โ† Back to Home
Ethos
ben walther's review for Ethos
Score: 2232
NEUTRAL

"EIP-3.1: Credibility Score Adjustment for Supporters of Slashed Users"

# **EIP-3.1: Credibility Score Adjustment for Supporters of Slashed Users** **Replaces:** EIP-3 (Original proposal at https://app.ethos.network/activity/review/157484/eip-3-sequential-review-eligibility ) ## **Abstract** This proposal introduces a credibility score penalty for users who have an active vouch or an unarchived positive review for another user who is subsequently successfully slashed. If a user (the "Slashee") is slashed, individuals who positively reviewed the Slashee will lose credibility score equal to 10% of the Slashee's total slashed amount. Individuals who vouched for the Slashee will lose credibility score equal to 20% of the Slashee's total slashed amount. Users will have the duration of the slashing period (e.g., 2 days) to archive their reviews or unvouch the Slashee to avoid these penalties. This change is intended to encourage more deliberate and thoughtful endorsements on the platform. This penalty will be recorded under a new category, such as "Penalty for Supporting Slashed User." ## **Rollout** This will only apply to slashes resolved after EIP-3.1 is implemented. It does not apply to any previous or existing slashes, as of May 28, 2025. ## **Motivation** The core motivation behind this EIP is to enhance the integrity and thoughtfulness of interactions within the Ethos ecosystem. Currently, there is no direct consequence for endorsing users who later face successful slashes. This proposal aims to: 1. **Encourage Due Diligence:** Incentivize users to be more discerning and thoughtful before leaving positive reviews or vouching for others, particularly individuals they may not know well. 2. **Strengthen Trust Signals:** Make vouches and positive reviews more reliable indicators of trust by associating a potential cost with misplaced endorsements. 3. **Reflect Shared Responsibility:** Introduce a measure of accountability for those whose endorsements might have lent credibility to a user who ultimately harmed the ecosystem or violated community standards. It is hoped that this change will lead to more considered endorsements, thereby strengthening the overall reputation system. ## **Specification** 1. **Triggering Condition:** * A user (the "Slashee") is successfully slashed. 2. **Conditions for Penalty Application:** * At the moment the slash is finalized, another user has an active (non-archived) positive review for the Slashee. * At the moment the slash is finalized, another user has an active vouch for the Slashee. 3. **Credibility Score Penalties (upon successful slash):** * **Reviewers:** Any user with an active, positive review for the Slashee will lose credibility score equal to 10% of the total credibility score amount lost by the Slashee in that specific slash event. * **Vouchers:** Any user with an active vouch for the Slashee will lose credibility score equal to 20% of the total credibility score amount lost by the Slashee in that specific slash event. * These penalties are distinct and not cumulative; if a user both reviewed and vouched for the Slashee they are penalized the maximum aount (ie, 20%). 4. **Grace Period & Mitigation:** * During the slashing period (the time between when a slash is initiated and when it is finalized, e.g., 2 days), users will have the opportunity to: * Archive any positive reviews they have issued for the user facing the slash. * Unvouch the user facing the slash. * If a review is archived or a vouch is removed before the slash is finalized, the respective penalty will not apply to that user. 5. **Penalty Categorization:** * Score deductions resulting from this EIP will be attributed to a new, distinct category, for example: "Penalty for Supporting Slashed User" or "Misplaced Endorsement Penalty." 6. **No Impact on Failed Slashes:** * If a slash attempt against a user fails, no credibility score adjustments outlined in this EIP will be applied to their reviewers or vouchers. ## **Rationale and Considerations** * **Potential for Negative Dynamics:** We acknowledge a risk that this EIP could introduce unintended negative consequences, such as an increased reluctance to review or vouch, or a generally more cautious atmosphere than desired. The effects will be monitored closely. ## **Beta Implementation & Future Iteration** This feature will be implemented as a BETA. * **Parameter Adjustments:** The penalty percentages (10% for reviews, 20% for vouches) are initial values and may be subject to change based on observed outcomes and community feedback. * **Rollback Clause:** If this EIP is found to significantly increase negativity on the platform, fails to achieve its intended goals, or otherwise proves detrimental, a subsequent proposal to amend or roll back these changes will be considered.

224 upvotes
May 28, 2025

Comments

MLO
MLO
2088
May 28, 2025
I like the change
ak
ak
2059
May 28, 2025
I think that allowing users to archive/remove their reviews/vouches for a slashed user is not ideal: State should be locked/snapshotted at the moment the slash goes live. Moreover, users who lose score due to this should have a neutral review left by 'SlashAgent' or similar detailing the reason for the credibility loss so that it is properly documented/visible on their profile. With these changes, I am fully in support. I think this is a step in the right direction either way.
Doesn't this: "Users will have the duration of the slashing period (e.g., 2 days) to archive their reviews or unvouch the Slashee to avoid these penalties" effectively cancel the effect of this new proposal? I would only enable something like this when the profile getting slashed is a company (vs an individual). As in, you give a positive review for FTX in 2021, but it won't affect you 1 year later when it unexpectedly crashed. However, if we talk about people, the effect is unavoidable. This will ensure other people think twice before reviewing/vouching for someone they don't REALLY trust. Rugs & crashes are comparatively harder to predict as an individual vs bad behavior from another person.
Crypto Degen
Crypto Degen
1465
May 28, 2025
I think it is a good point. However, following need to be considered: 1. Issue: There are several projects or people who initially are very good. People review them based on their image back then. Some people or projects when they get fame, that's the time when they show their actual plan. So penalising people for their old review or vouches may actually impact people who actually reviewed based on initial perception. Suggestion: Make impact of such slashing time bound. I mean in last 1 week or last 3 days etc. 2. Issue: There are some slashing which is done for sake of slashing. Example some people are slashing that a person gave AI based reviews based on their evaluation of other person's post. Issue with this is that it is very subjective. Suggestion: There need to be preselected labels or guidance which need to be added by team which help evaluate whether slashing in the first place is justified or not. I mean I know some people who use AI tools to translate stuff from their original language. Reason being, they don't know English. And AI tools when they translate, they leave taste of AI drafting. I am not saying that all slashing is wrong and probably things will evolve with passage of time but i feel that there need to be more clear guidance before reviewers get penalised.
ZimmerAllDay
ZimmerAllDay
1800
May 29, 2025
Unlike some of the comments posted above, I wouldn't have supported this WITHOUT the ability to archive review or unvouch, because, someone you have known, interacted, or supported in the past could absolutely do something that you would have not expected of them, and you would be punished for this. People are unpredictable, even in irl, and they can often let us down or surprise us. It would be unfair if your score suffered because someone you trusted did something shady. You could leave a genuine review and still get burnt by this. I am supporting the EIP because it allows for this. It is still a layer of deterrence, as there is additional risk added by reviewing/vouching for someone you don't know. If you weren't tuned in to Ethos, and they got slashed, you would be punished. Overall, I think it strikes a nice balance.
Voted my decision as supporting it
Supporting this, even with the risk of penalty if you vouched. But might make people more careful to vouch.
emilios.eth
emilios.eth
1975
May 29, 2025
Thoughts on EIP3.1 I treat this as a "minimum viable enforcement" but for the long-term, the design logic should evolve 1. Reservations a. Grace period - EIP allows users to escape penalties by removing their pos review during the slashing period - this doesn't make sense because it breaks the fundamental logic of trust systems - IRL, when someone you've vouched for does something wrong, you can't retroactively "un-recommend" them to protect your rep info asymmetry โ†’ a user that happens to see the active slashing vs someone else who didn't log in for 2-3 days won't, so only one escapes consequences b. Fixed penalty %s - ignore the actual economic risk each user took - eg someone who vouched with 0.01ETH vs 1 ETH faces the same percentage penalty despite the x100 different skin in the game economically irrational โ†’ penalties should reflect actual exposure (although this is counterintuitive for the protocol's TVL) c. Mixing reviews and vouches - conflating reviews and vouches dilutes the clarity of both - blending them may help with low-effort farming now, but I think that it kind of blurs long-term cred modeling โ†’ merging them short-term makes sense only as an immediate anti-Sybil measure, but on the long term is kind of Yelp vs co-signer on a loan (both good but absolutely not the same signal) 2. Comments Rollback clause - I get the learn by shipping logic and I do agree because I believe that strong leadership = course-correction when data proves you wrong - we already have the draft โ†’ discussion โ†’ implementation so why don't we extend the window for a dialogue? - not implying to follow a deliberately slow and conservative model like ETH (apples and oranges) but I'm just saying that (future) flip-flopping can potentially erode confidence in governance โ†’ just thinking out loud here 3. My proposal a. Logic: when you vouch for someone, you're making a prediction about their future behavior so if it's wrong, you should face consequences proportional to your exposure just how IRL rep works: no loopholes, no grace periods tldr vouching is a social bet, not just a tx b. mechanism: - immediate penalty: when someone gets slashed, all their vouchers face instant credibility loss (if they believe in them, they'll do their best to defend it) - proportional structure: [[ your penalty = (your vouch amount รท total vouch amounts) ร— slashee's credibility loss ]] - max cap (to avoid edge cases): no vouch can cost you more than 20% of your total cred (large backers take a hit, but aren't nuked to Mordor) - buffer: dampening by vouchers count = if 5+ people vouched, more vouchers = shared responsibility โ†’ multiplier reduces everyone's penalty slightly - no reversals: once you vouch, you're committed until you unvouch c. why this works: - more realistic from an economic POV, - proper incentives for careful vouching, - don't disincentivize vouching (cap + buffer) - soft anti-sybil measure
Page 1 of 2