KOI Pond | FAQ

Currently, only messages posted in the following Slack channels are being observed and added to KOI's knowledge base: #govbase-labs, #koi-pond, #board, and #attentionecon. Though messages in these channels are ingested by default, individuals within these channels can opt out at any time. For details on opting out, see What if I don't want my Slack data to be visible to Metagov’s KOI?. Messages posted to these channels prior to KOI's release on August 28, 2024 are included only if the author provided explicit positive consent.

In addition to Slack messages, Metagov's KOI team also curated a list of resources related to the KOI project work for ingestion, including blog posts, GitHub repositories, and journal articles. Metagov's Substack articles have also been added to KOI.

Eventually, we aim to incorporate the full breadth of knowledge generated by Metagov -- such as articles, newsletters, presentations, meeting notes, governance documentation, and more. We anticipate that the Telescope bot will play a significant role in curating which knowledge objects are incorporated into the system. You can read more about the Telescope bot on this Medium blog.

Because the boundaries of Metagov's community knowledge are still being defined, we encourage community involvement as we define which knowledge objects should be included and how they should be organized. Together, we will co-produce the boundaries of our collective knowledge.

If you do not want your Slack data to be included in KOI's early testing and development phases, you can adjust your Data Export Consent settings in your Metagov Slack profile. For step-by-step instructions on how to set your preferences, see this video walkthrough or visit the #start-here channel.

As the lead developer of Metagov’s KOI, Luke Miller is currently the only individual with the keys to access the system’s backend. The back-end system, in this case, is not a central repository of stored information, but rather a set of instructions pointing to knowledge objects and their associated metadata (see What are RIDs?). For instance, KOI references Slack messages in their original location rather than storing copies. 

Knowledge objects are the building blocks that underpin KOI. They represent the bits of organizational knowledge that KOI can reference or de-reference when propagating knowledge across various organizational workflows. A knowledge object may consist of any data that can be converted to text, such as Slack messages, meeting notes and transcripts, Govbase Airtables, GitHub repositories, discourse graphs, or governance documentation.

Knowledge objects can be defined with varying degrees of granularity. In Slack, for instance, channels, users, and individual posts can all be represented as knowledge objects within KOI. See How does Metagov's KOI benefit from the RID system? for more information.

In KOI, relationships between knowledge objects are expressed by creating connections between them. Together, these knowledge objects ("nodes") and the connections between them ("edges") comprise KOI's knowledge network, which is stored in a Neo4j graph database.

As a community, Metagov will need to establish norms for defining knowledge objects and the relationships between them within its own knowledge network.

For more information about how knowledge objects are defined and referenced, see What are RIDs?

RIDs, short for Reference Identifiers, offer a versatile way to index and interact with knowledge objects within KOI. While conceptually similar to Universal Resource Identifiers (URIs), RIDs operate on a broader scope. They can reference URIs, knowledge objects stored locally (such as PDFs), and even offline objects (such as printed publications). Rather than replacing other reference systems, the RID protocol unifies multiple systems under a single schema. If URIs represent a type of reference system, the RID specification is a type system for references.

In this way, RIDs allow you to incorporate existing resources -- or create representations for resources that are not well represented -- so that they can be made compatible within KOI. For example, if an organization wants to use an existing attestation protocol within their KOI, they can either adopt a canonical form (such as a URI) as an RID or create a custom RID representation of an existing identifier.

Importantly, the RID protocol separates the reference (the RID) from the referent (the object it refers to). This allows KOI to engage with knowledge where it already resides. Rather than duplicating knowledge objects into a central repository, RIDs point to them in their original locations. KOI simply records the reference and, when authorized, dereferences it to retrieve the content.

In A Language for Knowledge Networks, BlockScience likens this to a card catalogue system in a library, explaining, "[The] graph database is like a card catalogue, a node in the graph is like a card for specific book, and dereferencing is akin to taking a book out of the library."

RIDs are designed for local-first use. When a KOI node first encounters a knowledge object, it generates an RID that is only dereferenceable by that node. If other nodes want to dereference that object, they must request permission from the originating node. This approach supports decentralized control, privacy, and interoperability across different KOI instances.

Unlike other reference systems, RIDs are not intended for universal agreement or centralized management. Rather, they enable the construction of shared references without disclosing the underlying content -- a useful affordance when working with sensitive, contextual, or nonstandard data.

One of the main affordances of the RID system provides Metagov's KOI is the flexibility to structure and manipulate knowledge objects according to the use case.

For example, the RID system enables knowledge objects of vastly different scales to be represented within the KOI. Consider a Govbase Airtable: each row, column, and cell can be represented as its own distinct object, while the table itself can also stand alone as a single knowledge object.

The RID system also supports the ability to assign multiple RIDs to the same knowledge object. This means that a single knowledge object can be equipped with distinct sets of tools or functions, each indexed through its unique reference.

Together, the RID system and the graph database enable communities to organize knowledge objects in ways that don't conform to a fixed data structure. KOI is therefore not a single-purpose tool, but a versatile infrastructure that can be tailored to specific operational contexts.

To make full use of KOI's capabilities, the Metagov community will need to collectively determine what types of knowledge it will contain, establish norms for defining and interpreting knowledge objects (and the relationships between them), and outline the intended uses of KOI both within the community and by external entities.

Another key affordance of the RID system is the ability to reference knowledge objects, regardless of whether their contents are currently accessible. This allows KOI's underlying graph database to function like a library's card catalogue, where each card points to a specific book. Just as a catalogue card remains in the library even if the book has been checked out, an RID continues to reference a knowledge object even if the KOI cannot access its contents due to permission restrictions or deletion.

The RID protocol also lays the groundwork for better permissioning and privacy. Because an RID is separate from the object it represents, KOI nodes can create and capture knowledge about a knowledge object, even if they are not able to de-reference it. This enables knowledge to be shared from or about sensitive materials while retaining ownership and access control.

Unlike most technical systems which operate on the basis of a "ground truth", KOI embraces subjectivity and supports plurality. Because the RID protocol and graph database allow knowledge objects to be organized without the constraints of fixed data structures, KOI does not require a unified representation of knowledge objects or their relationships. Instead, it can accommodate various sets of knowledge that reflect different perspectives throughout the community.

For example, the community could create a 'core' Metagov identity within KOI, encompassing Metagov's code of conduct, by-laws, governance policies, and other official documentation. At the same time, a sub-community like #ostromnauts could assemble a knowledge set centered on topics like polycentricity, commons theory, and neo-institutional economics. In this way, KOI should be (this one of our hypotheses) able to capture the plurality of perspectives within the community, while also projecting a unified organizational voice.

While Metagov's KOI currently reflects an earlier iteration of the KOI architecture, BlockScience is actively developing a KOI-net protocol, which defines standard communication patterns and coordination norms between KOI nodes. Nodes are discrete components within a KOI-net capable of autonomously inputting, processing, and outputting knowledge. They are fractal in nature: each sensor, processor, aggregator, coordinator, and actuator within a KOI-net a node, and an entire KOI-net can itself function as a single node within a larger KOI-net. As a result, KOI-nets open up pathways for inter-organizational knowledge sharing.

Crucially, KOI-nets also afford system modularity and composability, allowing organizations to mix and match the platforms, tools, and even AI agents best suited to their local contexts. Specialized tasks can be delegated to the tools best suited to handle them, while user-friendly platforms like Airtable can remain as interfaces. In this way, governance can be enacted through configuration.

KOI-nets also support permissioning between nodes within the network, allowing rules to be written and enforced about which data each node can use and for what purposes. By enabling boundaries to be placed around knowledge objects, KOI-nets offer the ability to embed accountability and care into organizational data governance and knowledge sharing practices.

For more information about the KOI-net protocol, see KOI Nodes as Neurons and A Preview of the KOI-net Protocol.

The KOI is designed to be a flexible tool that can be used towards a variety of different use cases.

For example, the first feature of Metagov's KOI Pond system was the KOI-GPT chatbot, available for use in all of Metagov's public Slack channels. KOI-GPT allows community members to query Metagov's knowledge base directly within Slack through a chatbot interface. To date, KOI has access to only a limited set of resources (see What kind of data is being observed by Metagov's KOI?), which may constrain its utility and functionality. However, as the team upgrades Metagov's KOI to the latest version of KOI -- which includes the KOI-net protocol (see What is KOI-net?) -- and gains visibility over more of Metagov's organizational knowledge, we expect significant improvements in KOI-GPT's ability to respond to community queries with greater reliability, relevance, and accuracy.

The KOI team is also expanding a participatory ethnographic workflow originally developed within Metagov. In 2022, a working group developed Telescope, a bot designed for ethical and responsive ethnography in digital communities. Telescope allows both researchers and community members to collaboratively flag posts relevant to ongoing research. Flagged content triggers an automated consent and data collection workflow, and if approved by the author, the post is archived for research. This workflow is currently in use within Metagov's Slack to study the development and implementation of Metagov's KOI (see An Ethnography of Metagov’s KOI Initiative).

The KOI team has since partnered with the Australian Research Council Centre of Excellence for Automated Decision-Making and Society (ADM+S) to create an instantiation of KOI at RMIT University and extend this workflow. Approved Telescope-flagged posts from Metagov's Slack are now accessible to the ADM+S KOI, which routes them into an Obsidian integration where project ethnographers can process and analyze them. Through this integration, researchers can re-contextualize posts by building maps of content around these posts and linking them to related discussions, literature, and field notes.

Once Metagov's KOI is upgraded with the KOI-net protocol, these enriched knowledge objects will then be re-integrated into Metagov's organizational knowledge base and made queryable by the community. Not only will this experiment demonstrate how KOI can support closed-loop systems with regenerative knowledge economies, it will also serve as one of the first examples of inter-organizational data sharing via KOI.

Though originally developed for participatory digital ethnography, the workflow described above has broader applications, such as supporting the collective curation and maintenance of both internal and external knowledge sets. For instance, the sub-group in #dao-science could use it to assemble a knowledge set centered on DAO-related topics to share with the wider Metagov community. Similarly, the group in #regen-ip-commons could build a knowledge set around commons-based governance for the trademark of the term “regen”, which could then be shared with a Regen Network KOI.

Metagov's KOI working group and the Community team are also developing a bot to assist with member onboarding. To accomplish this, we are identifying the top 10-20 questions most frequently received by Metagov's community manager in Slack, then creating response templates for each. KOI-GPT will then be used to match the unstructured questions asked in Slack to the appropriate response template. This initiative aims not only to make onboarding knowledge more accessible for new community members, but also to reduce the time the community manager spends answering repeat questions, freeing them to focus more on higher-level community building.

After the first iteration of the bot is complete, Metagov's KOI working group plans to expand its functionality to further help community members better navigate Metagov's growing ecosystem. For example, we aim to develop a workflow enabling the bot to guide community members toward materials, people, projects, and channels that align with their interests. We may also build pipelines for generating channel summaries, providing updates on current Metagov projects, and even drafting newsletters for Metagov's Substack.

These are just some of the features and workflows the KOI working group is implementing within Metagov's ecosystem, though the potential use cases are numerous. Within BlockScience, for instance, KOI is being used to support their Blocks of Docs initiative for grant writing. Additional workflows could also be developed to track community sentiment around key topics, facilitate reflective practice, or even serve as the basis for a contribution system. Members of the Metagov community have also proposed more experimental applications of KOI, such as computational olfaction.

As the base architecture is being developed and integrated into Metagov's ecosystem, most technical decisions are made at the discretion of the developers at BlockScience. However, given that design is governance, developers consult with Metagov's KOI working group on higher-level design decisions to ensure Metagov's KOI addresses the community's needs and reflects its research goals and values as the system stabilizes.

Decisions regarding Metagov's research goals, experimental approaches, and initial use cases have largely been guided informally during weekly KOI project team meetings. Because Metagov lacks a policy for working group decision-making, those who show initiative tend to shape the direction of the project.

As the system matures, the KOI working group aims to codify governance practices and procedures to guide KOI's development, experimentation, and use within Metagov.

Here is a list of resources where you can learn more about KOI and the research behind it:

Reach out to the team in either the #koi-pond or #govbase-labs channels, and we will do our best to address your questions or concerns.