An Economic Analysis of the Computable Protocol

Building decentralized systems presents new challenges that are not often seen in traditional software development. In particular, the adage of “move fast and break things” is no longer a viable strategy as we’ve seen time and time again how even a single critical security vulnerability can be very difficult for a project to recover from. Also, the success of these protocols depends on the design of economic incentives that encourage balanced participation and growth between different types of users in order to create a multi-sided market that will ultimately accrue value. These incentive structures can be difficult to modify once deployed since there is no centralized, governing authority.

Computable aims to create a decentralized data market that will incentivize the curation of high-quality datasets at scale, while providing trust and transparency around data privacy and usage. The protocol aims to be flexible enough to accommodate data markets for different industry applications. For example, certain markets may have the property that a handful of large players own the majority of the relevant data, while the success of other data markets may depend on many individual users making contributions over time. Each dataset has a unique token associated with it to incentivize curation and growth, and the participants, or “agents” in the network are grouped into the following roles:

The mechanism for determining whether a listing is valid is similar to a Token-Curated Registry. The dynamics of TCR voting can resemble other blockchain systems such as Proof-of-Stake consensus and decentralized oracles. However, for the rest of this post we will focus on macro features that are more specific to the Computable Protocol.

There is a bootstrapping problem since each dataset has its own token, as these tokens will have limited liquidity and be hard to value initially. A Bonding Curve is a contract that determines token price when buying/selling and acts as an automated market-maker for the token to encourage early participation. The diagram below illustrates how one might use a bonding curve to issue tokens. Note that there are separate buy/sell curves, where the sell price is lower than the buy price, to discourage short-term price manipulation while allowing for organic price discovery and liquidity since market participants can agree to trade tokens at any price between two curves.

Computable uses a bonding curve for issuing tokens when patrons deposit cryptocurrencies to the reserve, and when tokens are issued to makers for providing data. For the rest of this post, we will use the term “Network token” to denote the tokens deposited to the reserve. In practice, Network tokens could be either be a token that is native to the Computable protocol and shared across multiple data markets, or it could be a token that is native to the underlying blockchain (e.g. ETH). We will show that the shape and parameterization of the bonding curves has a significant impact on network growth.

For this analysis we use a linear bonding curve defined as follows:

Buyer — We model demand to query data in aggregate, rather than as individual agents. This demand is realized in the form of payment for queries during each simulation time step. The demand is a function of the number of listings in the dataset, with a predefined upper bound (market size) and bounded growth rate per time step. The fee to query data (in Network tokens) is split as follows:

Datatrust — These agents will process queries if fee that they receive is greater than their marginal cost of doing the computation.

Maker — We assume an upper bound on the number of makers (i.e. there are only so many participants that have high quality data to contribute to the dataset), and that the number of makers that will want to list their data is a function of the expected utility of being listed. We also assume that each maker can have at most one listing. Makers can take the following actions:

Patron — Can buy or sell Market tokens via the bonding curve:

For the analysis below, we make the following assumptions:

A few observations:

Out of these three factors, we suspected that the shape of the bonding curve was likely to have the largest impact, so we decided to dig a bit further.

We re-ran the above analysis and got the following results:

A few observations:

Within the context of a single data market, simulation allows us to analyze mechanism design as a distributed constrained optimization problem. More broadly, the framework can generate a reference set of parameters and initial conditions that are catered to serving data markets with all sorts of different properties and industry applications. The goal is to design a system that maximizes buyer demand, while maintaining equitable incentives between datatrust providers, makers, and patrons in a way that is statistically verifiable.

Economic incentives are of paramount importance for the long-term security and success of blockchain applications. Trying to reason about incentive mechanism design without simulation is tricky as the emergent properties of a network can be difficult to predict from local changes, and people often resort to making overly simplistic assumption about user behavior in order to get tractable results or closed-form solutions. Agent-based simulation can be a valuable tool for helping developers to validate security assumptions, and understand how value is created for network participants over time.

Related posts:

Every day we heard about someone who made something big, something that made us feel kind of jealous. Maybe because we have 3 years thinking about publishing an application, ebook or a simple post…

We are delighted to announce that we have partnered with Harmony, a next generation public, scalabale blockchain infrastructure for the decentralized economy. Hydro Protocol empowers dApp developers…

I travelled around Kerala for 10 days using all forms of public transportation and accomodation. Here is what I gathered. Filtered bottled water is easily and readily available in most stalls, even…