Spiel.com Pitch Deck 2024
by Murat Ayfer, @mayfer

AI assisted governance protocols

Civilization runs on hierarchies of evolving protocols, but participation has never hit mainstream.

Corporate law, city council, senate are examples. There is self-similarity at each scale with how protocols govern resource allocation & permissions.

Governance is a nested, almost fractal concept.

Democracy is much more than voting, it's a set of protocols followed by a hierarchy of institutions.

How a council meeting is structured with roles & parliamentary procedures has more effect on changing the rules than the votes themselves.

By distilling the priors into a DSL, we can attach open-ended governance to any software that uses permissions & roles. Hierarchical structure is a bonus.

Using a few simple priors, it is easy to represent all popular governance structures, from direct democracy to representative democracy, to oligarchy, to dictatorship, to liquid democracy.

In fact, even these systems seem overly simplistic when you realize the full potential of algorithmic policies based on metrics.

Difficulty #1: Participation in changing the protocols is limited to people with programming knowledge of the DSL.

With traditional software (Reddit, Slack, Discord, Twitter, you name it), changing the core rules is limited to engineers (est. 0.000004% of users).

Even in legal systems, even though the laws are written in natural language, the process of changing them is limited to those who understand the formalism.

Ideally we'd bring participation up to the standard 1% - 10% rate

Even if the DSL is implemented as a no-code UI,
proposing / reviewing / voting on changes is still a high bar.

Solution #1: LLM's can convert natural language to the DSL and take care of the heavy lifting.

While still requiring input from humans (reviews, or even just votes!), scripting in a small, well-defined context is exactly what LLM's are good at.

Bonus advantage: the LLM can also generate data visualizations & transparency tools that help humans keep finger on the pulse

Votes are often too opaque. Imagine how vote distributions change between the general population vs. specific roles. Subjective projections of vote stats can be incredibly insightful.

An LLM that's already tapped into the governance DSL allows users to get visual answers to custom questions, which can and should guide new policy proposals.

Difficulty #2: "who is it for?" communities only need antifragile governance at large scales and long timelines.

The need for governance isn't felt until the community outgrows the moderation efforts. For this reason it can be difficult to find early adopters.

Solution #2: start out as toys & games, iterate on tooling, then move to more serious applications.

The first game is an MMO. It involves AI NPCs for fun & initial traction. Ultimately the goal is to build formal protocols that allow collective decision making & resource allocation for civilization building.

The educational implications of running civilization simulations have the potential to be massive.

Bonus research opportunity: the AI bots have the potential to be valuable for agent R&D -- a safe playground where bots can cooperate & compete alongside humans, within a system that mimics the real world.

If the governance tooling works in games, it can be used for any community that needs to manage shared resources.

Developing the governance tooling as a standalone framework that can be integrated into any other software is the long term goal.

I've tried to keep it as brief as possible.

I can continue with more technical details & key features (such as forking, templates, extensibility etc.) of the governance framework if you'd like to discuss further.

Thank you for your time.

by Murat Ayfer, @mayfer