Coronavirus, insurance, and crypto: the limits of distributed system design

I've been an insurance coverage lawyer for the better part of two decades, representing policyholders whose claims are denied. It's that experience, in part, that got me interested in blockchain technology (I've written some about that experience in the past here).

One of the things that has both frustrated and kept me occupied as a coverage lawyer is the fact that insurance companies happily and quickly take premiums for policies but exercise significant discretion when it comes to paying claims. 

We are about to see this writ large in the case of coronavirus-related losses which, by all accounts and any reasonable prediction, will be catastrophic. I look at this both as an insurance lawyer but also with the eyes of a technologist – with feet in two centuries, I suppose. 

On the one hand, I ponder how to get money right now for clients. On the other hand, I think about system design and technological solutions for the future. Programmable money that guarantees claims payments is a neat idea, but human mess creates practical implementation challenges that can't be ignored.

The question is whether there is something about this deterministic technology that can fix or address subjective, stochastic, indeterminate disputes over contract language. Stated differently, it's one thing to create a programmable one-way transaction to cover a single trigger loss like a flight delay; it's another (I think) to use distributed append-only contract automation to resolve commercial business interruption losses.

What do I mean? Perhaps it will be helpful to describe some of the issues that insurance companies and insurance policyholders are experiencing.

While I can't in this column design the solution to these complex problems, understanding the problem set might help illuminate potential opportunities for system designers and investors and also the limitations that crypto insurance innovators face. Many businesses carry property insurance policies that may cover loss or damage from business interruptions. I am oversimplifying a lot, but that is the gist (and it's also, to be honest, very much on my mind).

With that, consider the following complexities:

• Property loss and business interruption coverages may require physical loss or damage. It may be possible to argue that the presence of a virus on a property may constitute physical damage and trigger coverage. However, forms vary across the United States and policies are subject to varying judicial interpretation and regulatory oversight and approval.
• Some policies may have exclusions that specifically exclude coverage for losses caused by a virus, but many policies do not exclude such losses; exclusions for pollution, fungus or mold are not the same. Again, meaning here is not binary – there are innumerable contingencies.
• Governmental orders may trigger coverage for business losses under provision addressing losses caused by directives from or by a "civil authority." As with the preceding examples, there is a good deal of fuzz here, and it is fuzz that can be mustered in favor of coverage if there is any ambiguity.
• Policies may also include coverage for "contingent business interruption" that arises from supply chain or customer access disruption. Creating a fixed schema to define this sort of disruption is not something that blockchains make any more or less easy to sort.

These sorts of issues are and will be the source of considerable dispute between insurance companies and policyholders. I question whether this is complexity is something that distributed systems qua distributed systems can do much to address or solve on their own. 

There have been some interesting and notable attempts to fix insurance "on the blockchain."

The Etherisc flight insurance model – which uses a single trigger (delay) and is tied to external data sources is the one that has made the most sense so far. I'm increasingly skeptical, however, that complex chains of claim decisions can be automated or that so-called crypto-economic incentives will do much more than add complexity theater to a process that is already complicated enough.

Take Nexus Mutual, which uses a so-called "discretionary mutual" model, where humans "stake" funds and collectively make entirely discretionary decisions about whether or not to pay claims. According to the Nexus website, you can use Nexus to "[s]ecure risk and potential bugs in smart contract code [and b]e covered for events like The DAO hack or Parity multi-sig wallet issues." With a "tokenized mutual model … all members of the mutual benefit[] from the platform's success, aligned incentives will foster a community spirit rather than the existing adversarial and unbalanced relationship between individual and institution."

Kumbaya, y'all. 

But seriously, dropping a token and staking into an already complicated process seems really unlikely to make people more whole for their losses any more expeditiously. Indeed, the fact that there was a lengthy discussion among members before 16 claims were rejected is probably cold comfort to those who suffered a loss, as the emoji-laden announcement pasted below made clear:

I'm not saying that this decision is correct or incorrect, by the way. I'm saying that I don't see what blockchain did to make the process any better or how "staking" added anything to the result. I also wonder how anyone would feel if Chubb sent them a denial letter laden with emojis.

"What’s it all mean, Steve?"

Ultimately, if I were using blockchain/crypto to redesign insurance, I would ruthlessly remove human discretion from the process and focus on identifying binary/on-off decisions that don't require such discretion. Crypto-economics and complexity theater seem likely to do little but make a fraught and convoluted process even more so.