too: this (https://www.national-accident-helpline.co.uk/injury-and-accident-claims/accidents-at-work, under "How much work injury compensation could I claim?") claims injuries like ankle injury can range from £12,900 - £46,980. We expect that regulators choosing these will remove some uncertainty, else, they can choose to adjudicate through a similar system. Minor injuries, like falling down and bruising one's knee, would not be included
3. Yes, makes sense. Attribution is hard with children, particularly given that they can't assume responsibility for their own actions. Because the accidents are so rare, we generally establish criminal liability for serious injuries/accidents, that means there are already existing procedures for attribution.
4. We agree that premium increases will be large, but we don't think there will be many, if any, more accidents than currently, and because of the reputational costs already associated with serious accident/injury.
5. Similar to car insurance, there are tests/audits that an driver/nursery can take to certify to the insurer they're more safe than the median driver/nursery they're insuring (for driving, you have advanced driving courses) and therefore benefit from lower premiums. We could imagine large insurers implementing something similar, a self-reported audit, which could help reveal unsafe practice + incentivise correction.
Would insurance help for events that are very rare, and we have little to no past data? also a breakdown would have huge knock on effects that wont just be arbitrated in the court.
This is regarding using insurance for nuclear and AI. It seems specially impossible for the latter, where the direction and capability of tech may change at any time
1. Yes, you are correct in observing that it is difficult for insurers to calculate the probability of some events, though it's likely to much easier than regulators, who will therefore default to the most restrictive policy. The insurers are incentivised to find the accurate probability for such rare, high-cost event, similarly to how actuaries have good models of other rare, high-cost events, like wars, which we do sell insurance for.
2. Is what you are trying to ask with the "breakdown" that a lot more arbitration will be needed? Could you elaborate?
3. For AI, what we were pointing at was that this is a much more flexible regime that would perform better in situations where "the direction and capability of tech may change at any time", because the standards/regulation would not have to update to the specific process by which AI is produced, instead focusing on the societally beneficial outcomes (maximising interpretability, preventing accessible bioweapon production etc.) without requiring the technical expertise and speed needed for safety of each of the latest research paradigms.
I am surprised that war insurance exists since it highly depends upon the leader of the state and its neighbours which can swing with an election. but it must work since its being offered. but its fascinating research topic and I wonder what parameters they use. and why aren't they being talked about in forecasting world?! (I hadn't come across this fact)
for 2, I was referring to a disaster that causes public backlash and which might result in added regulation that cripples the industry. it happened with nuclear and might do with others too. Is it possible for insurers to prevent themselves from public whim that is highly arbitrary? or just a matter of taking some losses and de-risking by offering diverse options?
definitely agree with your note on using Insurance for AI. but since the predictions could be off in such a nascent industry, wouldn't mandatory insurance cripple the startups or even scare away new ventures by existing tech giants?
Fascinating! Having put the childcare plan in particular to an insurance expert, they give the below challenges and I'm interested in any pushback you may have on these points?
1. There are v few ‘mandatory insurance’ requirements in this country, for good reason. Insurers don’t find them profitable so it’s difficult to find insurers to offer the insurance - at an affordable price anyway.
2. Insurers would hate a menu of government regulations saying ‘child losing one finger = £10,000’ and so on.
3. For the most part there’d be contributory negligence because children do stupid things.
4. Any ‘fine’ for a child carer would be accompanied by a stinging premium increase from the insurer - thereby driving up child care fees for that carer, thereby nullifying savings due to fewer staff.
5. Unsafe practice only comes to light after an accident happens.
Thanks so much for your comments – broad take is that we agree among the principal implementation issues, and see our thoughts below.
TLDR: We agree the general point for childcare is that, in fact, serious accidents (of the kind that would be covered in compulsory insurance, for childcare and nuclear power) are extremely rare, and much of our response comes from this point. Though Ofsted doesn't keep numbers! (https://www.nurseryworld.co.uk/news/article/exclusive-concern-over-lack-of-data-on-nursery-deaths-and-injuries), because all deaths/injuries are often reported on national news, we'd be surprised if they were more than low single-digits a year, for ~4 million children 0-4 years old.
1. Re. Profitability: The requirements for childcare feel analogous to workplace insurance, despite fines and downsides presumably being similar (even businesses under £2m in turnover are liable for up to £450k in fines https://worknest.com/health-safety-fines/) - so given insurance is currently provided for even very small businesses it seems likely childcare would work also?
2. Yes, we agree that this might be costly to assess. For workplace insurers, there is also uncertainty over the payout for a specific injury because it is awarded by courts/tribunals, but they do range significantly too: this (https://www.national-accident-helpline.co.uk/injury-and-accident-claims/accidents-at-work, under "How much work injury compensation could I claim?") claims injuries like ankle injury can range from £12,900 - £46,980. We expect that regulators choosing these will remove some uncertainty, else, they can choose to adjudicate through a similar system. Minor injuries, like falling down and bruising one's knee, would not be included
3. Yes, makes sense. Attribution is hard with children, particularly given that they can't assume responsibility for their own actions. Because the accidents are so rare, we generally establish criminal liability for serious injuries/accidents, that means there are already existing procedures for attribution.
4. We agree that premium increases will be large, but we don't think there will be many, if any, more accidents than currently, and because of the reputational costs already associated with serious accident/injury.
5. Similar to car insurance, there are tests/audits that an driver/nursery can take to certify to the insurer they're more safe than the median driver/nursery they're insuring (for driving, you have advanced driving courses) and therefore benefit from lower premiums. We could imagine large insurers implementing something similar, a self-reported audit, which could help reveal unsafe practice + incentivise correction.
Would insurance help for events that are very rare, and we have little to no past data? also a breakdown would have huge knock on effects that wont just be arbitrated in the court.
This is regarding using insurance for nuclear and AI. It seems specially impossible for the latter, where the direction and capability of tech may change at any time
1. Yes, you are correct in observing that it is difficult for insurers to calculate the probability of some events, though it's likely to much easier than regulators, who will therefore default to the most restrictive policy. The insurers are incentivised to find the accurate probability for such rare, high-cost event, similarly to how actuaries have good models of other rare, high-cost events, like wars, which we do sell insurance for.
2. Is what you are trying to ask with the "breakdown" that a lot more arbitration will be needed? Could you elaborate?
3. For AI, what we were pointing at was that this is a much more flexible regime that would perform better in situations where "the direction and capability of tech may change at any time", because the standards/regulation would not have to update to the specific process by which AI is produced, instead focusing on the societally beneficial outcomes (maximising interpretability, preventing accessible bioweapon production etc.) without requiring the technical expertise and speed needed for safety of each of the latest research paradigms.
I am surprised that war insurance exists since it highly depends upon the leader of the state and its neighbours which can swing with an election. but it must work since its being offered. but its fascinating research topic and I wonder what parameters they use. and why aren't they being talked about in forecasting world?! (I hadn't come across this fact)
for 2, I was referring to a disaster that causes public backlash and which might result in added regulation that cripples the industry. it happened with nuclear and might do with others too. Is it possible for insurers to prevent themselves from public whim that is highly arbitrary? or just a matter of taking some losses and de-risking by offering diverse options?
definitely agree with your note on using Insurance for AI. but since the predictions could be off in such a nascent industry, wouldn't mandatory insurance cripple the startups or even scare away new ventures by existing tech giants?