# Workpackages

## Data

### Deployment

#VOICE will be deployed initially in local community contexts across London. We already have a range of communities in London who have committed to run community assemblies, using #VOICE as the data capture and synthesis tool in a collaborative and deliberative environment. Factory Labs will be organising, coordinating and facilitating these events.

During these events, #VOICE will onboard non crypto natives with familiar web2 tooling (SMS, email) into a web3 environment for consensual granular data collection, RAG LLM training and governance.

### Expert Models

#VOICE is a path to collaboratively built, human trained Expert Models. A pillar of our thesis on collaborative and decentralised AI, is that everyone has expertise in something, and our goal is to help users to identify and tap into that expertise.

Using a combination of tools which enable emergent consensus, semantic ballot voting and ongoing conversations with the LLM, we will be to identify and elevate expertise in any given hash tag. These experts can be rewarded with increased governance power and act as a check and balance against poor data entering the model.

## Governance

Sybil protection is of the utmost importance to the efficacy of #VOICE.
Using an iterative approach, we will move towards a quality set of initial high trust community contributors. Over time and via a mixed methodology of data science, onchain analytics and trust based distribution of NTTs (Non Transferable Tokens) we will develop sybil resistance user sets.

### Reputation Based Governance

Reputation is determined by the quantity of contributions that make it through the layers of Governance to the GRC20 FACT hypergraph.

Reputation is not universal. A high FACT reputation on the #Boston_Celtics, does not automatically qualify you for #NBA reputation. All reputation must be earned and is non transferable.

### Aragon OSx Plugin Development

Further alignment can be achieved through our planned integration with Aragon OSx. Our intention, is to develop a plugin that uses OSx for the onchain execution layer that connects our tools to GRC20.

## Token Economics

As Token Economics experts, Factory Labs designed non extractive incentive mechanisms to motivate users to continuously share their data in a win win scenario.

Contributors data should inherently be private and only inferred by an LLM when needed. Contributors should be paid for contribution, but if there data is a high quality and has a higher inference frequency than other contributions in the same set, they should be paid accordingly.