Deep Rewarding

The Deep Rewarding project is a mechanism designed to ensure fair and transparent reward distribution within decentralized communities. This system addresses a common problem: how to evaluate and distribute rewards equitably among various contributors in a decentralized setting. These contributors include developers, designers, documentation writers, and community organizers.

This reward mechanism is inspired by Deep Funding (a mechanism of public good funding proposed by Vitalik and Kevin Owocki) and uses a multi-step process for contribution evaluation and compensation:

1. Role Assignment: Community organizers assign roles to core contributors (e.g., developers, designers, promoters). We use a role-management protocol (currently utilizing Hats Protocol) to formalize role assignments.

2. Credit Exchange: Contributors can issue Assist Credits (ERC1155) to each other. This credit exchange serves as a way to express appreciation and track collaboration. For example, if a developer helps another team member, they may receive 100 Dev Assist Credits as thanks. We are also building an AI Agent to remind members to exchange credits. Often, community members forget to express appreciation or to send tokens, and existing community token exchange tools have proven ineffective. For our prototype, the AI Agent monitors conversations in the Telegram community group chat. When a member receives sufficient help from others, the Agent suggests sending Assist Credits to acknowledge the support.

3. Data Aggregation: The Assist Credit transactions are recorded on-chain, creating a rich data set of contributions.

4. Dependency Graph Generation: An AI agent analyzes transaction data of assist credit to generate multiple dependency graph patterns. These graphs visualize the interdependencies within the community, highlighting which contributors provide the most impact to others. The agent can generate hundreds of variations for human review.

5. Human Validation: Community members perform spot-checks to validate the generated dependency graphs, ensuring the model’s accuracy. This process safeguards against errors in the automated analysis.

6. Reward Calculation: Based on the validated dependency graph, the system calculates reward ratios. For example, if a DAO has a 100 USDC budget, the system might allocate 30% (30 USDC) to one key contributor and distribute the remaining amount proportionally to others.

7. Automated Distribution: Finally, funds are distributed using Sprits Protocol, which enables transparent and immutable on-chain transactions.

This mechanism helps maintain a sustainable and collaborative DAO by:

• Incentivizing contributions through peer recognition.
• Reducing administrative overhead with automated AI-driven analysis.
• Providing a transparent process for fair reward distribution.

Deployment (Frontend, Contracts and Subgraph)

• Frontend: https://deep-rewarding.vercel.app/
• Assist credit contract (Sepolia): 0x2939D7Dd2dF88f901A2de4B282367134480bBdC2
• The Graph endpoint: https://api.studio.thegraph.com/query/34436/agentethereum_fractiontoken/0.0.3/graphql

Gaia Network: We utilize Gaia Network to distribute our custom RAG models that are familiar with our proposed mechanism. With tools provided by Gaia Network, we can host these RAG models and create knowledge resources efficiently.

Collab.Land AccountKit API: We use the Collab.Land AccountKit API to enable Assist Credit exchanges via the Telegram bot on behalf of users. It provides us with a smart wallet, making it easier to facilitate on-chain interactions through Telegram. Also, The Collab.Land AI Agent Starter Kit was useful to learn about how to build an AI Agent product

The Graph: We maintain index of the history of Assist Credit exchanges. The graph structure allows efficient querying and visualization of past transactions.

Hats Protocol: Hats Protocol is used to manage roles in the community. This ensures clear role definitions and responsibilities for contributors.

Sprits Protocol: We used the Sprits Protocol to distribute rewards to all community members in a single transaction, ensuring transparency and efficiency.

elizaOS: We use elizaOS to build and manage the AI Agent. This framework helps with creating and maintaining agent-based infrastructure for our mechanism.