Blockchain Upgrade as a Coordination Game P2PFISY 2018 Stephanie - - PowerPoint PPT Presentation

Blockchain Upgrade as a Coordination Game

P2PFISY 2018 Stephanie Hurder Prysm Group July 27, 2018

Blockchain organizations are struggling with how to design decentralized governance systems

• Salient example: EOS governance

Long history in economics of studying social choice: how best to design a decision-making system for a group (Arrow (1951), Mueller (2003)) Unlike most standard governance settings, blockchain can experience hard forks

• An unhappy group of users can secede from the chain and start its own chain
• Bitcoin had many major policy-driven hard forks (Cash, Gold, Private, ...)
• The DAO hack → Ethereum hard fork (Classic)

The founders of a blockchain platform face two novel questions

• 1. How to evaluate when hard forks are socially optimal and should occur
• 2. How to design a governance system so that forks only happen when they are socially
• ptimal

P2PFISY 2018 | July 27, 2018

Motivation

We build a game-theoretic model of policy choice on a blockchain

• Community is choosing between a status quo policy or a policy upgrade
• Blockchain exhibits network effects: I get more utility from a blockchain when there are

more other users on my chain (Shapiro and Varian (1999))

• Users who don’t like the policy choice on the main chain can fork

We classify under what conditions a hard fork is an equilibrium, and in what cases it is socially

• ptimal for a community

We explore in what circumstances standard voting mechanisms (majority vote, quadratic voting) would guide the community to the socially optimal outcome

P2PFISY 2018 | July 27, 2018

This Paper

• 1. Hard fork is an equilibrium if the policy change is sufficiently big, and if both policies have

sufficient support

• 2. Hard fork can be socially optimal, but can also occur as coordination failures (everyone has ex

post regret)

• 3. Once hard forks are an equilibrium, which governance system is better is sensitive to

hard-to-observe population parameters

• 4. The most realistic method to prevent suboptimal hard forks is to change the policy proposal

process

P2PFISY 2018 | July 27, 2018

Results

➔ The math ➔ Results under individual choice ➔ When is forking socially

• ptimal

➔ Performance of standard governance systems Unit mass of users vij (x): Value to user i of participating on chain j with fraction of the community x

• No utility without other users: vij (0) = 0
• Network effects: v’ij (x) >0

Two types of users:

• Type A: Prefer status quo policy given fixed x (Fraction )
• Type B: Prefer upgrade policy given fixed x (Fraction 1-)

P2PFISY 2018 | July 27, 2018

➔ Model set-up ➔ Independent strategic chain choice ➔ When is forking socially

• ptimal

➔ Performance of standard governance systems What happens in a community with no governance? Individual users choose between two chains: status quo and upgrade qj: Fraction of community on status quo chain s.t. type j is indifferent between status quo and upgrade (Lemma 1) Proposition 1: Nash equilibria are

• Everyone on the status quo chain
• Everyone on the upgraded chain
• If ∈[qB, (1 - qA)]:

○ Users who prefer status quo on status quo chain ○ Users who prefer upgrade on upgrade chain

P2PFISY 2018 | July 27, 2018

➔ Model set-up ➔ Results under individual choice ➔ When is forking socially

• ptimal

➔ Performance of standard governance systems Two common measures of social welfare:

• Total surplus maximization: Maximize the sum of the value of the

user base

• Pareto Optimality: Can’t make one group better without making

another group worse Results:

• Everyone on a single chain is always Pareto Optimal
• One single-chain equilibrium dominates the other in total surplus
• A hard fork can be

○ Nash Equilibrium for a range of (Prop 1) ○ Pareto Optimal for smaller range of (Prop 4) ○ Surplus maximizing for even smaller range of (Prop 5)

P2PFISY 2018 | July 27, 2018

❴

➔ Model set-up ➔ Results under individual choice ➔ When is forking socially

• ptimal

➔ Performance of standard governance systems Two common measures of social welfare:

• Total surplus maximization: Maximize the sum of the value of the

user base

• Pareto Optimality: Can’t make one group better without making

another group worse = Fraction of users who prefer the status quo

P2PFISY 2018 | July 27, 2018

1 Total surplus max. Pareto Optimal Nash Equilibrium

➔ Model set-up ➔ Results under individual choice ➔ When is forking socially

• ptimal

➔ Performance of standard governance systems Introduce one of two standard governance systems:

• Majority rule: policy preferred by larger fraction of users is

implemented

• Quadratic voting: policy that maximizes total surplus is

implemented (Posner and Weyl, 2015) Order of events:

• 1. All users start on single chain
• 2. Voting on which policy to implement occurs
• 3. Chain implements policy selected by governance process
• 4. Hard fork occurs (if desired)

Assume naive voting: users vote assuming everyone will stay together

• n a single chain

P2PFISY 2018 | July 27, 2018

➔ Model set-up ➔ Results under individual choice ➔ When is forking socially

• ptimal

➔ Performance of standard governance systems

P2PFISY 2018 | July 27, 2018

Neither leaves the other group’s chain Minority leaves Majority’s chain Majority leaves Minority’s chain Each group leaves other’s chain Majority’s chain has highest total value Minority’s chain has highest total value Hard fork has highest total value Bigger policy change

➔ Model set-up ➔ Results under individual choice ➔ When is forking socially

• ptimal

➔ Performance of standard governance systems

P2PFISY 2018 | July 27, 2018

Neither leaves the other group’s chain Minority leaves Majority’s chain Majority leaves Minority’s chain Each group leaves other’s chain Majority’s chain has highest total value Minority’s chain has highest total value Hard fork has highest total value

Forks never happen Forks are socially

• ptimal and

always happen

Bigger policy change

➔ Model set-up ➔ Results under individual choice ➔ When is forking socially

• ptimal

➔ Performance of standard governance systems

P2PFISY 2018 | July 27, 2018

Neither leaves the other group’s chain Minority leaves Majority’s chain Majority leaves Minority’s chain Each group leaves other’s chain Majority’s chain has highest total value Minority’s chain has highest total value Hard fork has highest total value

Forks never happen Forks are

• ptimal and

always happen Fork under majority rule

Bigger policy change

➔ Model set-up ➔ Results under individual choice ➔ When is forking socially

• ptimal

➔ Performance of standard governance systems

P2PFISY 2018 | July 27, 2018

Neither leaves the other group’s chain Minority leaves Majority’s chain Majority leaves Minority’s chain Each group leaves other’s chain Majority’s chain has highest total value Minority’s chain has highest total value Hard fork has highest total value

Forks never happen Forks are

• ptimal and

always happen Fork under quadratic voting Fork under quadratic voting

Bigger policy change

➔ Model set-up ➔ Results under individual choice ➔ When is forking socially

• ptimal

➔ Performance of standard governance systems

P2PFISY 2018 | July 27, 2018

Neither leaves the other group’s chain Minority leaves Majority’s chain Majority leaves Minority’s chain Each group leaves other’s chain Majority’s chain has highest total value Minority’s chain has highest total value Hard fork has highest total value

Forks never happen Forks are

• ptimal and

always happen Governance matters

Bigger policy change

Standard governance processes do not copy-paste well to blockchain if forks are an option

• Policy proposal process is just as important as the voting process

Assumptions that could be relaxed in future work:

• Non-transferable utility: What happens if some users can (credibly promise to) pay off the
• thers?
• Naive voting: What if voters anticipate the possibility of a fork and vote accordingly?
• Social planner’s objective: What if he/she wants to keep the chain together at all costs?

Open question: a more effective, yet practical governance mechanism

• Vickrey-Clarke-Groves mechanism is optimal but shown difficult to implement (Rothkopf,

2007)

P2PFISY 2018 | July 27, 2018

Implications and next steps

For questions or for more information on Prysm Group, I can be reached at stephanie@prysmgroup.io Telegram: @stephaniehurder Thank you!

Prysm Group regularly published blockchain industry articles: https://medium.com/prysmeconomics https://medium.com/mit-cryptoeconomics-lab MIT Sloan Executive Education recently launched a six-week executive course for which we were industry contributors: https://executive.mit.edu/openenrollment/program/blockc hain-technologies-business-innovation-and-application/

P2PFISY 2018 | July 27, 2018