Collective Intelligence

The goal is to scale collective intelligence in the number of humans.

What is collective intelligence?

Collective intelligence describes the intelligence of a group as a whole.
The intelligence of the group is much higher than the intelligence of any individual.
Scaling collective intelligence in the number of people means:
The more people in a group, the higher the intelligence of the group - without diminishing returns.
The system should easily scale to billions of people.

What is intelligence?

Intelligence is the ability to understand the environment.
Understanding is the ability to create a model of the environment to make predictions.
Making predictions includes the ability to predict the future of the environment in combination with your own actions.
In particular, deciding on your own actions such that the state of the future environment achieves an imagined goal state.

Why is collective intelligence useful?

Enabling a large number of people to develop this ability allows them to solve problems that go far beyond the capabilities of an individual human being.

Example problems are climate change, economic inequality, overpopulation or misinformation.
Teams, hierarchical organizations and representative democracies would benefit from collective intelligence technology as well. It could enable true deliberative democracies and flat large-scale organizations. It's qualitative decision making that voting alone cannot provide.

Scalable collective intelligence architecture

The building blocks correspond to the properties of intelligence itself:

Common Goals

The base for a group to work together are common goals.
A goal is a desired future state of the world.
Finding and agreeing on common goals can be done collaboratively using specialized voting systems (Quadratic Voting, Schulze Method, Downvoting).
The outcome of these voting systems needs to represent the prioritized opinions of the whole group.
Populism is to be avoided, and the system needs to be designed to be sybil attack proof.
If there is no consensus on the goal or mission, the group cannot organize itself.

Selecting goals helps to focus attention and activity on knowledge creation in relevant areas.

Knowledge base

The knowledge base is a data structure that is expressive enough to capture knowledge and can be used for prediction.
It mostly consists of claims, arguments, evidence, causality, systems and ontologies (Canonical Debate, Hyperknowledge).
The primary use of this information is to estimate the outcome of possible future actions.

The model is curated by individuals in a gamified, reputation based environment similar to Stack Overflow.
Modifications include addition, redundancy removal and conflicting claim resolution.
Everyone can propose modifications, the community accepts them if they are useful.
The curation system is designed to be attack proof against e.g. sybil attacks.

Abstraction in the knowledge base enables individuals to navigate and discover related knowledge.
An abstraction is a community curated summary of a subset of knowledge.
A summary represents simplified and reduced knowledge to make it easier to understand the overall picture.
Abstractions can summarize other abstractions, form hierarchies and can be used for high-level reasoning.
They are part of the knowledge base and are complements of ontologies.

Knowledge creation usually happens via communication.
At every level of abstraction there are communication systems to aid the creation and curation of knowledge.
These communication systems must scale in the number of participants, else overview is lost quickly (continuous online world cafe, hierarchy of 1-on-1 chats).
Communication history is not stored permanently, created knowledge must be moved into the knowledge base.

Decision making and execution

Everyone can propose actions to achieve a desired goal.
In the best-case suitable actions can be derived from the knowledge base itself, e.g. by reasoning about causality.
The knowledge base is used to predict if the action helps achieving the goal, potential side-effects, risks and uncertainty.
Additionally the community defines acceptance criteria for possible actions.
If all criteria are met, the group made a collective decision and can work on its execution.
Learnings from the execution is knowledge that is captured in the knowledge base.
Over time changed goals or new knowledge that leads to changes in the prediction can invalidate the decision.
Execution of the action needs to stop at this point and evaluate alternative actions.

Gamification and AI

Individuals can be represented by groups (Society Library, Guilds) or AI bots.
They play by the same rules of reputation as everyone else and are therefore evaluated automatically by the community.
AI can assist humans in many possible ways throughout the whole process, but will never be allowed to make any decisions itself.

Page created with