Description

Perhap the highest priority for humanity is to build an open source, decentralized, privacy protecting, and fact checking global platform that can merge the knowledge and sentiment expressed in public conversations with billions of people around the world. Would we unite to realize the principles of the Universal Declaration of Human Rights and the Earth Charter? This github project is for a demonstration of concept of collective human and digital intelligence.

Components

Components could include an app, a website, or a phone number at the front end, a language model in the middle, and a vector and graph database such as Neo4j in the back end. A fusion of the Neo4j LLM Knowledge Graph Builder and the Zep Graphiti knowledge graph engine could serve as a starting point for a demonstration of concept.

Running small language models (SLMs) on smartphones can yield a 10x reduction in electricity usage compared with running LLMs in data centers. Making the platform accessible by phones could reach 90% of humanity.

Multilingual text embedding models could create indexes of problems and proposals discussed in local to global deliberations. There are embedding models that are small enough to run on smartphones.

Features

• The language model would extract knowledge and sentiment from selected parts of public conversations to merge into a shared vector and graph representation.
• The platform, with humans in the loop, would continuously fact check statements merged into the shared graph.
• The platform would report which issues are discussed by the most people, the sentiment expressed on proposed solutions, and show how this sentiment changes over time.
• The platform would classify, from the local to the global level, the jurisdictions to which the issues relate.

Outcomes

After the demonstration of concept, the platform could be scaled to merge the knowledge and sentiment expressed in millions of simultaneous conversations with people around the world, becoming a form of collective terrestrial intelligence, CTI. The prompt.txt file in this project is a draft set of instructions to a language model for constructing a knowledge graph of people, claims, evidence, proposals, and sentiment about proposals.

This would enable citizens to deliberate and decide directly on principles and policies from the local to the global level.

Contact

John Kintree, jkintree@gmail.com

Resources

• Lenny's memory, a Neo4j project to give AI agents memory, https://neo4j.com/blog/developer/meet-lennys-memory-building-context-graphs-for-ai-agents/
• Memlayer is a project to integrate LMs with structured memory, https://github.com/divagr18/memlayer
• Gliner2 schema-directed data extraction, https://github.com/fastino-ai/GLiNER2
• Asta, scientific research assistant, http://asta.allen.ai/
• Reinforcement learning fine tuning of language models to improve knowledge graph construction, https://arxiv.org/abs/2510.15339
• OriginTrail DKG Edge Node, https://docs.origintrail.io/graveyard/everything/dkg-edge-node
• Building Agentic GraphOS, https://medium.com/@aiwithakashgoyal/building-agentic-graphos-the-16-layer-architecture-behind-production-ready-knowledge-graphs-9ca632bc74c5
• This paper describes another way to use LMs for knowledge extraction, construction, and retrieval. https://arxiv.org/abs/2502.09956
• Cognee uses BAML to improve data extraction from unstructured text such as conversations, https://www.cognee.ai/blog/integrations/structured-outputs-with-baml-and-cognee
• Dgraph manages very large databases by sharding, https://github.com/dgraph-io/dgraph
• Another outstanding open source project to integrate LMs with graph databases, https://github.com/TencentCloudADP/youtu-graphrag
• Modular GraphRAG with Rust, https://autognosi.medium.com/state-of-the-art-graphrag-rust-implementation-with-modular-ai-architecture-8c6baf5312cd
• Memori is another project for structured management of memories of conversations, https://github.com/GibsonAI/memori
• Estonia has traveled uniquely far towards digital democracy, https://e-estonia.com/
• SmallThinker language models were trained specifically to run on local computers such as smartphones, https://arxiv.org/abs/2507.20984.
• Cactus is an app at the Google play store for running language models on Android smartphones, https://github.com/cactus-compute/cactus
• The Your Priorities platform for participatory democracy is being used for an Online COP 30 community to prepare for a global referendum on a worldwide carbon tax, https://ypus.org/community/381
• People Powered has a ranked list of existing platforms for participatory democracy, https://www.peoplepowered.org/digital-participation
• Zep Graphiti knowledge engine, https://github.com/getzep/graphiti
• Mem0 might be more computationally efficient than Zep Graphiti for storing and retrieving memories from conversations, https://mem0.ai/ and https://github.com/mem0ai/mem0
• Neo4j LLM Knowledge Graph Builder, https://github.com/neo4j-labs/llm-graph-builder
• Neo4j LLM Knowledge Graph Builder demo, 4 minutes, https://www.youtube.com/watch?v=LlNy5VmV290&t=188s
• Audrey Tang, 37 minutes video on The Power of Collective Intelligence, https://youtu.be/2OGKzOWplyQ?si=aeJV7ql3JQZKCNoA
• Knowledge Graph of Thought paper, https://arxiv.org/abs/2504.02670
• Project Alexandria, to construct a graph of facts extracted from scientific papers, https://arxiv.org/abs/2502.19413
• Be The World Congress, video less than 2 minutes, such extreme optimism, it's funny, https://youtube.com/watch?v=ecU0BWzmDa0&feature=share
• Talking About A World Constitution, video 3 minutes, a work of art, https://youtu.be/xzICL4v8SOw?si=ss_8YaIogt0sNgei
• Interview on collective intelligence, 34 minutes, https://youtube.com/watch?v=0tCFtsfMFqA&feature=shared