Starring: Tobin South, Leon Erichsen, Shrey Jain, Peter Maymounkov, Scott Moore, E.J. Glen Weil

Compiled by: Tiao

Translator's preface: This paper was published earlier this year. It is a protocol design for collaboration and self-governance within an organization, using a highly formalized language (mathematics). One interesting point for me is that the governance "token" in this protocol has no exchange value outside the protocol, but for the organization itself, it is the only "token". This makes people want to compare the many governance tokens and further think about the protocol design of LXDAO.

This article contains the first two sections of the article Plural Management. The last three sections will be published in the next issue.

1. Abstract

We introduce Plural Management, a model that partially replaces hierarchical organizational authority with plural mechanisms to allow networked authority to emerge. Participants gain influence by anticipating and fulfilling organizational priorities, and use this influence to set priorities and validate contributions, thereby facilitating a dynamic, merit-based power structure. We use open source software development as an example to illustrate this approach—it emphasizes valuable contributions and diligence without relying on hierarchical choke points; thus it enhances participation and helps achieve adaptive collective intelligence.

Multivariate Management Protocol Overview:

Organization members can dynamically earn governance points for their work. Through the issue board, members can use points to make quadratic contributions to determine the priorities of issues. These points are paid to members who resolve the corresponding issues through their contributions. Payments only occur after other members use their governance points to exercise their authority through quadratic voting. Organization administrators can choose to have members earn governance points by correctly predicting the results of votes, thereby rewarding those who perform due diligence on contributions and can foresee the preferences of managers on duty. This protocol achieves dynamic governance control without the need for a simple hierarchical structure through mechanisms that scale from small collectives of members to large organizations.

2. Introduction

The classic dichotomy between the rigidity of hierarchical organizations and the flexibility of flat organizations is a fundamental challenge for organizational design. Traditional hierarchies with clear command structures remain the orthodoxy, but are generally seen as inhibiting the dynamic capabilities that organizations need to thrive in today's complex environments. In contrast, flat structures, while inclusive and dynamic, often struggle to maintain coherent direction, momentum, and accountability, often falling into the "tyranny of structurelessness" (Ostrom and Hess, 2011). The classic alternative to this dichotomy is to use market mechanisms (Hamel and Zanini, 2020; Coase, 1995). However, an important role of the enterprise is to create internal partial public goods and take advantage of increasing returns, which the market is generally unable to provide efficiently (Samuelson, 1995). Therefore, Groves (1973) and Groves and Loeb (1979) advocated the use of public goods mechanisms to organize production within the enterprise as an alternative to hierarchy and market. However, these mechanisms are often considered cumbersome and impractical.

Recently, however, variations of public goods mechanisms such as quadratic voting (Lalley et al., 2016) and quadratic grants (Buterin et al., 2019) have become increasingly widespread and successful [1]. This paper aims to capitalize on these advances, return to Groves’ agenda, and outline a framework we call “pluralistic governance” to combine these mechanisms with other successful mechanisms to emulate many of the characteristics of organizational authority and collaboration without resorting to the simplistic hierarchical approach.

The traditional hierarchical management system is the backbone of modern corporate and organizational structures, with power dynamics following a top-down approach (Drucker, 1974). In such a system, employees tend to get promoted by working hard and aligning their actions with the culture articulated by those in authority, a practice sometimes derogatorily referred to as "cocking the sack." Critics point out that this practice can stifle creativity, reduce employee engagement, create decision-making bottlenecks, and often results in talent at lower levels in the organization being left with nowhere to realize their ambitions.

At the other end of the spectrum, lack of structured management has its drawbacks, such as the “tyranny of structurelessness,” where a lack of clear roles and responsibilities can lead to chaotic, inefficient, informal, and often unaccountable power structures (Friedman, 2007). Striking a balance between overly rigid hierarchies and a complete lack of structure has always been a complex task. Several innovative management approaches have been proposed and widely implemented, such as flat organizations that minimize layers (Laloux, 2014), Holacracy, which distributes decisions across overlapping teams (Robertson, 2015), and Sociocracy, which emphasizes consensus governance (Buck and Endenburg, 2012). These models all attempt to address the limitations of traditional hierarchies in more egalitarian and adaptive ways (Rothschild and Whitt, 1986). However, none of them have the mechanistic clarity of markets or hierarchies, which undermines their ability to avoid the challenges highlighted by Friedman. We aim to fill this gap by leveraging advances in multivariate mechanism design.

In our proposed model, governance points are a dynamic ledger of contributions and influence. Internal, non-financial points are initially allocated based on role or past contributions, and are subsequently earned through direct contributions and triage of situations. The expenditure of these points is constrained by a quadratic cost function when prioritizing and approving contributions to avoid over-dominance by those with greater authority and to achieve optimal public goods outcomes. Prediction markets are used to encourage those with limited authority to act as “analysts”, helping those with authority to triage and triage contributions, while the dynamic evolution of the priority of unresolved tasks acts as a dynamic auction-style bounty system, ensuring that tasks are resolved in a timely and prioritized manner. Therefore, in addition to presenting the specific design we use to illustrate this structure, we also aim to propose a general structure for implementing such a composite by combining different mechanisms.

The rest of this article is organized as follows:

Section 2 introduces the multi-governance protocol, describing in detail the high-level aspects of the system, the roles in the ecosystem, and the process of earning and spending governance points. It then discusses the practical application of multi-governance systems in open source software development, showing how they can help solve the long-standing governance problems of open source projects as they scale. Section 3 elaborates on this, describing a more detailed and easy-to-implement technical version of the protocol. Section 4 provides a detailed analysis of the protocol's features, studying voting, predictive behavior, and the optimal choice of parameters in the system. Finally, Section 5 explores some of the implications of the protocol and highlights the open problems and future work it raises.

3. Multiple management protocols

There are three roles in this ecosystem: Workers, who make direct contributions to the organization; Managers, who determine what work is important and whether the quality of a work is up to standard; Administrators, who can decide system properties to influence behavior. Importantly, a person can play multiple roles at any time and establish connections with multiple people according to these roles; there are no fixed roles, and organizational members are encouraged to play different roles when facing different individuals.

Instead of assigning everyone a set of hierarchical roles, everyone in such an organization has a set of management credits for the organization. These credits allow individuals to exercise authority in decision-making and be recognized for their contributions. We will walk through the steps for earning and using credits. These credits are specific to a particular organization, project, or community and have no value outside of it; in this sense, they are similar to "community" or "artificial" currencies (Blanc, 2018). As we will discuss later, these credits cannot be traded externally, and they are only used to control the flow of dynamic management potential.

Imagine an organization that has an issue board that lists all the major tasks or initiatives that need to be completed (similar to the open source issue tracker in Section 2.1). Individuals with admin points can set priorities for the organization by assigning a priority to an issue using their points. The priority of an issue is not simply the sum of the points assigned to it, but may be matched through a matching pool (provided by individuals in the admin role), which is consistent with the actual application of quadratic funding, which we will discuss further below.

Individuals can act as workers and provide solutions to issues in the form of contributions. If the contribution is accepted, the worker will receive points proportional to the total points assigned in the priority setting. Essentially, from the worker's perspective, the "bounty" attached to an issue may increase over time until enough rewards are provided to compensate the worker for the cost of solving the issue. This is like a reverse Dutch auction, although there is no guarantee that the reward will increase over time.

Once a contribution is made, it goes to a Contribution vote. Individuals can spend management points to vote on whether the contribution should be approved. If the vote passes, the worker will receive a reward; if not, the issue goes back to the board (the manager can increase the priority to offer a higher bounty). This voting uses a quadratic mechanism to ensure a balanced impact between individuals with different numbers of points.

In addition to voting, each person can choose to “bet” the amount of points they use to vote, predicting whether a contribution will be accepted or rejected. If the prediction is correct, this bet will receive double the vote points. With vote prediction, individuals can be rewarded for correctly predicting the wishes of the entire community. We introduce a prediction subsidy parameter that administrators can set for each contribution vote to reduce the cost of voting and increase the rewards for betting. By default, it is completely unprofitable to vote and then bet on the vote. However, in many cases, administrators may want to increase the subsidy to give individuals who can predict the needs of the community a chance to gain authority. For example, providing subsidies can incentivize individuals with fewer management points to participate in votes that would otherwise be too costly for them. This means that more people are doing due diligence on contributions. If there are many contributions in a large organization, betting on votes is similar to rewarding individuals for administrative review to “triage” contributions, thereby giving important divisive votes to managers with more authority.

Combining these two systems of quadratic agenda setting and hybrid voting-prediction, we are able to create a dynamic governance system where contributions are rewarded based on their public good demand when the broader organization collectively approves them, and individuals who have a deep understanding of or model of community preferences are rewarded and empowered for supporting the governance process.

Figure 1: Key components of the multi-governance workflow. Any member can be a contributor or manager, exercising authority by spending points, which are earned through contributions and correct voting predictions.

3.1 Open Source Applications

While polyglot governance protocols can be applied to a wide range of organizations and communities, they are particularly relevant to the world of open source software and other areas where peer production is prevalent (Benkler, 2017). Far from a niche industry, Git-based open source software powers over 93% of modern software applications (Daigle, 2023) and has been operated through a community governance model in which contributions in the form of code are evaluated for their quality and relevance before being merged into existing work. Despite these important contributions, the governance and management challenges facing open source communities are well known, most notably described by Eghbal (2020), including the following:

Although the contributions of open source contributors are recorded, it is difficult to track and trace the recognition of contributions because the relationship between the contributions and higher-level goals is unclear and the value is unclear. This reduces motivation and sustainability.

While contributions to open source projects are often open and participatory, their management (often referred to as "maintenance") is often in the hands of a "benevolent dictator for life", which contradicts its underlying democratic values ​​and leads to people with different opinions "forking" the project, fragmenting efforts.

Worse, the inability to leverage distributed participation to assist in governance places a heavy burden on project maintainers, who start projects highly motivated but have to maintain quality over the years, forcing them to identify and triage an increasing number of contributions of questionable quality with little community support.

Especially as a project grows and becomes more widely used, its potential directions for improvement grow exponentially, and the improvements most needed by users often lack clarity, resulting in projects with too many features but not enough usability.

By providing clearer direction and greater empowerment for contributors, multi-management can help founders gradually transfer management authority to those who have proven their value in the community through code contributions, diligence, or support. Because this model is lightweight, iterative, and autonomous, it is well suited to the agile environment and tools commonly used in open source communities.

Take GitHub, the most popular open source hosting platform. For any project, there is a code repository set up by an administrator or maintainer, and any contributor can be listed as a member. This repository comes with an "Issues" board (this is very similar to the issue board we described, but without the clear numerical priority). Anyone on GitHub can create contributions on this board in the form of a "pull request" (PR) to deal with one or more open issues. After discussing the pull request in the comments, the community can decide to accept or reject it, and then the maintainer can add or "merge" the contribution to the code repository. With minimal changes to the workflow to use multi-management, any maintainer or administrator can set a priority label on GitHub and associate it with a price expressed in points, thereby driving more contributions to their code repository as a first step to eventually improve their low average bus factor (i.e. the number of key people on the project) (Metabase, 2022).

It's worth noting that in general, while contributions are usually code, anything can become a PR. For example, if someone is going to be appointed as a social media manager for a project, an issue will be raised stating the need for a social media manager. Once someone is appointed, the new social media manager can raise a simple PR to add their name to the community record. If the community votes to approve the new role, the social media manager will receive additional management points reflecting their new role.

3.2 A Simple Use Case

One specific application of polyglot governance outside of the usual open source context is the book Plurality, a Git-based experiment in collective authorship. Initiated by E. Glen Weyl and Audrey Tang, 50 members from around the world contributed to the book Plurality: The Future of Collaborative Technology and Democracy without any expected reward. The project uses a polyglot governance agreement to gradually transfer ownership of future improvements to the book, including content updates, translations, and further links to related materials. Over time, the most meaningful contributors will help lead not only the book, but also the field of study itself.

Imagine a political economy undergraduate from a little-known university. She sees a typo, opens an issue and submits a PR. This action doesn't earn many points during voting, but the small amount of points allows her to participate in priority setting. Inspired, she continues to look for opportunities to contribute, and finds that her paper work can help with an open issue around the addition of content to a chapter. She submits a PR on the issue board, adds some key references cited in the book, and is rewarded with a large number of points.

Given the challenges of inclusion in contemporary higher education settings, such students may never have the opportunity to engage in such work without the permissionless and community-judged structures that diversity management provides ( Gvozdanović and Maes, 2018 ).