This article is Part 4 of a series of articles on Optimism Retro Funding voting designs based on GovXS’s Voting Design Evaluation. The framework allows the assessment of Retro Funding voting systems and enables the Optimism Collective to achieve design goals (i.e., Impact=Profit, Simplicity for Voters) by measuring the performance of optimal voting rules. The framework applies both axiomatic analysis and agent-based simulations to ensure comprehensive evaluation.
GovXS’s resources provided for Retro Funding voting design evaluation include:
Part 1: Making Optimism Retro Funding Strategyproof (Incentive Compatibility in Optimism Retro Funding)
GovXS Presents: Evaluating Voting Designs for Optimism Retro Funding (Workshop with Optimism Badgeholders)
GovXS Evaluating-Voting-Design-Tradeoffs-for-Retro-Funding (Open-source simulation framework to measure how different voting designs perform against several typical retro funding design goals)
A Social Choice Analysis of Retroactive Funding (Formal Description)
Optimism Retro Funding is a system designed to reward positive contributions to the Optimism ecosystem. With 850M OP earmarked for Optimism Retro Funding’s evolution, this initiative has the potential to influence the successful development of Optimism and the Superchain significantly.
Central to the program’s success is its voting design. To support this, GovXS developed a framework that rigorously assesses voting designs against six crucial and common design objectives for retro funding programs.
For Optimism, GovXS evaluated the following five design goals and published detailed reports on each:
Resistance to Malicious Behavior: Evaluate the extent to which malicious voters can influence voting outcomes.Report: Securing Retro Funding against Malicious Behavior
Incentive Compatibility: Determine whether the voting design is incentive compatible, meaning that each participant, or a coordinated group of participants, achieves their best outcome by voting according to their true preferences.Report: Making Retro Funding Strategyproof
Simplicity for Voters: Assess how easily voters understand the voting design and and how to best behave to achieve their goals.Report: Measuring Simplicity for Voters
Incentive Alignment: Determine if the voting design aligns the incentives of voters with those of the Optimism Collective.Report: see below
Impact = Profit: Evaluate the effectiveness of the voting design in distributing funding according to the principle "Impact=Profit."Report: see below
For Optimism Retro Funding, we did not evaluate Majority vs. Diversity
This article summarizes findings regarding Incentive Alignment and Impact=Profit and proposes a roadmap for further refining the voting design of Optimism Retro Funding based on our evaluation results.
In evaluating the design dimensions for Optimism Retro Funding, we developed a cumulative score encompassing all assessed metrics. Our evaluation covered all voting rules applied in Optimism Retro Funding Rounds 1–4. Here’s what we found:
Incentive Compatibility is lacking across all voting rules. We recommend implementing Majoritarian Phantoms or non-normalized median rules.
R1 Quadratic Voting performs best overall, showing particular strength in resistance to malicious behavior and achieving the highest scores in Impact=Profit (see below).
R2 Mean demonstrates strengths in Simplicity for Voters and Impact=Profit but is more susceptible to malicious behavior.
R4 Capped Median emerges as the strongest median voting rule, offering robust resistance to malicious behavior, whereas R3 Quorum Median shows serious vulnerabilities.
Finally, all voting rules lack sufficient Incentive Alignment between voters and the Optimism Collective (see results below).
The following sections focus on the key issues identified in the design goals “Incentive Alignment” and “Impact=Profit.”
We’ve analyzed the following voting designs:
Round 1 (R1 Quadratic Voting): For Round 1, Optimism implemented Quadratic Voting, allowing voters to allocate tokens with increasing costs to cast multiple votes.
Round 2 (R2 Mean): For the second round, the funding allocation was determined by averaging the votes across all participants, applying the Mean Rule.
Round 3 (R3 Quorum Median): In the third round, a minimum vote and token threshold were introduced, with the Normalized Median rule being applied to reduce the influence of outlier votes. Only projects that met a minimum number of votes and funding were considered.
Round 4 (R4 Capped Median): For Round 4, Optimism introduced the Impact Metric Score, where voters prioritized KPIs rather than direct funding allocations. Projects submitted KPI data to validate their achievements, and funding was distributed based on voter preferences using the Normalized Median, with caps on allocations and redistribution of overflow funds. GovXS includes a version that skips calculating the Score based on Impact Metric Share for better comparison.
The formal specification of all voting rules in the evaluation can be found here.
Incentive alignment is a core challenge in principal-agent problems, where conflicts arise as one party (the agent) acts on behalf of another (the principal). These issues typically emerge when the agent holds more information and may have different interests than the principal, making it difficult for the principal to ensure the agent’s actions align with their best interests. This dynamic is evident in Retro Funding voting, where the Optimism Collective, as the principal and funder of the program, relies on voters (the agents) and the information they hold to evaluate proposals and vote in the Collective’s best interests.
A second dimension of the principal-agent problem is the incentive structure itself. Actions beneficial to the principal might be costly or effortful for the agent, and the principal cannot easily observe how the agent acquires information or makes decisions. This holds true for Optimism’s Retro Funding, where properly assessing projects demands effort from voters, and Optimism lacks control over how voters gather their knowledge.
Consequently, the objectives of voters and the Optimism Collective might be at odds. To solve this, incentive mechanisms are essential to align voters’ choices with the overarching goals of the Optimism Collective.
To assess incentive alignment in Optimism Retro Funding, we start with a straightforward question: Is it beneficial for a voter to participate? We can measure this with the “participation” metric:
This metric is critical for ensuring voters have no disincentive to participate, thereby supporting the system’s legitimacy.
Interestingly, only the R2 Mean rule satisfies the participation criterion. In R2 Mean, if voters abstain, their influence is removed, typically resulting in less favorable allocations for their preferred projects. Abstaining doesn’t lead to higher utility; therefore, R2 Mean satisfies participation.
For R1 Quadratic Voting, this does not apply. Consider an election without a voter where the output without is [0.8, 0.2]. If the true vote of is [0.8,0.2], casting a vote can still push the mean allocation away from their preference due to the quadratic influence of votes.
Similarly, in R3 Quorum Median and R4 Capped Median, participation might lead to less favorable outcomes for voters. In these voting rules, quorum constraints and normalization apply. If voter prefers to allocate a small amount to project , participating in the vote may help meet the minimum quorum, potentially leading to a much higher allocation for than envisioned. In this case, abstaining from voting might lead to an outcome closer to the actual preference of voter .
The Participation metric helps assess if voting participation benefits voters, yet incentive alignment encompasses more aspects, such as voter incentives and skin in the game.
Our next evaluation asks whether voters benefit from voting in the Optimism Collective’s best interest and if they participate in the success of Retro Funding. Conversely, we ask whether they risk any loss, in other words, if voters have “skin in the game”.
Voter incentives and skin in the game can take various forms, such as:
Monetary rewards: Voters receive compensation for their time spent evaluating projects.
Performance-based rewards: Voters earn a share of funds allocated to winning projects if their assessment aligns closely with the final allocation.
Access rights: Voters increase their voting weight when voting beneficially or gain preferred access to future votes.
Penalties: Voters lose rewards or privileges a)–c) if they act against the Optimism Collective’s interests.
In Retro Funding Rounds 1–4, before the actual voting, a group of badgeholders reviews applications for eligibility and are compensated for their time. Additionally, a code of conduct exists, with potential loss of voting rights for violations.
However, beyond reviewer compensation, no further voter incentives or risk-based penalties exist in Retro Funding, which could threaten its success. First, without incentives, there’s little to reward voters for thorough project evaluation. A voter might invest substantial time refining allocations across applications or choose to allocate all funding to a favorite project ignoring all other applications. Regardless, voters cannot expect any gain from the success of “good” projects, nor do they face any risk for supporting “bad” projects.
Second, this lack of “skin in the game” makes bribery a low-risk option. With no rewards at stake, a rational voter may find it more beneficial to accept even minimal bribes over voting according to their true preferences.
In Social Choice/Epistemic Democracy, there is a concept called Ground Truth that refers to the “right” allocation that should ideally be achieved.
In Optimism Retro Funding, we assume that there is an objective truth, a positive impact made on the Optimism Collective. We also know that there’s an optimal distribution of the funds: Impact=Profit, ”positive impact to the Collective should be rewarded with profit to the individual”.
The voting conducted in Retro Funding aims to find this objective truth using crowd wisdom since we don’t have any better instrument to measure impact and define profit. Otherwise, we could simply write an algorithm that computes project funding based on application data.
In Optimism Retro Funding voting, we assume every voter provides a noisy estimate of this objective truth. Each voter’s preference can be seen as an approximation of the truth, influenced by own information, biases, and constraints.
In simulations, we can assign a voter’s deviation of the true optimal allocation and compare the aggregated output the different voting rules give. For our experiments, we’ve created artificial voting data. GovXS uses the Mallows model to generate a matrix of cumulative votes for 𝑛 voters and 𝑚 projects. Each voter’s total vote sums to 𝐾 (in this case 𝐾 = 1), and the votes are generated with some randomness controlled by the parameter 𝛼. We use 𝛼 = 0.5, representing a balance between homogeneity (all voters being similar in their vote distributions) and heterogeneity (each voter having individual differences). The detailed specifications for our experimental design are available here.
With the metric Alignment with Ground Truth, we find a voting rule that is as close as possible to (any) true optimal allocation. Note: we don’t define the objective truth itself here; we measure how well the voting rule outputs the objective truth (Impact=Profit) by processing the same randomly generated voting profile across all voting rules tested.
The R1 Quadratic Voting has the lowest L1 distances, and a narrower interquartile range (IQR), suggesting more consistent results. In contrast, with the R2 Mean and both median rules (R3, R4), L1 distances are higher and show wider IQRs, hence greater distance between votes and outcome. These results indicate that the L1 Quadratic Voting rule best supports finding the ground truth in funding allocation.
The L1 distance metric does more than evaluate a voting rule’s effectiveness; it enables ongoing monitoring of each voting round to see if opinions converge toward a shared ground truth (indicated by a decreasing L1 distance) or if perspectives on fair compensation remain divergent. By examining L1 distance on a project-by-project basis, we can gauge the community's alignment on each project’s deserved funding: a high L1 distance reflects a broad range of opinions, while a low L1 suggests consensus on the appropriate funding level for a given project. Of course, any numbers on alignment are only reliable if we can be sure that voters vote according to their true preferences.
For insight into the current state, we analyzed 39 projects that successfully received Retro Funding Round 3 and 4 funding. Round 3 took place in October 2023, with a 30M OP round size and 146 eligible voters assessing 643 projects. Round 4 distributed 10M OP in funding in July 2024, following 108 voter assessments on 230 projects.
We calculated the average L1 distance across all voters per project to compare voter alignment. The results provide a real-world snapshot of how aligned or varied voter opinions were regarding funding allocations.
The chart above shows all projects, in descending order according to the funding allocation they received. In Round 3, L1 distances are generally higher than in Round 4, indicating stronger alignment among voters in the later round—a promising trend. Round 3 also shows greater volatility in alignment, with some projects sparking highly divergent opinions on their funding, regardless of the actual funding allocation. By Round 4, while some outliers remain, these cases are significantly fewer.
To validate this, we analyzed the correlation between funding amount and voter alignment. We found a positive correlation, with the highest L1 distances in Round 3 corresponding to projects with the largest funding amounts. However, this correlation weakens in Round 4, suggesting a shift toward tighter voter alignment over time. Overall, this trend toward increased alignment from Round 3 to Round 4 signals progress toward achieving the objective truth that “Impact=Profit.”
Alignment on Ground Truth can be monitored at the project level and across all projects to gauge whether the voting community is converging toward a shared standard in funding allocation. It also serves as a potential signal for shaping voter incentives, encouraging participants to consider the collective perspective in their funding choices. When balanced with additional KPIs (such as impact metrics) to avoid Groupthink, alignment with ground truth becomes a valuable metric for rewarding thoughtful voter engagement.
Based on the overall evaluation results, we propose the following prioritized roadmap to develop Retro Funding voting designs towards fairness, resilience, and alignment with Impact = Profit:
Establish Incentive Compatibility and Strategyproofness:
Establish Incentive Compatibility as the fundament for subsequent optimization steps. Implement voting mechanisms that encourage voters to express their true preferences, reducing any advantage gained from strategic voting, e.g Majoritarian Phantoms or non-normalized median rules.
Address Collusion and Enhance Group-Strategyproofness:
Develop safeguards against voter collusion by designing incentives that make collective manipulation unprofitable. Explore group-strategyproof rule modifications and voter incentives that deter groups from coordinating votes for personal advantage.
Monitor Voter Extractable Value (VEV) and Strengthen Resistance to Malicious Behavior:
Continuously assess the system’s VEV and susceptibility to malicious activity. Introduce voter incentives that decrease the likelihood of bribe-taking or control attacks.
Track and Refine Alignment with Ground Truth:
Monitor voter behavior and measure alignment on an objective Ground Truth round by round, optimize voting rules so that they best support Impact=Profit.
GovXS Voting Design Evaluation Framework, Figure 1: OP Evaluation Reports / Simulation Framework Optimism Retro Funding Evaluation Scores, Figure 2: We calculate the score based on the average result across all metrics per design goal; on a scale between + (worst) and ++++ (best) Incentive Alignment, Figure 3: results of axiomatic analysis x = does not satisfy property, ✓ = does satisfy property / results of agent-based simulations on a scale between + (worst) and ++++ (best) Alignment with Ground Truth, Figure 4: Voter type: Mallows Model / Simulation rounds: 100 / Round size: 8M OP / projects: 63 / voters: 40 / quorum R3: 17 / min. funding R3: 0 OP / min. funding R4: 1000 OP / max. funding R4: 500K OP Average L1 Distance per Project in Round 3 and 4, Figure 5: Voter type: R3 Voting Matrix / projects: 644 / voters: 109 and Voter type: R4 Voting Matrix / projects: 230 / voters: 108 Average L1 Distance vs. OP Received (R3, R4), Figure 6: Voter type: R3 Voting Matrix / projects: 644 / voters: 109 and Voter type: R4 Voting Matrix / projects: 230 / voters: 108
Kicked off with Optimism, the GovXS Voting Design Evaluation Framework is a tool to secure robustness, fairness, and trust in Retro Funding Voting Systems across all ecosystems. It covers design objectives like Resistance to Malicious Behavior, Incentive Compatibility, Simplicity for Voters, and more.
The open-source framework enables the prioritization of design goals and analyzes a voting design with formal rigor, applying axiomatic analysis and agent-based simulations.
GovXS Presents: Evaluating Voting Designs for Optimism Retro Funding (Workshop with Optimism Badgeholders)
GovXS Evaluating-Voting-Design-Tradeoffs-for-Retro-Funding (Open-source simulation framework to measure how different voting designs perform against several typical retro funding design goals)
A Social Choice Analysis of Retroactive Funding (Formal Description)
GovXS is a research initiative under Token Engineering Academy. Team members include Nimrod Talmon, PhD, Angela Kreitenweis, Eyal Briman, and Muhammad Idrees. GovXS is a member of the Token Engineering Academy Applied Research Network. Sign up to receive further updates related to GovXS.