EIP-2: Review Impact - Community Sentiment Adjustment

This EIP proposes enhancing the Review Impact component of the Ethos Reputation Score by incorporating community sentiment. This introduces a mechanism where the weight of a review in score calculations is adjusted based on the collective upvotes and downvotes it receives. This empowers the community to moderate the influence of reviews, reducing the impact of those perceived negatively while maintaining the significance of well-regarded ones.

Motivation

The existing reputation system weights reviews based on author reputation (ELO) and the review's stated sentiment (positive/negative/neutral). However, it lacks a direct feedback loop for the community to collectively validate or dispute the influence of individual reviews. This proposal addresses the need for community oversight by:

1. Moderating Low-Quality Reviews: Allows the community to reduce the score impact of reviews they deem inaccurate, unfair, or low-quality through downvoting.
2. Incorporating Community Consensus: Directly integrates broader community agreement or disagreement into the review weighting process.
3. Increasing Score Resilience: Makes the reputation score less susceptible to undue influence from individual reviews that lack broad community support.
4. Reducing Governance Overhead: Provides a system-level moderation tool, reducing pressure for centralized appeals.

Specification

Overview

The V3 Review Impact mechanism introduces a Community Sentiment Adjustment factor that modifies the baseline influence of a review (derived from the author/subject ELO comparison). This adjustment is determined by the pattern of upvotes and downvotes a review receives:

1. Positive Sentiment: Reviews receiving more upvotes than downvotes retain their full calculated influence.
2. Significant Negative Sentiment: Reviews accumulating a substantial number of net downvotes (exceeding a defined threshold, e.g., 5 net downvotes) will have their influence on the score calculation effectively nullified.
3. Moderate Negative Sentiment: Reviews with a smaller number of net downvotes will see their influence proportionally reduced. The more net downvotes a review has (up to the threshold), the less impact it contributes to the subject's score.

This ensures that reviews must generally align with community perception to exert their full potential influence on reputation scores.

Rationale

This adjustment-based approach (rather than a simple filter) was chosen because:

1. Nuance: It allows for a graduated response. A review with one net downvote is treated differently than one with five, reflecting varying degrees of community disagreement.
2. Stability: It prevents minor disagreements or small numbers of downvotes from completely silencing a review, requiring a clearer community consensus (reaching the threshold) for full nullification.
3. Preserves Signal: Reviews that are not strongly contested still contribute some information to the score, albeit potentially reduced.
4. Balances ELO and Sentiment: It ensures that both the author's relative standing (ELO) and the community's perception (votes) contribute to the final weight.

Considerations

• Vote Manipulation: Future EIPs will need to balance coordinated voting "brigades" and "sybils" reducing impact of the same set of individuals coordinate downvotes.