Overview
This represents my opinion and not necessarily W3F’s.
Recently, there has been extensive discussions in the Polkadot community about the current landscape of OpenGov and how to improve it. Many interesting and provocative ideas have been put forward. Irrespective of my opinion on them individually, this shows how engaged and invested many members are. I believe that there has never been a time where Polkadot was more vivid than today.
While reading many of the comments and proposals, I had recurring thoughts that are better summarised here in a single document than to scatter them in multiple comment sections.
Firstly, it’s important to note that my observations are rooted in the intuition that the outcomes yielded by OpenGov are generally “good”. Making this claim is hard to defend, since it is impossible to judge whether preferences in referendums have been aggregated “accurately”; That is the whole problem at hand in collective decision making. However, one observation is that there has never been a runtime upgrade accepted that introduced malicious code and broke the network and faulty referendums were quickly “nayed” by a huge majority. Additionally, controversial referendums have seen extensive discussions on many platforms resulting in a prolonged voting period, giving enough time for everyone to inform themselves and show up to vote. Based on these observations, I want to give a few thoughts on Polkadot’s OpenGov in its current state and ways to further improve it.
Turnout: an ambivalent metric
A key metric in evaluating OpenGov is turnout as an important proxy of how well things are going. We can define it as how many voters (or in Polkadot’s case, DOT) show up to express their opinion in a referendum (potentially as % of the total DOT that are available to vote). High participation and turnout is generally indicating that things are going well, because it means that we can be more certain that an outcome is representative, i.e,. we could approximate the overall preferences of all token holders properly.
To get a better understanding about the current turnout in OpenGov, we can take a look into the data. There are many degrees of freedom here, but I decided to take the following perspective: Let’s aggregate all the unique DOT (with the simplifying assumption that they haven’t moved during the time of analysis) that have voted in at least one referendum in the last 30 days (in fact since Referendum #398). Under these conditions, there are 132.8M DOT (without conviction) that turned out. Normalizing this by the number of DOT that technically can vote (denoted as active issuance (i.e., removing tokens in treasury, crowdloans, and staking pools), this is roughly 9.8% of eligible DOT.
This number is far from ideal, but it is not too bad either, especially once we consider that a large share of the active issuance is held within custodial solutions, which (unfortunately) generally do not offer voting. It is crucial to note that a low turnout does not necessarily mean the outcome is not representative. Here, things are more nuanced: Generally, people vote on things that they care about. But, people do not necessarily care about all the proposals put forward in OpenGov. Also, many voters might not feel educated enough to vote on certain proposals. Then, the better thing to abstain, because otherwise they would dilute the aggregation of preferences of those that actually care or are informed by simply adding noise to the equation. This also emphasizes the importance of the conviction mechanism which effectively maps the weight of an opinion in the election to the (subjective) confidence or importance attested by the voter to their choices. Then, in the case the election result was in line with their vote, they incur costs by longer locking period for higher conviction.
Another aspect less often discussed is the potential inefficiency of achieving high voter turnout for every referendum. Consider, for instance, a runtime upgrade that merely cleans up existing code. If such an upgrade receives a few million DOT with 100% approval, it might be unnecessary, even inefficient, for the remaining 1.3x billion DOT to also directly participate in voting. This inefficiency arises for a couple of reasons. First, voting generates network traffic, leading to transaction fees for users. Secondly, the time invested by individuals in understanding a proposal that’s already undisputed could be seen as excessive.
However, this example illustrates a dangerous pitfall. Observing referendums with overwhelming approval should not lead voters to simply trust the witnessed majority and abstain or, even worse, jump on the bandwagon and blindly support the winning side. It should also not cause experts to fall victim to variants of social loafing, reducing their effort in reviewing proposals only because they see other experts expressing their choices. Uninformed opinions would only be adding noise to the aggregation of preferences, leading to potentially dangerous side-effects.
Despite the argument that a lower turnout can still lead to representative outcomes, the question about resilience remains. A motivated and sufficiently backed entity could try to “slip through” a referendum, hoping to exceed low resistance faced from generally low turnout. This would mean that a low turnout is not necessarily explained by a lack of interest or expert opinion, but rather by the arguably sometimes slow response of many voters. While this is generally a risk, OpenGov is designed to be resilient towards that by scaling the time required to approve a referendum with turnout & approval rates. While the degree depends on the respective track, generally, a referendum executes slower if the turnout is low or if approval is low (i.e., it is disputed). This design makes it resilient towards the case where an entity wants to push through a malicious or controversial proposal quickly, as it would either reflect in low approval or low turnout. Then, the voting period is extended significantly (or rather not shortened), which allows the social protection layer to unfold. All voters have time to rally allies, make others aware of the situation. To a certain degree, we’ve seen this already with extensive mobilisation of voters through social medialeading to much above-average turnout rates in controversial referendums (for example in referendums #431, #401, and #213).
While observing low turnout in voting does not inherently lead to unrepresentative results – it may indicate a consensus on acceptability or rejection of a proposal, or simply a lack of strong opinion among many token holders – it is important to recognize the value of high voter turnout. High participation ensures a diverse range of opinions and reinforces the legitimacy of the outcomes. Therefore, even though low turnout might sometimes reflect a proper outcome, seeking ways to increase turnout remains a key objective. This ensures that decisions are made with the broadest possible engagement, capturing a more comprehensive snapshot of the community’s views.
Increasing turnout
Before diving into ideas that could improve the situation, I’d like to urge caution with an approach that is frequently proposed: incentivizing voting.
The dilemma of incentivizing voting
A popular idea floating around to increase turnout is to provide extrinsic, specifically monetary incentives to vote: Psychologists and behavioral economists, however, have long established that observed outcomes for actions that are mainly driven by intrinsic motivation can suffer from perverse effects after introducing extrinsic incentives. This observation has been labelled the “crowding-out effect”. It appears that an intrinsic motivation is not reinforced and strengthened by the extrinsic one, but rather replaced. Then, if the extrinsic one is not large enough, people stop caring about the action and either abstain from performing it or stop putting sufficient thought into it. As it turns out, voting and participation in governance is something inherently driven by intrinsic motivation. Rationally, most people should not even vote, because they are unlikely to affect the outcome, an observation known as the paradox of voting.
While the weighted voting scheme with conviction on Polkadot makes it currently more likely to be pivotal, i.e., affect the outcome, the underlying consideration remains: People are driven by other benefits from voting. For example, many feel a civic duty to vote, do it to build / maintain their reputation, or want to contribute to the social norm of voting by giving a good example. The paper “Extrinsic Rewards, Intrinsic Motivation and Voting” investigates whether monetary incentives lead to increased turnout in a real field setting (Panagopolous (2012)). The author found that monetary payouts, in the best case, only marginally increased turnouts and that “[…] this study suggests there are likely limitations to the mobilization capacity of even sizable incentives” (p.277). In addition to weak evidence that monetary payouts even increase turnouts, the author was unable to determine whether, even with larger turnout, the outcome of the election is of “higher quality”. Indeed, as outlined above, we don’t only want more people to vote. We want more people to inform themselves before voting to make a quality assessment of the problem at hand. It’s impossible to tie a monetary incentive to making an informed vote, because most referendums need to be judged by individual preferences. It is easy to visualise situations in which these incentives lead to individuals casting votes to collect their payouts. This, in fact, can already be observed from 1kv validators that used bots to auto-nay referendums. Even the most motivated ones might find themselves short of time to properly assess the situation, joining the bandwagon and voting for the winning side. In contrast to the aforementioned study, using automated voting bots are easy to be used, likely leading to a degradation of the quality of outcomes.
In short, providing monetary incentives will inevitably lead to (undetectable) noise in the referendums that dilute informed opinions potentially leading to worse outcomes. While turnout might be increased with sufficient incentives, it is doubtful that the quality of vote aggregation increases.
Removing disincentives to vote
Giving direct incentives to vote is a delicate topic and could lead to perverse effects, so the question remains what can be done to increase turnout. A safer and therefore better bet is to reduce the disincentives to vote. In other words, we should strive for ways to reduce the costs of voting and/or the costs of acquiring information on the proposals. This, if anything, makes it more likely that people that want to vote will vote. The first and most direct impact would be to reduce or even eliminate the transaction fees for voting. The implementation of this is not trivial, because it must be reasonably sybil resistant, in other words, preventing malicious entities from clogging the system by spamming votes. This, however, might become a reasonably fixable problem if governance is moved to a system parachain.
Another important lever is to reduce the costs of acquiring and condensing information about the various proposals. As we strive for votes that are informed, these are large costs that a voter needs to pay (in terms of time and energy) before voting. There are recent advances in that direction, such as having expert talks, governance roundups and public discussions. Those efforts should be continued, aiming to condense information on proposals to make them easy to understand for voters.
Give a stage to delegates
Delegating votes is a powerful tool to further strengthen the robustness and resilience of Polkadot. The underlying (but reasonable) assumption is that delegations are made towards experts that are actively voting and kept in check by their motivation to preserve their reputation. It is also an effective tool to increase turnout by less active voters that want to give their voice to dedicated delegates that represent their interests.
While still in its early days, a political spectrum within Polkadot is emerging, giving rise to fundamentally different opinions across referendums. The most notable formation of this can be seen in the Treasury track where some voters appear to be rather fiscally conservative while others are less so. This opinion transcends the individual proposals and can better be described as a general stance towards treasury spending. With further decentralisation efforts, other aspects such as the evaluation of protocol changes will increasingly fall into a political spectrum, too. With different entities within the Polkadot community proposing changes, they will more commonly go beyond fundamental features and optimizations but rather fundamentally steer the direction that Polkadot is developing into.
To navigate this emerging complexity, delegating votes to experts, potentially different experts per track, is technically possible on Polkadot. While delegation faces challenges mentioned earlier, such as high transaction fees and intricacy, a specific issue stands out: Currently, there is a lack of platforms providing a forum for potential delegates. This makes it challenging and less convenient for voters to find representatives to whom they can delegate their voting power. These platforms should provide avenues for delegates to showcase their voting histories and articulate their positions on the political spectrum. Such transparency and clarity will empower voters to make informed decisions, enabling them to find and delegate their tokens to representatives whose views align with their own. While individual voters are still important, it would increase turnout among those voters that are less active but still want to see their opinion represented.
Increase turnout potential
The previously mentioned measures aim (primarily) to enhance participation among already activated voters. But, we could also slightly shift the perspective and focus on currently inactive voters. The metric would then be driven by the question of how many DOT could turn out in a specific time frame if they chose to. This could be denoted as the turnout potential. While it is challenging to quantify this metric precisely, it is vital to explore initiatives to increase it.
Before discussing ideas that could have a positive effect here, let us first discuss why this metric might be important after all. Being able to rally large amounts of voters in a short period of time is crucial in two distinct scenarios. First, if we encounter a malicious entity that tries to damage Polkadot via governance and exerts strong pressure with large amounts of DOT and/or with high conviction. In this case, we’d need to be able to rally a large proportion of voters to turn the approval in the referendum and prevent it from passing. While this is definitely a realistic scenario (and varying in degrees of severity), it also has to be noted that the economic costs the entity would endure is relative to its voting power. In other words, the more likely it can pass something against many other voters, the more stake it needs to have. Second, and potentially more likely, is the situation where we need to respond to critical issues such as bugs and implement urgent upgrades. Depending on the severity, we’d require a rapid, widespread voter response. In contrast to the previously mentioned scenario, where we need to flip the approval, here we’d need to exceed barriers imposed on the support (which is simply turnout in favour of the proposal) in order to execute the referendum as quickly as possible. For example, to execute a whitelisted call in less than 24 hours, we’d require a turnout of 20% and beyond.
If we want to find ways to increase turnout potential, we’d need to remove obstacles that prevent voters from turning out. Addressing these barriers can significantly increase the potential turnout.
There are several avenues that positively impact the turnout potential. First, we should move DOT that are locked in staking pools into the active issuance, an issue already being worked on here. Second, we should encourage custodians to allow their customers to participate in governance. Here, the “Split Vote” mechanism comes in handy, because it allows one to represent the preferences of numerous individuals with a single vote (by a single stash). Pressuring custodians to implement voting is a challenging task. The first step would be to encourage individuals that hold DOT in custody to actively ask for that feature. With enough pressure and the mechanic of split vote, achieving this task might be realistic. Another endeavour is to incentivize people to learn to vote (and potentially move DOT to self-custody). While a critic could make a similar point as mentioned above, that non-voters are simply not interested enough, I’d argue those voters might not even know that they are interested in the proposals that are up for vote, because they lack the entry to the whole process. Here, we could try to leverage the Treasury and conduct experiments to help onboard voters. Building on ideas proposed in referendum #431, the idea might be interesting:
- Conduct this experiment on Kusama.
- Start a referendum that aims to distribute (some reasonable amount) of tokens to new voters voting in the referendum.
- Make only accounts without a voting history eligible to receive tokens.
- Require a minimum 4x (or even 5x) lock-in for eligibility.
- Repeat the proposal for 2 or 3 months.
- Afterwards, analyze whether these voters continue participating and stick around. Also, we could cross-reference how many of the new DOT are transferred from (identified) custodial addresses.
Note, that this does not incentivize voting in general, but rather incentivizes to learn to vote (a kind of “pay to learn” scheme). While it is somewhat prone to sybil attacks (a user could transfer tokens to a new account without voting history), the fact that many current voters are already locked with conviction (or voting) and the fact that their first participation would exclude them from later rounds, seem to make it a reasonable approach.
A positive byproduct of increasing turnout potential is that it almost inevitably increases actual turnout. This happens as more individuals learn about voting or gain access to their tokens, and subsequently find themselves engaged in other referendums.
Conculsion
OpenGov is an exciting evolution in collective decision-making, bringing together a vibrant and engaged community. Generally, things are going quite well, though there is always room for improvement. Turnout can serve as approximation of engagement of voters and representativeness of election outcomes, but it remains a nuanced metric. We should further explore measures to increase turnout, while also considering potential (and large) adverse effects certain approaches may entail.
Focusing on removing barriers to participation, rather than implementing direct voting incentives, appears to be a more conservative but safer path. This strategy helps preserve the intrinsic motivation behind voting, crucial for maintaining the quality and integrity of decision-making.
Furthermore, enhancing the delegation process by creating platforms that enable potential delegates to present their views and voting records could empower voters to make informed decisions. This would not only facilitate a broader representation within the governance process but also potentially increase turnout by engaging those less active in direct voting. Through these measures, we can further refine Polkadot’s governance system, ensuring it remains responsive, inclusive, and representative of its diverse community.
Lastly, it is essential to consider the turnout potential and explore strategies to increase it, thereby further strengthening Polkadot’s resilience and prepare us for future challenges.