Introduction

Traditional top-down data management systems have provided firms with significant advantages, but fall short in addressing the dynamic nature of knowledge within non-hierarchical groups, including DAOs. Such organizations need to enact collective governance over knowledge. In addition, existing approaches do not enable effective knowledge-sharing between groups, which is essential for collectives seeking to address shared problems, expand their networks or make use of an adjacent community’s expertise.

The KOI project is a Metagov initiative in collaboration with Block.Science and aims to reimagine knowledge management by developing and implementing an advanced system using Reference Identifiers (RIDs) within the graph-based KOI (Knowledge Organization Infrastructure) architecture developed by our partner, BlockScience. This initiative will enhance the ways in which collectives manage, share, and build upon their knowledge, fostering more effective and cohesive collaboration.

KOI will also make Large Language Models (LLMs) more directly useful to collectives and DAOs. While LLMs offer new capabilities, they can be prone to unreliability and raise issues regarding confidentiality in an organizational context. Retrieval-Augmented Generation (RAG) systems provide more specific and reliable data for LLMs to draw from but, on their own, do not enable groups to specify how knowledge should be used in different contexts or to collectively curate sets of knowledge as a group. By adding Reference Identifiers (RIDs) to a graph of connected knowledge objects, KOI provides the means for groups to specify how LLM’s access specific sets of knowledge, whilst keeping some knowledge confidential.

Our project will address the unmet knowledge needs of collectives and communities of practice, including DAOs. Collectives, DAOs, and communities of practice often struggle with static, inflexible data structures that do not adapt to the evolving nature of knowledge that characterizes permissionless or voluntary contexts. Existing graph databases are designed primarily for companies using business process management and are not sufficiently adaptable or governable by collectives. The KOI introduces a dynamic, mutable approach that allows groups to organize and reorganize knowledge objects as their needs and contexts change (expressing beliefs about relations between knowledge sets). This flexibility is crucial for maintaining the relevance and utility of shared knowledge.

The KOI’s RID system (defined below) also facilitates governance practices in relation to knowledge, allowing groups to define and agree on the structure and boundaries of their collective knowledge. Metagov’s aim is to develop tools for making these processes straightforward and engaging. For instance, using the KOI, a group can create an empty container that members fill with knowledge objects (via RIDs), collectively defining and agreeing on what belongs there. Or a member can decide to link one set of knowledge with another, which may require a process for others to explore or dispute that link. Such governing will ensure that the knowledge infrastructure is continuously shaped by the contributions and consensus of the group.

Methodology

Low Funding - $1000: Documentation of the current RID system + conceptual integration with the Canvas-based architecture being developed at Blockscience

• Document the current functionally RID system for the Metagov Instance of the KOI.
• Determine the core use-case of the canvas based interface based on Ellie Rennie’s research at RMIT

Mid-Funding - $3000: Create a persistent canvas connected to the Metagov KOI instance

• All of the above
• Use the defined use-case to create a proof of concept of the canvas-based interface with an LLM.
• Create a persistent host location of this canvas
• Allow this canvas to update and add additional context into the KOI upon use.

High-funding budget - $6000: Test and Iterate with the Metagov community

• All of the above
• Adding governance of the underlying datasets to allow groups to dynamically work within the Canvas
  • User permissions for updating, editing, or viewing documents in KOI
  • Privacy permissions for those who don’t want their work included in KOI
  • Draft canvases, private canvases, or public canvases.
• Upon completion of the Proof of Concept, we would work with the Metagov community or within the KOI research group to test the tool capabilities and run a user feedback session
• The remaining time would be used to update the KOI canvas based on that feedback.

The budget here is allocated for development time from Luke Miller which will be documented via a google sheet and paid out through Metagov’s page on Open Collective. The contributions from Ellie Rennie, Michael Zargham, and Eugene Leventhal will be funded by their respective organizations RMIT, Block.Science, and Metagov.