WeTrust recently announced the success of its four-week-long experiment in Liberal Radical (LR) charitable donation matching, which it's calling the first implementation of the LR equation as applied to fundraising for 501(c)3 verified nonprofits.
In the experiment, donors on WeTrust's Ethereum-based platform were provided with a list of causes they could choose to contribute Ether to. Contributors did so with the knowledge that their donations would be matched by WeTrust according to the design framework laid out in a September paper titled "Liberal Radicalism, Formal Rules for a Society Neutral among Communities," authored by Vitalik Buterin, Zoe Hitzig, and E. Glen Weyl.
In the end, the donation drive gathered over 190 Ether from contributors, and WeTrust contributed an equal amount, totaling roughly 380 Ether distributed across the 16 organizations that garnered contributor support during the drive (there was one cause that garnered no support). However, the matched funds were not distributed equally in accordance to the dollar amount contributed to each fund. In the end, the "winning" cause, the Machine Intelligence Research Institute, garnered more than 147 Ether from 64 donors, which was met with a contribution of almost 174 Ether from WeTrust.
Liberal Radical Donation Matching
According to the LR model, the number of total contributors matters more than the amount contributed by each individual. In math terms, "the amount received by the project is (proportional to) the square of the sum of the square roots of the contributions received." Put more concretely: Under the standard donation-matching model, if there is one person who contributes $1,000 to one cause, and there are 10 people who contribute $100 each to another cause, and both funds are matched, then both funds would receive an additional $1,000. Under the LR model, the fund with 10 contributors will receive much more money from the matching pool than the fund with one contributor. Specifically, the fund with only one $1,000 contributor would receive $1,000, whereas the fund with ten $100 contributions would receive $10,000. By adding the square root of each individual contribution together, and then squaring that number, the Liberal Radicalism model favors causes with more public support than causes that have only a few wealthy supporters.
WeTrust's implementation of this is slightly different and more complicated than the above model because the matched funds are limited. WeTrust agreed to match up to $100,000 worth of Ether. To do so, WeTrust followed a variation of the LR mechanism, which is available on its website.
This funding model is a part of the bigger Liberal Radical movement, which seeks to radically rethink existing social, economic, and political frameworks and expand "competitive, free and open market mechanisms to reduce inequality, build widely-shared prosperity, heal global political divides and build a richer and more cooperative social life." And the ideas core to LR, including this funding model, are pretty radical and untested on chain and off. So WeTrust's experiment is an important one, both in terms of studying the incentives the model creates, and in testing the viability of an LR crowdfunding model on Ethereum.
In a conversation with ETHNews, Hitzig said that in the original Liberal Radicalism white paper the authors assumed a model wherein all contributions were made blindly and at the same time. The effect of this is that donors would not be influenced by the knowledge of the behavior of other donors. Under these conditions, one benefit is that people might be more likely to contribute in line with their "true" preferences, something Weyl and other liberals (in the classical sense) tend to place a lot of importance on. While there may be value in that, Hitzig stated that the original paper is somewhat limited in scope, because "many implementations will be both dynamic and public," as was the case with WeTrust's experiment.
That being the case, she says, it is important to study "how public information on existing contributions may shape individuals' decisions" in an LR model. By looking at a dynamic picture of contributions, a researcher could "better understand how incentives to contribute might change as causes gain – or fail to gain – support."
To this point, information about the donation drive – the pseudonyms of donors, the amount they donated, and when they donated – is available on each individual cause's donation page, though detailed information regarding dates and times of contributions are not provided. If you visit these pages, you'll notice that the donation drives for each organization are still ongoing, just without the promise that contributions will be matched (the fund matching period ended on Giving Tuesday, November, 27 at 11:59 pm PST). Further, though it is not clearly stated on the donation pages, the funding drive began roughly two weeks before WeTrust announced it would match donations using the LR model, which happened on November 14.
To make this information more intelligible, WeTrust provided ETHNews with a graph representing the number of donations-per-day made to all causes, shown below.
While this data may provide some insight into the effect of public information on contributors' decisions, the graph is somewhat limited. A more detailed graph – one marking the date of the announcement of the LR matching experiment, and including the rate of donations since the conclusion of the experiment – would allow for more analysis. Further, for any meaningful conclusions to be drawn, more experiments in LR funding models are necessary, as well as more detailed data. As it stands, it is not possible to account for a number of factors that may have contributed to donor behavior, such as differing levels and types of promotional efforts between causes, or potential differences in donor base after the announcement of the LR fund matching experiment. It's also probable that the spike in donations had something to do with Giving Tuesday.
The Identity Problem
Another hurdle in meaningfully analyzing the results of this experiment is that that a public, dynamic donation process might make it easier to game the system, since potential bad actors would have the data at their disposal to do so most effectively. And, as the Liberal Radicalism white paper points out, without identity verification, sybil attacks (one person pretending to be many) are possible even if the donation period is static and private – something of particular relevance to a mechanism that intentionally considers the number of contributors over amounts contributed by each. This makes analyzing results from any Ethereum-based LR experiment difficult.
The WeTrust experiment took this concern into consideration by asking donors to create accounts, though all a person needed to provide was an email address. Donors without accounts were also able to contribute, but all donations from anonymous contributors were combined and considered as if they came from a single donor. Though, WeTrust admits this procedure only "thwart[s] the most rudimentary sybil attacks."
Li admits this is a low threshold and that individuals could simply use different email addresses. This is because, he said, WeTrust values user privacy, and it's well-understood that solid identity protocols are hard to come by in the crypto space. Ultimately, though, while identity verification is something WeTrust cares about, Li did not see this as a major concern in this particular instance. He said:
"To a certain degree, at least for this experiment, if you want to really game the system … you would just be helping a particular nonprofit more than you would have, which I think is not the end of the world, because these are nonprofits that we vetted and are legitimate, but in the future as we plan to open this up to enable … any organization to onboard, then [identity solutions] will need to be more rigorous because you could then be donating to yourselves."
The available information regarding donors suggests possible attempts to game the system, as there are a number of donors who contributed small amounts multiple times, perhaps not understanding that all contributions made by a single pseudonym would be considered as a single donation. Additionally, there are a number of contributors with handles that vary only slightly from each other. Li stated that WeTrust's data shows all registered donors came from different IP addresses, though also admitted that this could be faked.
A Better Experiment
While the lack of identity verification might not have presented any immediate issues, in a Twitter exchange with Buterin, Weyl raises concern that without it, data from the experiment would not be particularly helpful.
A better LR experiment, the two conclude, would involve a trusted funding organization, such as the Ethereum Foundation (EF), using trusted donors (of "the current technocracy," in Buterin's words) to provide the initial donations to determine fund-matching distribution by the EF. Buterin suggests a retrospective analysis would also be necessary to determine the efficacy of the projects under this funding model. Weyl adds that such a study would also require a control group where donations were matched under a different principal.
Perhaps WeTrust's LR implementation does not stand as an ideal experiment, but that's not the mark of success for WeTrust. Li said that the hype around LR seemed to attract donors, so the organization may run similar LR-based donation drives again in the future.