I've been thinking about these kinds of issues too, in the context of setting up a new block chain where the money supply is more flexible.  Here's a system that I think might be workable, and would like to submit it to criticism from others.

Right now, there is a fixed monetary inflation schedule that we all "agree" on by dint of network effects, and there is no way to realistically change it.  It would be better if we could dynamically agree on how the money supply should grow or shrink.  The block chain is a fantastic tool for building consensus without trusting your peers.  So we could set up something like a voting scheme for what the monetary inflation/deflation rate should be.  Each miner that produces a block would add to it in a new field of the block header what they think the rate of monetary inflation/deflation should be (right now, it's fixed at 50 BTCs per block).  The actual inflation rate would then be the median of the voted inflation rates of the previous, say, 100 blocks (this bears resemblance to how "difficulty" is agreed upon nowadays).  If a block's coinbase transaction created anything but what the "consensus" inflation rate dictates, the block would be declared invalid.  Monetary deflation could then be implemented by forcing the miner to forfeit a portion of the transaction fees (if they don't, the block they produce is declared invalid by the network).  This is how a "transaction tax" might be implemented in a way that doesn't allow free-riders to ignore it.

The reasons I think this might work are the following:

a) In current monetary systems, a central bank rather opaquely decides how much money to create/withdraw, and any money created is hard to distribute without political influence.  While I happen to personally think that this system works reasonably well most of the time, I can empathize with the queasiness that lack of transparency produces.  Having a more transparent decision mechanism is certainly better (although shooting ourselves in the foot with the current transparent system that's transparently dysfunctional is much worse, IMHO).  I have a feeling that this queasiness is an important emotional factor in Bitcoin adoption today.  Voting on decisions about the money supply continuously would allow a decentralized consensus to emerge (dare I say, "letting the market decide").  Additionally, the money created goes to miners, instead of politically connected banks and treasuries, and the miners really are providing a valuable service (securing the blockchain). Additionally, in principle anyone can arrange to receive a part of the money created by joining a mining pool.

b) Using the median of the votes from a large number of blocks prevents all sorts of problems from free riding.  If, say, we were to allow any miner to decide how much to inflate/deflate without needing consensus, then the individual incentive is to inflate as much as possible, which, when unregulated, screws everyone over (cf. the tragedy of the unregulated commons).  By voting and forcing a consensus, you lose the immediate incentive to inflate everyone else away, and can vote on deflating or keeping the money supply steady.  If everyone else disagrees with you, at least you get to partake in the money creation.

c) The incentives are set up to keep investment in mining at a reasonable level.  If the consensus is to keep the supply steady or deflate, then miners that just want to get-rich-quick will be discouraged.  This would make it easier for miners that really want to increase the money supply to be the ones that find successful blocks, but they are forced to put their money where their mouth is and temporarily mine at a loss in order to register their votes for an increase.  This might occur when lots of new users join the network and/or there is pent-up demand for additional money due to increased desire for trade: the money supply should grow to accommodate the additional economic activity they'll bring online.  If, on the other hand, the consensus is to inflate, miners that want to deflate have to put their money where their mouth is and increase their mining activity to register enough down votes: in the time it takes to do this, they are participating in the inflation, which would subsidize the additional investment.  It also forces early-adopters to invest their hoards, lest their value be inflated away, which would mitigate the initial distribution problem.  In all cases, there is no threat that mining will be so discouraged that the security of the block chain is in jeopardy.

What are people's thoughts on these ideas?  Virtues?  Flaws?

Deflation shouldn't be artificially limited. What you are discussing is a central plan. All central plans fail due to the economic calculation problem.  The greater the value of a bitcoin, the greater the incentive to spend/sell it. This is a natural limit. We have seen this principle at workk for several weeks now and there is no reason to believe it will stop working when bitcoin production slows.

I haven't, have no intention of doing, and said "in the context of setting up a new block chain where the money supply is more flexible".  The underlying technical side of Bitcoin is rock solid, and there are many many people in this forum are thinking of ways of building on *that* without having to subscribe to deflationary economics.

You know what they say, if it ain't broke...

That is a problem best solved with companies that will offer private insurance funds and other classic hedging solutions long in use in mutlicurrency trading.   There is no need to burden the currency itself and hence everyone with those costs.

Whether everyone should be forced to participate in an insurance pool or just those who choose to is open to debate. Insurance can be prohibitively expensive when it is elective. Where can I purchase forward bitcoin contracts in non trivial volumes now? How much does this insurance cost?  How do i know that the insurer won't pull a 2008 banking crisis and fail to pay? I don't think usable insurance for bitcoin will ever materialize without revamping the generation protocol. This in my view is an Achilles heel.

Perhaps there are different ways of registering these votes?  The reason why I thought miners might do the trick is that the barrier to entry to mine is not too high, so anyone who wants to register their opinion about inflation rates can start mining and join a mining pool.  Right now, deepbit (the dominant mining pool) accounts for ~50% of new blocks.  I should also add that one might also implement speed limits on the rate at which the target inflation rate can be changed, in the same way that difficulty now can't be increased or decreased by a factor of more than 4 every two weeks.

Originally, I had liked Suggester's old suggestion of tying mining rewards to hash rates, so that the money supply would at least roughly track the number of active users weighted by their investment in the system.  But that doesn't really allow for monetary deflation when it's necessary, and it also ties up huge resources in mining, neither of which seems like a good idea.

I'd be happy to hear of other ways in which the target inflation rate might be determined.  Personally, I would favor something democratic and dynamic over an algorithmic solution, as a way to acknowledge that we're all fallible and can't perfectly predict the future, so we need a mechanism for allowing consensus decisions to be changed.

Most viable would be imho a theoretical exchange rate against a basket of all major fiat currencies. But if you don't believe in them, that's probably not an option. Historians have a similar problem and afaik they often use a very basic and overall consistent measurement, the price of a cow.

I agree that Euros and dollars will have a more stable long run valu than difficulty, but this requires a trusted mechanism for reporting exchange rates to the system. I don't see how this could happen without centralization. If someone has a clever mechanism for incentivizing nodes to make truthful reports to the blockchain, I would be very excited by that. (nerdgasm) It needs to be relatively low cost and incentive compatible (truth must always be more profitable than falsehood). Good luck with this designing these systems is challenging/maybe impossible.