EIP-10: Revision to Invitation Bond Calculation

Status: Draft
Author: Ethos Labs
Created: 2025-09-22

---

Summary

This proposal updates the calculation method for the Invitation Bond component of the Ethos scoring algorithm. Instead of using a weighted average across all accepted invitations, the new approach will only take into account the scores of the top 2 and bottom 2 accepted invitations.

---

Abstract

The Invitation Bond currently averages the scores of all invitees accepted through a given member. While this ensures broad coverage, it disproportionately lowers the Invitation Bond of members who have successfully invited a large number of users, as the inclusion of many average or below-average scores dilutes the impact of high-performing invitees.

This proposal modifies the algorithm so that only the highest two and lowest two invitee scores are considered. This balances recognition of reliable, high-scoring invitees with accountability for weaker invitees, while avoiding the flattening effect of large sample sizes.

---

Motivation

• Fairness to highly active inviters: Members who have invited dozens of people see their Invitation Bond averaged down, even when they have successfully brought in high-scoring contributors.
• Preserving accountability: By including the lowest two invitees, inviters are still responsible for the weakest members they bring in.
• Rewarding reliability: The highest two invitees ensure that strong contributions are properly reflected in the inviter’s score.
• Better signal value: The adjusted method provides a sharper measure of inviters’ ability to identify high-reputation members without punishing scale.

---

Specification

• Let I be the set of accepted invitations sent by a given member.
• If |I| <= 4, the Invitation Bond will be calculated as the average of all invitees’ scores (unchanged from the current method).
• If |I| > 4, the Invitation Bond will be calculated as the average of:
  • the top 2 invitee scores in I
  • the bottom 2 invitee scores in I

Formally:

If |I| <= 4:
IB = average(score(i) for i in I)

Else:
IB = (max2(I) + min2(I)) / 4

where max2(I) are the two highest scores in I and min2(I) are the two lowest scores in I.

---

Rationale

This design preserves the accountability mechanism of including weaker invitees while ensuring that top performers meaningfully influence the Invitation Bond. It prevents the "flattening effect" seen when active inviters’ bonds converge toward mediocrity due to large sample averaging.

By taking both extremes into account, the measure captures both the inviter’s best judgment and their weakest choices, producing a balanced yet more accurate reflection of their contribution.

---

Backwards Compatibility

This change directly alters the Invitation Bond calculation, and therefore will cause retroactive shifts in scores. These shifts are expected and intended, as the updated algorithm provides a more accurate reflection of inviter contributions.

No other components of the Ethos score algorithm are affected.