There have been a couple proposals floated for ways to increase our confidence in alternative incentive mechanisms for filecoin in evaluating evolution / changes to the incentive program. One current proposal is that an experimental network can be spun up in parallel for a fixed period of time with a compatible PoRep where data being onboarded to filecoin can also be opted-in to being opted in to the experimental network. Then based on the subset of the network that participates in the side net we can evaluate the outcomes of the alternate mechanism.

The basic premise of this is attractive - we want to increase our confidence and get a sense of what an outcome will be before we commit to it.

We do need to think through the behavior of a rational actor and the outcomes that are incentivized by such a process that that in developing a way to experiment we make sure that outcomes that are good for the long term health and growth of the network are able to succeed.

In the above straw man proposal, one of the important factors that isn’t spelled out is what incentive is provided for participation in the experimental network. A side network would have it’s own token / block rewards, and it would be possible for the network to reimburse participants / provide a pot of concrete value that can be used to reward participants as the experiment concludes as a way to incentivize participation. The economics for how to price such an experiment are pretty tricky, because both extremes can be argued to incentivize “the wrong thing” and it’s unclear there’s a value in the middle that is as aligned as we might hope.

• If the price paid out for an experiment is negligible compared to primary filecoin block rewards, there isn’t a direct financial incentive for participation / change of behavior - e.g. if I’m an SP that isn’t serving retrievals and that costs me money, I wouldn’t set that up if there’s no money being provided and I have to pay a real cost to participate in the experiment. This means there are a few rational motivations to join:

  • If you’re already configured to meet the criteria of the experiment you may participate in order to get the minimal additional free value at no cost.

  • If you think a successful outcome of the experiment leads to evolution of the network in a direction that is lucrative for you, you may invest in the network.

• If the price paid out for an experiment is sufficient make participation profitable / matching the expected participation incentive for filecoin, we have different problems: it becomes extremely difficult to find funds to run an experiment, and the short term nature of the experiment participation don’t insure the costs of long-term participation

  • If the cost is too high, it becomes a question of who has money to run an experiment, and only those with significant stake become eligible to do so.

  • Getting costs / modeling set appropriately will require a full crypto-economic re justification for each such experiment, which has not lowered the cost of running the experiment, as was the original intention. If we want to make sure pricing is set to incentivize behavior in the same way as it will if the change is implemented, we would need to understand if we’re setting collateral for the 3 month term of the experiment, vs for the anticipated year that it would actually be stored - and how costs that are amortized over that longer period, like bandwidth and other operational costs - are amortized / understood given the known fixed period of the experiment. Even once we do such an analysis, it seems like the complexity of the behavior economics are difficult to communicate to provide full rational behavior by participants. The mechanism of ‘payout at end of fixed term’ also seems to make it very difficult to structure an experiment that is designed to incentivize long term health of the network at the cost of short-term profit. If everything concludes after 3 months, there’s no way to discourage participation by participants who are unwilling to hedge or provide service for the prospect of future rewards.

Reflecting on this space, this experimental framework seems to further the interests of a proof of stake network / existing token holders who are either exclusively able to shape the development of the network and who are the existing power-table participants who are able to choose to opt-in/make any such an experiment ‘ a success’ through their participation.

I think instead, we do better to develop/provide an evaluation modeling framework and ask proposed changes to demonstrate the effects of the change within the model. The model can provide a general demonstration of the outcomes of rational actors given an incentive change, along with a space for the experiment proposers to provide color with their chosen insights into the specific outcomes they’re attempting to modify.

I believe part of this framework has been already built in several context, but I’m not sure how general or how easily adapted current versions are to upcoming proposals. I would encourage us to see if an experimental framework can be made public and encouraged for proposals.