Solve sybil-resistance & create a transparency loop for Gitcoin to be governed by the community, thus saving the planet by restoring the balance of funding between public and private goods. Sounds simple, right? On second thought, it wasn’t that simple. Since the first grants round, we knew that at some point these questions would come up. Sybil attacks are an inevitable part of open systems with an unknown number of participants and unknown motivations of said participants. It is quite…
Solve sybil-resistance & create a transparency loop for Gitcoin to be governed by the community, thus saving the planet by restoring the balance of funding between public and private goods. Sounds simple, right?
On second thought, it wasn’t that simple. Since the first grants round, we knew that at some point these questions would come up. Sybil attacks are an inevitable part of open systems with an unknown number of participants and unknown motivations of said participants. It is quite literally an unsolved research question.
A unique data set from the first ever implementation of quadratic funding presented an excellent opportunity. We first engaged with Blockscience to help us better understand the problem after round 6 and found their results to be incredibly insightful. Here is an example from the most recent grants round.
Their reviews prompted us to more deeply understand the potential for multi-disciplinary community involvement in governance of the grants ecosystem. When Token Engineering Commons brought the idea of an Open Science collaboration to study our unsolved problem, we were ecstatic to see how it could help.
TE Academy was created as an 8-week research program. They directed how we could use our influence in the community to find participants from all over the world, specialists in many fields of study. They devised a plan which incentivized productive participation. Their methods directly created this successful experiment.
Working with BlockScience and the cadCAD organization for this open science initiative, they were able to leverage each of our orgs capabilities and turned them into possibilities. After a few discussions about what we would like to learn, TE Commons facilitated the program which ended up exceeding not only our expectations, but the participants expectations for the experience as well.
The co-researchers in the program, focused on their own research questions, provided insights by bringing multiple skills to the table: PhD mathematicians working with their students to improve the speed of the QF simulations, physicists and mathematicians battling for the best approach to define the optimal attack strategy in a grant round, machine learning experts designing novel methods for detecting sybil collusion, and node operators building community data access tools to analyse contribution networks. We came out of this experience with 6 presentations showing the results of these findings. Each of the research group members met with the gitcoin team providing a wide array of ideas and suggestions.
Here are the areas of research explored during the program:
Investigating the Optimality Gap as an Indicator of Collusion on Gitcoin Grants
Perform an exploratory analysis over the optimality gap metric, seeking to characterize the associated distribution as well as exploring emergent properties that may arise.
Strategic Attacks as an Optimization Problem
Understanding how a user or group of users could manipulate the system to achieve maximum matching for a particular grant or group of grants.
Permissionless Data for the Gitcoin Grants Community
Empowering the community to govern through transparency according to Ostrom’s 8 principles of common pooled resource management.
Node2Vec for Collusion Detection
Taking a look at how to build a feature set when looking at collusion detection as a semi-supervised machine learning problem. Identifying attack vectors.
Quadratic Funding Linear Algebra
What is the optimal strategy for collusion? By understanding via provable mathematical theorems, we can better select mitigation strategies.
Optimality Gap Sensitivities
A look at the tradeoffs in the sensitivity of the algorithm. First, develop an algorithm to detect collusion. Then adjust sensitivity to identify false positives and false negatives.
Graphing Sybil Score Characteristics
Find out how the sybil score performs in detecting fraudulent accounts and their subsequent behavior throughout the round.
From our end the experience was simple. Clearly communicate our open questions. Provide the needed data. Let Token Engineering do their magic.
The insights turned out to be useful, robust, and actionable. Most surprising was the number of suggestions outside the specific research questions that came from a group of incredibly intelligent individuals with a deep knowledge of the problem.
The Gitcoin team has received enough suggestions for new valuable workstreams that we needed to internally workshop how we prioritize addressing each of the insights generated by the program. The TE Academy research program served as a catalyst to understand what issues we face today, what issues will emerge tomorrow, and the many ways we can refine our product to better serve its users.
As we move forward, Gitcoin will progressively decentralize the grants platform. This begins with community governance. TE Commons showed us a model for actively engaging with top talent in the community and sped up the timeframe in which we can offer more control of the system to the users.
“We see this an excellent way to spur high quality community participation in solving complex problems for many projects in the ecosystem. We are looking forward to collaborate again with TE Commons and TE Academy to support the community-driven development of this new domain Token Engineering, that will be so vital for the success of Web3.” Kevin Owock, Founder of Gitcoin.