Thoughts on Liberal Radicalism

In their new paper Vitalik Buterin, Zoe Hitzig, and E. Glen Weyl propose a mechanism for a more efficient allocation of resources towards public goods. The mechanism, called Liberal Radical (LR), seeks to create “formal rules for a society neutral among communities”. Which essentially means a way for a society to deploy collective resources towards public goods while remaining neutral with regard to its constituents often competing preferences.

The mechanism builds on a similar mathematical foundation as that of Quadratic Voting, however, rather than voting constituents direct personal contributions towards public goods and the LR mechanism is used to amplify those contributions. The funds contributed in this way result in the public good recieving an amount proportional to the square of the sum of the square roots of contributions directed towards the public good. As a result of this process, the impact of many small contributions are magnified through a subsidy funded by a deficit/tax model or philanthropic entity. The paper concludes that based on some assumptions about the behavior of economic agents the mechanism should result in a more efficient allocation of resources towards public goods than one would expect from either pure market capitalism or through a democratic governance process.

Mechanism Overview

In order to get an intuitive grasp of the proposed mechanism it was helpful for me to create a simple excel model based on a limited set of participants and funding recipients. The model looks at 5 participants (Column A) and 3 public good recipients (Column B).

In above scenario Alice has a strong preference for X and makes a direct contribution of $10 dollars but she is the sole contributor. Since she is the only one to contribute to X the square of the sum of square roots ends up being exactly what she alone contributed. This result can be seen in cells F2 and G2.

In contrast all five participants contributed a small amount to Y, each contributes $1 dollar for a total direct contribution of $5. After applying the LR mechanism this results in a subsidy of $20 for a total contribution to Y of $25.

Finally, two of the five participants, David and Edward, both have a relatively strong preference for Z and contribute $5 each. After applying the LR mechanism the contribution doubles from $10 dollars to $20.

From these examples it’s clear that depending on the distribution and amounts the “deficit” between direct contribution and the amount prescribed by the LR mechanism can become arbitrarily large. If the mechanism’s source of funding must be constrained, then deficit amounts can be adjusted proportionally to fit within an arbitrary budget. This process is described in the paper as Capital Constrained Liberal Radical (CLR). In the screenshot above this is shown in column I and J based on an arbitrary budget of $10.

Simulating how the LR and CLR mechanism worked to allocated funds helped me form a more intuitive understanding of the mechanism, and the mechanism seems to have some really interesting and desirable properties. However, when thinking of how this could be applied in practice there are a number of challenges to consider…

The challenge of defining a “Society”

We use the term “society” to refer to the set of all participants and the word “community” to refer to groups that fund a particular public good; however, in many applications, the relevant “society” is itself a community within the broader world.

Based on this definition a “society” can be many different things, from a macro perspective we are all part of a global human society, and if we zoom in we have nation states, states, and local municipalities. We can even imagine societies that exist in the digital realm where there are no clear geographic borders and a much more fluid definition of what constitutes a participant.

In the context of these more fluid societies, such as a blockchain network, the public goods are non-excludable even to those external to the society, or whom participate in technically adjacent societies. This means that even though the solution attempts to address free-rider issues within a “society”, there may still be free-rider issues between societies. This is especially true if immigration and emigration from a society is as low friction as making a trade on an exchange.

Imagine if Ethereum were to fork, with one version allocating resources using the capital constrained liberal radical (CLR) mechanism funded by a fixed inflation rate, and one fork maintains the status quo. Users who subject themselves to additional inflation by holding ETH on the CLR chain would be subsidizing public goods that would benefit both chains, and users who choose to support the non CLR chain would be able to free-ride on the public goods produced as a result of the CLR fork. There likely exists a non-zero inflation rate allocation for public goods funding that would be sufficiently low as to not result in a free-rider chain. A similar situation could also exist within a single chain, with users choosing to avoid inflation by using other assets as a store of value rather than ETH as response to an LR policy.

In an ideal case we could use the unconstrained version of LR in order to reach a more optimal funding rate for public goods, however, it seems likely that competitive pressure and overlap between public goods might prevent that from being practical.

The Challenge of sybil-resistance

Consider a society of N citizens i = 1, . . . N ; we take as an assumption throughout what follows that we can verifiably distinguish among and identify these citizens

As with Quadratic Voting, the LR mechanism imposes rules on individuals which can be circumvented if an individual is able to control multiple identities. Someone who manages to control multiple identities can split their contributions into smaller amounts and cause the LR mechanism to trigger a subsidy to their chosen destination. If there are no restrictions on funding recipients the attacker may even select a destination that is effectively their own personal bank account!

For some applications it may be sufficient to do KYC checks on each participant to ensure that it is difficult to create multiple accounts, however, in the context of decentralized networks relying on a centralized KYC service may be unappealing.

Decentralized solutions to sybil-resistant identity are generally not feasible right now, though perhaps in the future a web of trust solution like BrightID could help address this issue.

The Challenge of Collusion

If we assume that we have a reasonable way to prevent individuals from directly possessing multiple accounts, there is still an issue with collusion. If multiple participants coordinate their efforts they can manipulate the mechanism for their own financial gain.

If perfect harmonization of interests is possible, Capitalism leads to optimal outcomes. LR is intended to overcome such lack of harmonization and falls prey to manipulation when it wrongly assumes harmonization is difficult. This is a bit of a paradox: while LR seeks to foster community direction through its design, in doing so it relies on strong ties of community flowing outside the design not existing.

In the context of smart-contract enabled blockchain networks this harmonization of interests is fairly trivial. Someone can deploy a smart contract which will distribute the funds it collects equally to each individual who contributes some minimal amount and propose that contract as a public good. Since the LR mechanism subsidizes small contributions from large groups, the total amount collected by the contract would be significantly larger than what was contributed and so when distributed back to each participant they will all profit at the expense of the LR funding source.

It seems the only way to prevent this, at least in the context of smart contract platforms, is to have some sort of curation process that restricts what can be added as an available public good that is eligible for funding. However, if this curation process is required, it undermines a core value propositions of the LR mechanism— that it doesn’t presume centralized knowledge about the preferences towards specific public goods.

The challenge of irrelevant public goods

If funding is coming from a particular source, whether a community or specific philanthropist, it may be important to restrict eligible recipients to meet some specific criteria to prevent the funding of public goods that are irrelevant to the community or philanthropist. This is particularly important in cases where the expectation is that the public goods will have a positive impact on the societies economy that justifies the economic costs incurred by subsidizing the public goods.

For example, if the source of funding is related to the usage of a digital software commons (e.g. the HCL) it might be reasonable to expect funds to be allocated for the purpose of growing that software commons and not for other arbitrary purposes like improving the roads in a specific neighborhood. Similarly, if the Ethereum community choose to implement an inflationary policy as a means to fund a LR mechanism, funding R&D on Bitcoin or EOS would probably not be a desirable result.

Much like above, the solution to this seems to require some sort of curated whitelist of approved public goods.

Conclusion

The Liberal Radical mechanism is very interesting and while there are some challenges to implementing it in practice the idea feels worth exploring more.

I’m generally left with the following questions:

1. If it ends up being the case that a whitelist is required in order to make the mechanism practical, would just using Quadratic Voting offer similar benefits?
2. If we wish to apply this to funding public goods in the context of a decentralized blockchain network can the issue of preventing sybil-attacks be resolved without dependence on a centralized authority for identity verification?
3. If we must rely on a centralized authority for identity verification, is it acceptable given that the decentralized network can coordinate to fork the subsidy away in the event that the centralized identity verification entity becomes corrupt?