Conviction Voting: Time-Weighted Governance for DAO Resource Allocation

Most governance systems capture a snapshot of voter preference at a single moment in time. A proposal is submitted, a voting window opens, participants cast their ballots, and the window closes. The result reflects the balance of power at that particular instant — which may bear little relationship to sustained community sentiment. Flash loan attacks, last-minute vote swings, and strategic timing all exploit this temporal vulnerability.

Conviction voting takes a fundamentally different approach. Rather than counting votes at a fixed deadline, it allows participants to continuously signal their preferences and accumulates conviction over time. The longer a voter supports a proposal, the more weight their signal carries. Conviction builds gradually — following an exponential curve — and decays when support is withdrawn. A proposal passes not when it reaches a majority at a deadline, but when its accumulated conviction crosses a dynamic threshold.

How Conviction Accumulates

The mathematical model behind conviction voting uses a continuous accumulation function. When a token holder allocates their tokens to support a proposal, their conviction begins at zero and increases over time according to a half-life parameter. After one half-life period, their conviction reaches fifty per cent of its maximum value. After two half-lives, seventy-five per cent. The conviction asymptotically approaches a maximum determined by the number of tokens staked.

If the holder withdraws their support, conviction decays following the same half-life curve — dropping by half with each period. This decay mechanism prevents hit-and-run governance attacks where participants briefly commit tokens to swing a vote and then immediately withdraw.

The half-life parameter is a critical governance design choice. A short half-life (say, three days) allows proposals to pass quickly but provides less protection against transient attackers. A long half-life (thirty days or more) ensures deep community commitment but makes governance sluggish. Most implementations settle on values between seven and fourteen days.

The Threshold Function

Conviction voting does not use a fixed quorum. Instead, each proposal has a dynamic threshold that depends on the proportion of the treasury or resource pool being requested. A proposal requesting one per cent of the treasury has a relatively low threshold — achievable with modest conviction from a minority of token holders. A proposal requesting fifty per cent requires massive, sustained community support.

This relationship between request size and threshold is typically modelled as a hyperbolic function. As the requested amount approaches the total available resources, the threshold approaches infinity — making it effectively impossible to drain the entire treasury through a single proposal. This built-in protection is one of conviction voting’s most attractive features for treasury management.

The formula creates a natural prioritisation mechanism. Small, broadly supported proposals pass quickly. Large, contentious proposals require extended periods of overwhelming support. The system self-regulates without requiring human moderators to set quorum levels or voting periods.

Origins and Implementations

Conviction voting was developed by the Commons Stack and BlockScience, with significant theoretical contributions from Michael Zargham and Jeff Emmett. The mechanism draws on ideas from control theory and dynamical systems rather than traditional political science.

1Hive was among the first DAOs to deploy conviction voting in production through its Gardens framework. Community members stake HNY tokens on proposals to fund public goods, with conviction accumulating continuously. The system has processed hundreds of proposals and demonstrated that conviction voting can function at scale in a live governance environment.

Giveth adopted conviction voting for its GIVgarden, using it to allocate funding to charitable projects. The continuous signalling mechanism proved well-suited to philanthropic allocation, where sustained community interest is a better indicator of project value than a one-time vote.

Token Engineering Commons (TEC) uses conviction voting as its primary governance mechanism, treating it as both a practical tool and a research platform. The TEC’s experience has generated substantial empirical data on conviction dynamics, attack vectors, and parameter optimisation.

Advantages of Conviction Voting

The mechanism offers several properties that address known weaknesses in traditional DAO governance.

Attack resistance is perhaps the most significant advantage. Flash loan governance attacks — where an attacker borrows a large token position, votes, and returns the tokens within a single transaction — are impossible under conviction voting because conviction takes time to accumulate. An attacker would need to maintain their position for days or weeks, making the attack economically costly and detectable.

Continuous participation eliminates the problem of governance fatigue caused by discrete voting periods. Token holders do not need to monitor proposal deadlines or interrupt their schedules to cast ballots. They simply allocate their tokens to the proposals they support and let conviction build. This persistent engagement model has shown higher effective participation rates than snapshot-based systems.

Proportional resource allocation emerges naturally from the threshold function. Rather than binary pass/fail outcomes, conviction voting allows multiple proposals to draw from a shared resource pool simultaneously. This is particularly valuable for grant programmes where the goal is to fund multiple initiatives rather than choose a single winner.

Minority voice amplification occurs because small groups of deeply committed supporters can accumulate sufficient conviction to pass modest proposals. In linear voting systems, these groups would be perpetually outvoted by larger but less engaged constituencies. Conviction voting rewards the intensity and duration of support, not just its magnitude.

Limitations and Challenges

Conviction voting is not suited to every governance context, and several limitations merit careful consideration.

Binary decisions — yes/no votes on protocol upgrades, parameter changes, or constitutional amendments — are poorly served by conviction voting. The mechanism was designed for resource allocation among competing proposals, not for up-or-down decisions on a single question. DAOs that adopt conviction voting typically maintain a separate mechanism for binary governance votes.

Urgency handling is problematic. Because conviction builds slowly, conviction voting cannot accommodate time-sensitive decisions. A critical security vulnerability requiring an emergency parameter change cannot wait for conviction to accumulate over days. Most implementations include an emergency governance bypass — often a multi-sig — for such situations.

Complexity presents an adoption barrier. The exponential accumulation curves, dynamic thresholds, and half-life parameters are mathematically elegant but difficult for non-technical participants to understand intuitively. Effective conviction voting requires sophisticated user interfaces that translate the underlying mathematics into clear, actionable information.

Strategic manipulation remains possible despite the attack resistance. A wealthy participant can maintain large positions on multiple proposals simultaneously, accumulating conviction across the board. While this is more expensive than a flash loan attack, it still allows plutocratic influence — just expressed through time rather than through a single transaction.

Parameter sensitivity means that the choice of half-life, threshold curve, and maximum conviction parameters profoundly affects governance dynamics. Poorly chosen parameters can make governance either dangerously fast or paralysingly slow. Finding the right calibration requires simulation, testing, and iterative adjustment.

Conviction Voting Versus Other Mechanisms

Understanding where conviction voting sits in the governance design space requires comparison with alternative mechanisms.

Compared to delegated voting, conviction voting replaces representative democracy with continuous direct participation. There is no need for delegates because every token holder can signal their preferences persistently. However, this means there is no mechanism for expertise aggregation — a limitation that delegated systems address.

Compared to quadratic voting, conviction voting addresses plutocracy through temporal weighting rather than mathematical cost curves. Both mechanisms reduce the dominance of large holders, but they do so through different channels. QV makes influence expensive; conviction voting makes influence slow.

Compared to optimistic governance, conviction voting is inclusively constructive — proposals require active support to pass. Optimistic governance is permissively constructive — proposals pass unless actively opposed. The mechanisms embed different assumptions about whether the default governance state should be action or inaction.

Design Recommendations

DAOs evaluating conviction voting should consider several practical guidelines.

Start with treasury allocation. Conviction voting excels at distributing resources among competing proposals. Grant programmes, public goods funding, and community initiative budgets are ideal initial use cases. Protocol governance decisions should use complementary mechanisms.

Invest in simulation. Before deploying conviction voting parameters to production, run extensive simulations using tools like cadCAD or TokenSPICE. Model various attack scenarios, participation levels, and request distributions to identify parameter ranges that produce healthy governance dynamics.

Build transparent dashboards. Participants need real-time visibility into conviction levels, threshold proximity, and the expected time to passage for each proposal. Without this information, conviction voting feels opaque and arbitrary. The Gardens interface provides a useful reference implementation.

Plan for parameter evolution. Initial parameter choices will almost certainly need adjustment as the community grows and governance patterns emerge. Build in a mechanism — potentially using a different voting system — to modify conviction voting parameters based on observed performance.

Combine with complementary mechanisms. Conviction voting should be one tool in a governance toolkit, not the only tool. Pair it with snapshot voting for binary decisions, multi-sig governance for emergency actions, and delegation for protocol upgrades.

The future of conviction voting lies in its integration with broader governance architectures rather than in its standalone deployment. As DAOs develop more sophisticated multi-mechanism governance frameworks, conviction voting’s unique properties — continuous participation, attack resistance, and proportional allocation — will make it an essential component of the governance design palette.

Donovan Vanderbilt is a contributing editor at ZUG DAO, the decentralised governance intelligence publication of The Vanderbilt Portfolio AG, Zurich. His work examines the intersection of governance design, institutional economics, and on-chain coordination.