Augur doesn't answer my original concern. If you have a smart contract between, for example, two companies, that information is private and it doesn't work like a popular baseball game. If one company doesn't follow the contract you need to solve the problem in the real world.
Also, we don't know yet how public blockchains (where Augur runs) work with State or powerful actor attacks, when the incentives are not purely financial.
One way you could solve the problem in the real world is track down the delinquent counterparty to the contract, drag them into an expensive legal system, subject them to the costs thereof, levy a penalty on them, hope that they pay it, in the even that they do not or cannot, seize their assets, in the event that they have none or you cannot, punish them with prison, in the event that they resist, kill them.
Sure, that's the traditionally accepted way to go about solving the problem in the real world. But there are many other ways to solve the same problems that we're all intimately familiar with by now based around reputation systems. When it becomes public that a counterparty to a contract has engaged in delinquency, they suffer a reputational hit. If their reputation has value (and there is nothing about pseudonymous entities that mean that they cannot accrue reputational value, rating systems work on darknet markets quite well and have since their inception), and the amount of damage to their reputation is higher cost than making good on the contract, then they will make good on the contract (and even if they can't, you've punished their delinquency by damaging their reputation, instead of dragging them through an expensive inefficient corrupt legal system involving violence as the final sanction).
All sharing economy and peer to peer systems that currently exist work on systems of this kind, and cryptocurrencies to the extent that reputations are valued and contracts made similarly so, and this will only become more prevalent as time goes by.
Reputation doesn't work in problems with a lot of money/power at stake. You can easily hack the system behaving like a good citizen for a long time until you consume all your reputation in a big transaction. Not only that, reputation is a hard problem if you are a newcomer.
That's why many people say "Everything is an exit scam in waiting". Even your legitimate meatspace bank will cheat if it thinks the economics make sense.
You could use transparency to tell how much a given escrow agent is holding and don't use them if it approaches the value of their reputation. And/or use a system where they don't hold the funds, they merely choose to unlock a transaction to pay to A or B.
Nothing works if the payoff for defection outweighs the costs of enforcement, that's why we have wars, crime, and all of the other things that the old fashion legal and diplomatic systems of the nations of the world have to deal with. However, conflicts of this type are rare compared to the very standard ones that are quite capable of being handled by reputation systems, escrow, etc. They solve these ordinary problems more efficiently, quickly, and at a lower cost.
That should be embraced and noted, and where appropriate migrated to, not pushed aside because there is a rapidly decreasing class of problems not suited to resolution with this model.
You don't make loans to people unless you have evidence that they can and will pay it back. BTCJam for example has reputation scores, and people actually say what they intend to use the money for.
The default rate on BTCJam, which was one of the first loan services is 10%, this doesn't compare well with mainstream rates however, Bitlendingclub is supposedly one third of that, which compares favourably to mainstream rates https://fred.stlouisfed.org/series/DRCLACBS
