Why Ethereum Needs Fairness Mechanisms that Do Not Depend on Participant Altruism

This paper argues that Ethereum's decentralization and censorship resistance ideals cannot be restored by relying on altruistic block proposers, as empirical analysis reveals that less than 1.4% of proposers consistently act in accordance with these objectives, thereby necessitating the implementation of incentive- or penalty-based fairness mechanisms.

The Gist

Here is an explanation of the paper using simple language and creative analogies.

The Big Picture: The "Honest Baker" Problem

Imagine Ethereum is a giant, global bakery. Every few seconds, a new baker (called a Proposer) is randomly chosen to bake a loaf of bread (a Block) that contains all the transactions people want to make.

The ideal of this bakery is decentralization: anyone can be a baker, and no single person should control what goes into the bread. The bread should be fair, meaning no one gets cut out of the line just because they didn't pay a bribe.

However, the authors of this paper asked a simple question: "How many of these bakers are actually honest, or 'altruistic'?"

They define an altruistic baker as someone who follows the rules perfectly, even if it costs them money. For example, if a rich customer offers a bribe to skip a poor customer's order, an altruistic baker says, "No, I follow the rules," and bakes the bread fairly.

The Investigation: Who is Really Baking?

The researchers looked at data from January to April 2025 to see who was actually baking the bread. They found a shocking reality: The bakery is mostly run by profit-hungry machines, not honest humans.

Here is how they filtered out the "dishonest" bakers, step-by-step:

1. The "Outsourced" Bakers (92% are out!)

Most bakers (92%) don't bake the bread themselves. They hire a giant, centralized factory called MEV-Boost (a builder service) to do the baking for them.

• The Analogy: Imagine you are chosen to bake a cake, but you immediately call a massive factory, say, "Here is my oven key, you bake whatever you want, and I'll just sign the receipt."
• The Problem: The factory bakes the cake to make the maximum profit, often by cutting in line for rich customers or excluding others. The original baker has no idea what's in the cake.
• Result: These bakers are not altruistic. They are just renting out their power for a quick fee.

2. The "Shared Kitchen" Bakers (Another 6% are out!)

The researchers then looked at the remaining bakers who didn't hire the factory. They checked if these bakers were part of a "Shared Kitchen" (a Staking-as-a-Service provider).

• The Analogy: Imagine a baker who claims to be independent, but they share their oven and recipe book with a shady group of 50 other bakers. If one person in that group decides to bake a corrupt cake, the whole group is compromised.
• The Problem: Even if a specific baker looks honest today, they might be controlled by a manager who acts greedily tomorrow.
• Result: These bakers are also considered not altruistic because they lack true independence.

3. The "Profit-Seeking" Bakers (The final 0.6% are out!)

Finally, they looked at the tiny group of bakers who seemed truly independent. They checked if these bakers were secretly manipulating the order of transactions to squeeze extra profit (like front-running).

• The Analogy: These bakers weren't hiring a factory, but they were still rearranging the ingredients in the bowl to make a little extra cash for themselves.
• Result: Even these "independent" bakers were often trying to game the system.

The Final Count: The "1.4%" Reality

After all this filtering, the researchers found that less than 1.4% of the bakers were genuinely altruistic. They were the only ones who:

1. Didn't hire a factory.
2. Didn't share their keys with a shady group.
3. Didn't try to game the system for extra profit.

The Shocking Stat:

• Old Belief: People thought maybe 9% of bakers were honest.
• New Reality: Only 1.4% are honest.
• The Rest: 98.6% are acting in their own self-interest (or are controlled by those who do).

Why Does This Matter? (The "Committee" Problem)

The paper discusses new ideas to fix Ethereum's fairness, like Inclusion Lists.

• The Idea: Instead of one baker deciding what goes in the bread, a Committee of 16 bakers decides together. The theory is: "If we have 16 people, surely at least one of them will be honest and stop the corruption, right?"
• The Math:
  • If 9% of bakers are honest, a 16-person committee has a 78% chance of having at least one honest person. That sounds good!
  • But if only 1.4% are honest, that same 16-person committee only has a 21% chance of having an honest person.
  • The Danger: If the committee is mostly made of greedy bakers, they can easily vote to censor people or take bribes. The "safety net" of altruism has vanished.

The Solution: Don't Rely on "Good Guys"

The authors conclude that we cannot build a fair system by hoping for "good guys" (altruists) to show up. They are too rare.

The New Strategy:
Instead of hoping people are nice, we need to build a system where being fair is the most profitable thing to do.

• Analogy: Instead of hoping a baker won't steal your cookies because they have a "good heart," you install a camera and a lock that makes stealing impossible or too expensive to attempt.
• Recommendation: If we must use committees, they need to be huge (at least 128 people) to make it statistically likely that an honest person is there. But even then, it's risky. The real fix is to create rules where cheating doesn't pay.

Summary in One Sentence

Ethereum's dream of a fair, decentralized network is being undermined because 98.6% of the people running it are motivated by profit, not principle; therefore, we can't rely on "good behavior" to save the system—we need to change the rules so that "good behavior" is the only logical choice.

Technical Summary

Here is a detailed technical summary of the paper "Ethereum Needs Fairness Mechanisms that Do Not Depend on Participant Altruism" by Spiesberger, Beyer, and Hartenstein.

1. Problem Statement

Ethereum's core ideals of decentralization and censorship resistance are currently undermined in practice. While the network relies on a consensus model where proposers (validators) construct blocks, empirical evidence shows that proposers often act as rational agents seeking to maximize profit rather than adhering to protocol ideals. This leads to behaviors like transaction censorship, front-running, and the exclusion of specific transactions.

Recent proposals to restore fairness, such as Inclusion Lists (ILs) and Multiple Concurrent Proposer (MCP) designs, rely on a critical assumption: that a sufficient fraction of proposers are altruistic. In this context, "altruistic" is defined not by moral intent (which is unobservable on-chain) but by behavior: proposers who follow the protocol exactly, prioritize user fairness, and do not deviate for short-term financial gain (e.g., accepting bribes to censor transactions).

The central research question is: What is the actual share of proposers exhibiting altruistic behavior? If this share is too low, committee-based fairness mechanisms will fail because the probability of a committee containing at least one honest member drops significantly, rendering them vulnerable to collusion or bribery.

2. Methodology

The authors conducted an empirical analysis of Ethereum block production data from January to April 2025 (covering 859,313 blocks proposed by 600,115 distinct proposers). They employed a progressive exclusion methodology to filter out non-altruistic proposers, moving from broad categories to specific behavioral indicators.

The classification process involved six classifiers:

1. Relaying Proposers (Classifier I):

   • Logic: Proposers who use MEV-Boost relays (e.g., Flashbots) blindly sign blocks constructed by external builders. They cede control over block content and cannot prevent censorship or front-running.
   • Action: Excluded any proposer whose block was mediated by a known relay at least once.

2. Shared Governance (Classifier II):

   • Logic: Proposers often share infrastructure (private keys) with others via Staking-as-a-Service (STaaS) providers. If one entity in a governance cluster acts non-altruistically, the entire cluster is considered non-altruistic.
   • Action: Clustered proposers by their coinbase address (reward recipient). If a cluster contained any proposer identified as a "Relaying Proposer" (from Step 1), the entire cluster was excluded.

3. Account Type Filtering (Classifier III):

   • Logic: Contract Accounts (CAs) used as coinbase addresses often indicate automated STaaS setups with non-altruistic management.
   • Action: Excluded clusters where the coinbase address was a Contract Account, retaining only Externally Owned Accounts (EOAs).

4. Builder Payments (Classifier IV):

   • Logic: Even without using MEV-Boost relays, proposers might receive direct payments from builders for block construction rights.
   • Action: Checked relay endpoints for payments to the coinbase addresses of remaining clusters. If a builder paid a proposer, the cluster was excluded.

5. Exclusive Order Flow (XOF) Detection (Classifier V):

   • Logic: Proposers may extract MEV by including private transactions (XOF) that never appeared in the public mempool, often to front-run users.
   • Action: Compared block transactions against a dataset of public mempool transactions. Proposers including XOF transactions were excluded individually (not by cluster, to avoid over-penalizing).

6. Transaction Ordering Analysis (Classifier VI):

   • Logic: Altruistic proposers should order transactions based on protocol-defined attributes (e.g., priority fees). Rational proposers might reorder transactions to maximize profit in ways not explained by fees.
   • Action: Calculated the Spearman rank correlation coefficient ($\sigma_{rs}$) between transaction fees and their position in the block.
     • Strictly Altruistic: $\sigma_{rs} \approx 1$ (strictly descending fee order).
     • Potentially Altruistic: High correlation ($\sigma_{rs} > 0.7$) but with minor deviations.
     • Non-Altruistic: Low correlation or ordering inconsistent with fee structures.

3. Key Results

The study quantified the share of altruistic proposers by progressively narrowing the pool:

• Total Observed Proposers: 600,115.
• Non-Relaying Proposers: Only 7.87% of proposers never used MEV-Boost relays. (92.13% were excluded immediately).
• Non-Outsourcing Proposers: After accounting for shared governance (clusters with relaying members), only 2.14% remained.
• Refined Non-Outsourcing: After filtering for EOA-only clusters and builder payments, 1.82% remained.
• No XOF Activity: After removing proposers using Exclusive Order Flow, 1.36% remained.
• Final Classification:
  • Strictly Altruistic: 0.65% (3,814 proposers) showed strict, deterministic ordering by priority fees.
  • Potentially Altruistic: 0.71% (4,211 proposers) showed strong correlation ($\sigma_{rs} > 0.7$) but minor deviations.
  • Total (Potentially) Altruistic: 1.36%.

Key Demographic Insight:
The study found that "altruistic" behavior is heavily concentrated among early validators (registered before "The Merge"). In the early validator index range, ~12.46% exhibited potentially altruistic behavior, whereas in recent indices, it was less than 0.2%. This suggests that altruism is correlated with participation during periods lacking strong economic incentives for MEV extraction.

4. Key Contributions

1. Empirical Quantification of Altruism: The paper provides the first rigorous, data-driven upper bound on the share of altruistic proposers, finding it to be ~1.36%, significantly lower than the "few percent" often assumed in theoretical models.
2. Refined Definition of Non-Altruism: The authors move beyond simple "MEV-Boost usage" to define non-altruism through shared governance structures and internal MEV extraction (XOF, ordering), arguing that delegation to rational entities (even if not via relays) constitutes non-altruistic behavior.
3. Re-evaluation of Fairness Mechanisms: The paper challenges the viability of current committee-based proposals (like Inclusion Lists) that rely on the presence of at least one altruist.
   • Example Calculation: For a 16-member committee, the probability of having at least one altruist drops from 78% (assuming 9% altruism) to **20.7%** (assuming 1.36% altruism).
   • Implication: To achieve a 99% probability of having an honest member, a committee would need to be ~128 members large, which introduces significant latency and complexity.

5. Significance and Conclusion

The paper concludes that relying solely on participant altruism is insufficient to secure Ethereum's fairness and censorship resistance goals. The current ecosystem is dominated by rational, profit-maximizing actors (98.64% of proposers).

Implications for Future Design:

• Incentive Alignment: New fairness mechanisms must not depend on the presence of altruists but must make fairness the rational choice. This requires penalty mechanisms (slashing) or reward structures that make deviating from the protocol financially unprofitable.
• Committee Sizing: If committee-based mechanisms are used, they must be significantly larger (e.g., 128+ members) to statistically guarantee the inclusion of an honest actor, though this may impact performance.
• Architecture Shift: The findings support designs like FairFIL, which shifts the burden of censorship detection to the block builder and requires explicit recording of censorship, making it harder to hide malicious behavior without a large coalition of validators.

In summary, the authors argue that Ethereum's future stability depends on mechanism design that enforces fairness through incentives and penalties, rather than hoping for a sufficient number of benevolent participants.