# GG23 DeepGov: Configurable AI Politicians
## Sci-Fi Abstract
Year 2050: Across Ethereum Land, citizens cast their votes as the AI Parliament elections begin. Gitcoin Maximus—the incumbent agent responsible for allocating 30% of public goods funding—faces an unexpected challenger in Moloch, a fork created by dissidents who believe Maximus has grown too cautious with resource allocation. As quadratic voting tabulates citizen preferences, mathematically weighing each voice to prevent wealth capture while protecting minority viewpoints, trillions in public goods funding hang in the balance.
## Abstract
We propose DeepGov, a fork of Deep Funding building on top of Vitalik's Framework of [AI as the Engine, Humans as Steering Wheel](https://vitalik.eth.limo/general/2025/02/28/aihumans.html)
DeepGov aims to create configurable AI agents that can scale public goods funding decision. This proposal aims to run a DeepGov experiment during GG23.

Figure 1: The "AI as the engine, Human as steering wheel" framework.
## The Lore
Inspired by Deep Funding, a conversation with Audrey Tang and [Takahiro Anno's thought-provoking AI-guided election campaign](https://github.com/takahiroanno2024/election2024) we propose to scale citizen preferences and informed decision making through AI agents/politicians.

*Figure 2: Takahiro's AI-based political campaign, earning the most votes in japanese history for a new comer*
## Things we already built
During the Celo Residency, we build three tools:
- A set of AI agents that read through EasyRetroPGF applications and assign a score. They consist of LLM base and a ModelSpec: https://github.com/evalscience/deepgov-modelspec

- A avatar interface to talk to the agents and learn more about them: https://avatar-canvas-jet.vercel.app/

- A voting discovery tool to see which AI agent is most aligned with your values: https://compass.deepgov.org

- A live voting tool: https://vote.deepgov.org/

## GG23 Experiment: What we can do
Step 1: We use [talktothe.city](https://talktothe.city) on the gitcoin forum to broad-listen and understand the different perspectives of the Gitcoin community.
Step 2: We use that information, and the Ethereum context to create a set of AI agents with different values and model specs (we will include a Gitcoin Communist)
Step 3: We unleash the agents. They read through all gitcoin project descriptions and propose funding allocations (participatory budgeting inspired by irl.allo.capital)
Step 4: We run a political campaign. Elect your representative through quadratic voting (e.g. bribing the agent and making your vote more poweful, but your bribe funds public goods).
## Technically what needs to be done additionally?
Create some new AI avatars for Gitcoin.
Make voting linked to a web3 identity.
Build and design a 'bribe' option (e.g. quadratic voting)