Case study: Fairness of the user discovery methods for Mastodon

Motivation

  • The presence of users is oligopolized by few celebrities in the user-generated content platforms.

  • Decentralization of users is expected.

User recommendation systems for Mastodon

  • Lists of recommended users

    • User Matching for Fedierse (Vaginaplant's)

    • Mastodon Recommended User Search (Cucmberium's)

      • Recommended users

      • Similar users

    • Recommended Followers (Osa's)

    • Potential friendship (Mastodon's official/embedded)

  • Who to follow panels

    • Pawoo

    • Pleroma

    • Halcyon

    • Misskey

Other user discovery methods

  • Lists of users

    • By their number of followers

    • By their number of boosts

    • Newcomers

    • Recently posted

    • Manual

    • Public profile endorsements (Mastodon's official/embedded)

  • Timelines

    • Home

    • Federated

    • Local

  • Search engines

Summary of fairness and unfairness

  • Fair

  • Newcomers

  • User Matching for Fediverse (Vaginaplant's)

  • Similar users (Cucmberium's)

  • Search engines

  • Potential friendship (Mastodon's official/embedded)

  • Local timeline

  • Federated timeline

  • Home timeline

  • Recently posted

  • Pawoo's who to follow panel

  • Misskey's who to follow panel

  • List of users by their number of boosts

  • Public profile endorsements (Mastodon's official/embedded)

  • Manual guidebooks

  • Recommended users (Cucmberium's)

  • Recommended Followers (Osa's)

  • List of users by their number of followers

  • Unfair

Case 1: List of users by their number of followers

Implementation

  • Mastodon Ranking [4] by User Local, Inc.

Specification

  • Ordered by the number of followers.

Unfairness

  • Oldies effect: Once we had followed an user, we rarely unfollow the one, even if the user becomes not attractive anymore.

  • Positive feedback: If we follow the users in this kind of list, it directly accelerates itself.

  • Shrink-wrap effect: The list of users shows screen names, bios, and avatars. No content is shown apparently. We are attracted by the famous names and aesthetic avatars.

Fairness

  • No power user effect: By definition.

Summary

  • One of the most unfair user discovery methods ever.

Case 2: Recommended Followers

Implementation

  • Recommended Followers [5] by Osa

Specification

  • When you use this recommendation engine, it shows the users who are followed by the users who you follow.

Unfairness

  • The same oldies effect, positive feedback, and shrink-wrap effect as the list of users by their number of followers.

  • Recommends the celebrities in the small society.

Fairness

  • No power user effect: By definition.

Summary

  • One of the most unfair user discovery methods ever.

Case 3: Recommended users in Mastodon Recommended User Search

Implementation

  • Mastodon Recommended User Search [6] by Cucmberium

Specification

  • The users who your similar users follow.

Unfairness

  • The same oldies effect, positive feedback, and shrink-wrap effect as the list of users by their number of followers.

Fairness

  • No power user effect: By definition.

Summary

  • One of the most unfair user discovery methods ever.

Case 4: Manual guidebooks

Implementation

  • Mastodon Famous Accounts [7] by Takumi

Specification

  • A blog post which is manually written.

Unfairness

  • Oldies effect: Manual guidebooks are rarely updated.

  • Shrink-wrap effect: The authors of the guidebooks are attracted by famous names or aesthetic avatars.

  • The authors of the guidebooks sometimes just follow some existing authority.

Fairness

  • No positive feedback: Our action does not affect to the static documents.

  • No power user effect: By definition.

Summary

  • Super unfair user discovery.

  • Fair minded authors of the guidebooks are expected.

Case 5: Public profile endorsements

Implementation

  • Mastodon's officiel/embedded [8]

Specification

  • The user pins other users on one's profile.

Fairness and Unfairness

  • Almost same as the manual guidebooks.

Summary

  • Super unfair user discovery.

  • Endorsement is worship or flattery [9].

Case 6: List of users by their number of boosts

Implementation

  • Mastodon Ranking [4] by User Local, Inc.

Specification

  • Ordered by the number of boosts.

Unfairness

  • Oldies effect: The number of boosts is accumulated ever.

  • Positive feedback: If we follow the user who is boosted well, we boost such kind of user's posts more often.

  • Power user effect: More posts, more boosts.

Fairness

  • No shrink-wrap effect: We read the post before we boost it.

Summary

  • Super unfair user discovery.

  • Better than the user discovery methods which depend on the number of followers.

Case 7: Misskey's who to follow panel

Implementation

  • Misskey [10]

Specification

  • Posting in the last 7 days.

  • Local users.

  • Many followers.

Unfairness

  • The same positive feedback and shrink-wrap effect as the list of users by their number of followers.

Fairness

  • Less oldies effect: Requires posting in the last 7 days.

  • No power user effect: By definition.

Summary

  • Remarkably unfair user discovery.

  • Better than the oldies effected user discovery methods.

Case 8: Pawoo's who to follow panel

Implementation

  • Pawoo [11, 12]

Specification

  • Following in pixiv.

  • Popular (many favorited, many followers, many pictures) users.

  • Active (many recent posts) users.

Unfairness

  • Many favorited and many followers criteria causes oldies effect, positive feedback, and shrink-wrap effect.

  • Many pictures criteria causes oldies effect.

  • Many favorited and many pictures criteria cause power user effect.

Fairness

  • Less oldies effect: Requires recent posts.

Summary

  • Remarkably unfair user discovery.

  • Better than the oldies effected user discovery methods.

Case 9: Recently posted

Implementstion

  • Profile directory [13] (Mastodon's official/embedded, default)

Specification

  • List of the users who recently posted.

Unfairness

  • Power user effect: By definition.

  • shrink-wrap effect: By definition.

Fairness

  • No oldies effect: By defintion.

  • No positive feedback: By definition.

Summary

  • Still unfair user discovery.

  • Power user effect remarkably brings unfairness.

Case 10: Home timeline

Implementstion

  • Embedded

Specification

  • Following users' posts.

  • Following users' boosts.

Unfairness

  • Positive feedback: If you follow some users and boost their posts, it accelerates itself.

  • Power user effect: By definition.

Fairness

  • No oldies effect: By definition.

  • No shrink-wrap effect: By definition.

Summary

  • Still unfair user discovery.

  • Positive feedback remarkably brings unfairness.

Case 11: Federated timeline

Implementstion

  • Embedded

Specification

  • Local users' posts.

  • Posts by the remote users who are followed by the local users.

Unfairness

  • Positive feedback: If a local user follows a remote user, the other local users may follow the remote user via their federated timeline.

  • Power user effect: By definition.

Fairness

  • No oldies effect: By definition.

  • No shrink-wrap effect: By definition.

Summary

  • Still unfair user discovery.

  • Mixed feature of the home and local timelines.

Case 12: Local timeline

Implementstion

  • Embedded

Specification

  • Local users' posts

Unfairness

  • Power user effect: By definition.

Fairness

  • No oldies effect: By definition.

  • No positive feedback: By definition.

  • No shrink-wrap effect: By definition.

Summary

  • Still unfair user discovery.

  • Power user effect remarkably brings unfairness.

Case 13: Potential friendship

Implementstion

  • Mastodon's official/embedded [14]

Specification

  • When an user engages (replies, favourites, and boosts) other user, the potential friendship counter is bumped.

Fairness and Unfairness

  • Mixed feature of the home, fedarated, and local timelines [15].

Summary

  • Still unfair user discovery.

Case 14: Search engines

Implementation

  • Tootsearch [16]

  • Mastodon Realtime Search [17]

  • Mastodon Search Portal [18]

  • Mastodon Search by Google [19]

Specification

  • Search posts by keyword.

Unfairness

  • Power user effect: More posts, more hits.

Fairness

  • No oldies effect: By definition.

  • No positive feedback: By definition.

  • No shrink-wrap effect: By definition.

Summary

  • Still unfair user discovery.

  • Not suit for the user discovery use. Explicit keyword is mandatory.

Case 15: Similar users in Mastodon Recommended User Search

Implementation

  • Mastodon Recommended User Search [6] by Cucmberium

Specification

  • The users who follow who you follow.

Unfairness

  • Shrink-wrap effect: By definition.

Fairness

  • No oldies effect: By definition.

  • No positive feedback: By definition.

  • No power user effect: By definition.

Summary

  • Neutral for fairness and unfairness.

  • Small cliques or clusters are formed.

Case 16: User Matching for Fediverse

Implementation

  • User Matching for Fediverse [20] by Vaginaplant

  • Who to follow panel in Pleroma [21]

  • Who to follow panel in Halcyon [22]

Specification

  • The users who have the vocabulary which similar to you.

  • Vocabulary is abstracted from one's screen name, bio, and posts.

  • Users who have too much followers (500 by default) are omitted.

Unfairness

  • Shrink-wrap effect: By definition.

Fairness

  • No oldies effect: By definition.

  • Reverse positive feedback: Users who have too much followers are omitted.

  • No power user effect: By definition.

Summary

  • Fair user discovery.

Case 17: Newcomers

Implementation

  • Profile directory [13] (Mastodon's official/embedded)

  • Newcomers in Fediverse [23] by Vaginaplant

Specification

  • List of newcomers.

Unfairness

  • Shrink-wrap effect: By definition.

Fairness

  • Reverse oldies effect: By definition.

  • No positive feedback: By definition.

  • No power user effect: By definition.

Summary

  • Fair user discovery.

References

[1] Hakaba Hitoyo, 脱中央集権のためのデザイン: セレブのためのインターネットを99 %の手に取り戻す, https://hakabahitoyo.wordpress.com/2017/12/24/decentralization-by-design

[2] Hakaba Hitoyo, マストドンのユーザーレコメンデーションがブームに?, https://hakabahitoyo.wordpress.com/2018/03/29/mastodon-user-recommendation-2018

[3] Hakaba Hitoyo, マストドンのユーザーレコメンデーションについて追記, https://hakabahitoyo.wordpress.com/2018/04/18/misskey-halcyon-recommendation

[4] User Local, Inc., Mastodon Ranking, http://mastodon.userlocal.jp

[5] Osa, Recommended Followers, https://followlink.osa-p.net/recommend.html

[6] Cucmberium, Mastodonおすすめユーザー検索, https://nocona.cucmber.net/eryngium

[7] Takumi, マストドン (Mastodon) ユーザなら必ずフォローしたい! アカウント一覧, https://takulog.info/mastodon-famous-accounts

[8] Gargron, Public profile endorsements (accounts picked by profile owner) #8146, https://github.com/tootsuite/mastodon/pull/8146

[9] Hakaba Hitoyo, Endorsement is some kind of worship or flattery, https://github.com/tootsuite/mastodon/pull/8146#issuecomment-411702072

[10] https://github.com/syuilo/misskey/blob/master/src/server/api/endpoints/users/recommendation.ts

[11] pixiv Inc., Pawoo, https://pawoo.net

[12] https://github.com/pixiv/mastodon/blob/pawoo/app/models/suggested_account_query.rb

[13] Gargron, Add profile directory to web UI #11688, https://github.com/tootsuite/mastodon/pull/11688

[14] Gargron, Re-add follow recommendations API #7918, https://github.com/tootsuite/mastodon/pull/7918

[15] Hakaba Hitoyo, ユーザーレコメンデーションがマストドン本体に導入された: フェアネスの評価は「現状維持」, https://hakabahitoyo.wordpress.com/2018/07/03/gargrons-official-user-recommendation

[16] Kirino Minato, Tootsearch, https://tootsearch.chotto.moe

[17] User Local, Inc., マストドンリアルタイム検索, http://realtime.userlocal.jp

[18] マストドン検索ポータル, http://mastodonsearch.jp

[19] Osa, マストドン検索 by google, https://mastosearch.osa-p.net

[20] Hakaba Hitoyo, User Matching for Fediverse, https://distsn.org/user-match.html

[21] lain, Pleroma, https://pleroma.social

[22] Neetshin, Niklas Poslovski, Halcyon, https://www.halcyon.social

[23] Hakaba Hitoyo, Newcomers in Fediverse, https://distsn.org/user-new.html

[24] Hakaba Hitoyo, User Recommendation Systems for Mastodon and their Fairness, https://hakabahitoyo.gitlab.io/slides/recommendations

Contents
Motivation
User recommendation systems for Mastodon
Other user discovery methods
Summary of fairness and unfairness
Case 1: List of users by their number of followers
Implementation
Specification
Unfairness
Fairness
Summary
Case 2: Recommended Followers
Implementation
Specification
Unfairness
Fairness
Summary
Case 3: Recommended users in Mastodon Recommended User Search
Implementation
Specification
Unfairness
Fairness
Summary
Case 4: Manual guidebooks
Implementation
Specification
Unfairness
Fairness
Summary
Case 5: Public profile endorsements
Implementation
Specification
Fairness and Unfairness
Summary
Case 6: List of users by their number of boosts
Implementation
Specification
Unfairness
Fairness
Summary
Case 7: Misskey's who to follow panel
Implementation
Specification
Unfairness
Fairness
Summary
Case 8: Pawoo's who to follow panel
Implementation
Specification
Unfairness
Fairness
Summary
Case 9: Recently posted
Implementstion
Specification
Unfairness
Fairness
Summary
Case 10: Home timeline
Implementstion
Specification
Unfairness
Fairness
Summary
Case 11: Federated timeline
Implementstion
Specification
Unfairness
Fairness
Summary
Case 12: Local timeline
Implementstion
Specification
Unfairness
Fairness
Summary
Case 13: Potential friendship
Implementstion
Specification
Fairness and Unfairness
Summary
Case 14: Search engines
Implementation
Specification
Unfairness
Fairness
Summary
Case 15: Similar users in Mastodon Recommended User Search
Implementation
Specification
Unfairness
Fairness
Summary
Case 16: User Matching for Fediverse
Implementation
Specification
Unfairness
Fairness
Summary
Case 17: Newcomers
Implementation
Specification
Unfairness
Fairness
Summary
References