Public Education Forum Moderation
@pef-moderation.org
Read Pinned post for mission
Startup admin: stevetownsend0.bsky.social
Appeals: admin@pef-moderation.org
Follow for PSAs
Committed to no use of !hide !warn etc outside illegality
Startup admin: stevetownsend0.bsky.social
Appeals: admin@pef-moderation.org
Follow for PSAs
Committed to no use of !hide !warn etc outside illegality
Pinned
This labeler's axioms:
1. Subjective content moderation by platform is doomed to fail and should be confined to legality (e.g. remove CSAM) and authenticity - expand verification, remove bots and fakes.
Focus on takedown, not subjective post-level decisions.
Trying to do more is a costly distraction
1. Subjective content moderation by platform is doomed to fail and should be confined to legality (e.g. remove CSAM) and authenticity - expand verification, remove bots and fakes.
Focus on takedown, not subjective post-level decisions.
Trying to do more is a costly distraction
Reposted by Public Education Forum Moderation
"@pef-moderation.org is not an official Bluesky labeler.
For the most part, they label things sensibly and understandably, but basically anyone can create their own labeler and label everyone however they want."
h/t @nightlythievery.bsky.social
Subjective - ymmv, esp. note our pro-Israel labeling
For the most part, they label things sensibly and understandably, but basically anyone can create their own labeler and label everyone however they want."
h/t @nightlythievery.bsky.social
Subjective - ymmv, esp. note our pro-Israel labeling
🚨 #BREAKINGNEWS Beiträge des Außenministeriums der Trump-Administration werden auf BlueSky als Desinformation gekennzeichnet. Anders als bei X werden Propaganda und Desinformation dort nicht toleriert. 🚨
Via @FordJohnathan5.x
Via @FordJohnathan5.x
October 19, 2025 at 10:45 PM
"@pef-moderation.org is not an official Bluesky labeler.
For the most part, they label things sensibly and understandably, but basically anyone can create their own labeler and label everyone however they want."
h/t @nightlythievery.bsky.social
Subjective - ymmv, esp. note our pro-Israel labeling
For the most part, they label things sensibly and understandably, but basically anyone can create their own labeler and label everyone however they want."
h/t @nightlythievery.bsky.social
Subjective - ymmv, esp. note our pro-Israel labeling
State of the Union post regarding platform moderation from the team
whtwnd.com/bnewbold.net...
whtwnd.com/bnewbold.net...
AT Moderation Architecture | bryan newbold
The AT network is becoming more heterogeneous in practice, with independent PDS hosts, apps, and alternative bsky AppViews establishing themselves. This means that more complex inter-service moderatio...
whtwnd.com
October 7, 2025 at 4:40 PM
State of the Union post regarding platform moderation from the team
whtwnd.com/bnewbold.net...
whtwnd.com/bnewbold.net...
@antibot4navalny.bsky.social reported on these in detail a couple of days ago. I have labeled as many as I could find.
mame tu noveho troll-a / bota / ai account ktory odkazuje na TG kanal, ktory odkazuje na "Slovanske Noviny"
prizemna srackoidna ruZka propaganda.
vy viete co mate robit. report, block etc...
prizemna srackoidna ruZka propaganda.
vy viete co mate robit. report, block etc...
October 3, 2025 at 10:15 PM
@antibot4navalny.bsky.social reported on these in detail a couple of days ago. I have labeled as many as I could find.
The latest antisemitic atrocity, this time in the UK, got me thinking about balance in moderation.
Revisited two frequently-used lazy slurs, both of which are auto-labeled as abusive/disinfo by this service.
Bluesky live post/account counts for:
ZioN*zi : 3912/927
PaliN*zi : 57/34
Revisited two frequently-used lazy slurs, both of which are auto-labeled as abusive/disinfo by this service.
Bluesky live post/account counts for:
ZioN*zi : 3912/927
PaliN*zi : 57/34
October 2, 2025 at 4:51 PM
The latest antisemitic atrocity, this time in the UK, got me thinking about balance in moderation.
Revisited two frequently-used lazy slurs, both of which are auto-labeled as abusive/disinfo by this service.
Bluesky live post/account counts for:
ZioN*zi : 3912/927
PaliN*zi : 57/34
Revisited two frequently-used lazy slurs, both of which are auto-labeled as abusive/disinfo by this service.
Bluesky live post/account counts for:
ZioN*zi : 3912/927
PaliN*zi : 57/34
Labeled hundreds of accounts just now found in this search
bsky.app/search?q=aip...
I did a manual review of all results but might have missed a couple all the same. Please let me know if so.
bsky.app/search?q=aip...
I did a manual review of all results but might have missed a couple all the same. Please let me know if so.
bsky.app
September 29, 2025 at 1:26 PM
Labeled hundreds of accounts just now found in this search
bsky.app/search?q=aip...
I did a manual review of all results but might have missed a couple all the same. Please let me know if so.
bsky.app/search?q=aip...
I did a manual review of all results but might have missed a couple all the same. Please let me know if so.
the post below revisits the other main reason for labeling NYT as disinfo. With its resources, recognizing their useful idiocy here - which they did not, they excused themselves unconvincingly - is not an out
bsky.app/profile/capi...
bsky.app/profile/capi...
This is the explanation @nytimes.com got labeled "disinformation".
Which imho is a correct labeling, after the Heitmann scandal.
Thanks, moderation team!
Which imho is a correct labeling, after the Heitmann scandal.
Thanks, moderation team!
July 15, 2025 at 11:07 AM
the post below revisits the other main reason for labeling NYT as disinfo. With its resources, recognizing their useful idiocy here - which they did not, they excused themselves unconvincingly - is not an out
bsky.app/profile/capi...
bsky.app/profile/capi...
PSA: see here for future PSAs on Github FAQ page.
Will relink as needed as this grows irregularly over time
github.com/SteveTownsen...
Will relink as needed as this grows irregularly over time
github.com/SteveTownsen...
FAQ
Public Education Forum Ozone moderation tools. Contribute to SteveTownsend/pef-moderation development by creating an account on GitHub.
github.com
July 14, 2025 at 4:54 PM
PSA: see here for future PSAs on Github FAQ page.
Will relink as needed as this grows irregularly over time
github.com/SteveTownsen...
Will relink as needed as this grows irregularly over time
github.com/SteveTownsen...
Just to emphasize that manual reporting of posts is now preferred. Account-label labeling decision results from a holistic view of its auto- and manual-report history.
Moderation workflow example.
Screenshot of a labeled account, taken from the platform's Ozone tool that we use to manually review the backlog.
This is the moderation event stream on the account, filtered for only reports.
Screenshot of a labeled account, taken from the platform's Ozone tool that we use to manually review the backlog.
This is the moderation event stream on the account, filtered for only reports.
April 23, 2025 at 2:52 PM
Just to emphasize that manual reporting of posts is now preferred. Account-label labeling decision results from a holistic view of its auto- and manual-report history.
Moderation workflow example.
Screenshot of a labeled account, taken from the platform's Ozone tool that we use to manually review the backlog.
This is the moderation event stream on the account, filtered for only reports.
Screenshot of a labeled account, taken from the platform's Ozone tool that we use to manually review the backlog.
This is the moderation event stream on the account, filtered for only reports.
April 23, 2025 at 2:44 PM
Moderation workflow example.
Screenshot of a labeled account, taken from the platform's Ozone tool that we use to manually review the backlog.
This is the moderation event stream on the account, filtered for only reports.
Screenshot of a labeled account, taken from the platform's Ozone tool that we use to manually review the backlog.
This is the moderation event stream on the account, filtered for only reports.
Reviewing appeals, some clarifications.
Automated labeling has been cleaned up and new labels introduced to destigmatize the presence of occasional labels on an account. Some users may be misunderstanding the semantics.
Automated labeling has been cleaned up and new labels introduced to destigmatize the presence of occasional labels on an account. Some users may be misunderstanding the semantics.
April 23, 2025 at 11:02 AM
Reviewing appeals, some clarifications.
Automated labeling has been cleaned up and new labels introduced to destigmatize the presence of occasional labels on an account. Some users may be misunderstanding the semantics.
Automated labeling has been cleaned up and new labels introduced to destigmatize the presence of occasional labels on an account. Some users may be misunderstanding the semantics.
Possibly of interest to some: latest filters for auto-reporting and auto-labeling are at github.com/SteveTownsen...
@maryshine.bsky.social rules with websites are the tracked sites. "track = true" is needed as well - standalone field, not substring of field "actions", which is legacy format.
@maryshine.bsky.social rules with websites are the tracked sites. "track = true" is needed as well - standalone field, not substring of field "actions", which is legacy format.
github.com
April 22, 2025 at 11:39 AM
Possibly of interest to some: latest filters for auto-reporting and auto-labeling are at github.com/SteveTownsen...
@maryshine.bsky.social rules with websites are the tracked sites. "track = true" is needed as well - standalone field, not substring of field "actions", which is legacy format.
@maryshine.bsky.social rules with websites are the tracked sites. "track = true" is needed as well - standalone field, not substring of field "actions", which is legacy format.
Report backlog progress. System is humming.
Completed a pass through accounts with over 100 pending reports.
Almost all either were or are now labeled.
Almost all were for anti-Israel and petrostate disinfo spam.
Surprisingly little traditional russian nonsense. Needs a different DB query imo.
Completed a pass through accounts with over 100 pending reports.
Almost all either were or are now labeled.
Almost all were for anti-Israel and petrostate disinfo spam.
Surprisingly little traditional russian nonsense. Needs a different DB query imo.
April 18, 2025 at 4:45 PM
Report backlog progress. System is humming.
Completed a pass through accounts with over 100 pending reports.
Almost all either were or are now labeled.
Almost all were for anti-Israel and petrostate disinfo spam.
Surprisingly little traditional russian nonsense. Needs a different DB query imo.
Completed a pass through accounts with over 100 pending reports.
Almost all either were or are now labeled.
Almost all were for anti-Israel and petrostate disinfo spam.
Surprisingly little traditional russian nonsense. Needs a different DB query imo.
This reply thread is really a rambling "state of the union" announcement regarding service philosophy and status. Specifically, it recaps labeler stance on
1. antisemitism
2. auto-labeling of slurs
which are the major current points of contention among label recipients.
bsky.app/profile/pef-...
1. antisemitism
2. auto-labeling of slurs
which are the major current points of contention among label recipients.
bsky.app/profile/pef-...
This account was labeled recently. History in the moderation tool is shown, redacted to obfuscate accounts involved.
You maybe don't understand this community labeler is distinct from platform moderation. We cannot take content down, just label it for "the community".
"Your moderators" = me.
You maybe don't understand this community labeler is distinct from platform moderation. We cannot take content down, just label it for "the community".
"Your moderators" = me.
April 17, 2025 at 1:14 PM
This reply thread is really a rambling "state of the union" announcement regarding service philosophy and status. Specifically, it recaps labeler stance on
1. antisemitism
2. auto-labeling of slurs
which are the major current points of contention among label recipients.
bsky.app/profile/pef-...
1. antisemitism
2. auto-labeling of slurs
which are the major current points of contention among label recipients.
bsky.app/profile/pef-...
Following up on a post regarding Al Jazeera's expansive 60/40 disinfo network, in the context of labeling well-known anti-Ukraine propagandists.
Process is to google the person, then review and include their amplifying websites for auto-labeling: no good outlet uses them.
bsky.app/profile/stev...
Process is to google the person, then review and include their amplifying websites for auto-labeling: no good outlet uses them.
bsky.app/profile/stev...
60/40 disinfo screenshot.
The oil-producing Arab states spend vast amounts of money to whitewash their regimes, deflecting towards the west, especially Israel.
Here is Al Jazeera's 'west-friendly' brand AJ+ cheerily sampling Native American cuisine, juxtaposed with anti-US 'eugenics' conspiracism.
The oil-producing Arab states spend vast amounts of money to whitewash their regimes, deflecting towards the west, especially Israel.
Here is Al Jazeera's 'west-friendly' brand AJ+ cheerily sampling Native American cuisine, juxtaposed with anti-US 'eugenics' conspiracism.
April 9, 2025 at 4:31 PM
Following up on a post regarding Al Jazeera's expansive 60/40 disinfo network, in the context of labeling well-known anti-Ukraine propagandists.
Process is to google the person, then review and include their amplifying websites for auto-labeling: no good outlet uses them.
bsky.app/profile/stev...
Process is to google the person, then review and include their amplifying websites for auto-labeling: no good outlet uses them.
bsky.app/profile/stev...
PSA: WIP to clarify labels continues.
Two news ones today, in order to support existing useful rules and future ones.
All use of "fag*ot" and "tr*on" slurs will be labeled realtime soon.
Two news ones today, in order to support existing useful rules and future ones.
All use of "fag*ot" and "tr*on" slurs will be labeled realtime soon.
April 6, 2025 at 4:43 PM
PSA: WIP to clarify labels continues.
Two news ones today, in order to support existing useful rules and future ones.
All use of "fag*ot" and "tr*on" slurs will be labeled realtime soon.
Two news ones today, in order to support existing useful rules and future ones.
All use of "fag*ot" and "tr*on" slurs will be labeled realtime soon.
PSA: more label clarification deployed.
Reiterating commitment to clean up labeling semantics by end of April 2025, including revisiting all auto-labeled content to correct or remove existing labels.
The goal is to alert about content considered harmful without overly infuriating casual amplifiers.
Reiterating commitment to clean up labeling semantics by end of April 2025, including revisiting all auto-labeled content to correct or remove existing labels.
The goal is to alert about content considered harmful without overly infuriating casual amplifiers.
April 5, 2025 at 12:27 PM
PSA: more label clarification deployed.
Reiterating commitment to clean up labeling semantics by end of April 2025, including revisiting all auto-labeled content to correct or remove existing labels.
The goal is to alert about content considered harmful without overly infuriating casual amplifiers.
Reiterating commitment to clean up labeling semantics by end of April 2025, including revisiting all auto-labeled content to correct or remove existing labels.
The goal is to alert about content considered harmful without overly infuriating casual amplifiers.
Reposted by Public Education Forum Moderation
Wish that you spent your Friday doing something besides checking if TikTok was banned or not?
Try out @skylight.social — a video app built upon the same open network as Bluesky. You can login with your Bluesky account, and all of your followers seamlessly come with you.
Try out @skylight.social — a video app built upon the same open network as Bluesky. You can login with your Bluesky account, and all of your followers seamlessly come with you.
🎉 150,000+ USERS!! 🎉
Huge thanks to everyone who's supported us as we build this unbannable TikTok!
There's still so much more we want to build— your support makes it possible.
You all are the best 🥰
Join: skylight.social
Follow Skylight Squad:
@reedharmeyer.bsky.social
@buildwithtori.com
Huge thanks to everyone who's supported us as we build this unbannable TikTok!
There's still so much more we want to build— your support makes it possible.
You all are the best 🥰
Join: skylight.social
Follow Skylight Squad:
@reedharmeyer.bsky.social
@buildwithtori.com
April 4, 2025 at 10:18 PM
Wish that you spent your Friday doing something besides checking if TikTok was banned or not?
Try out @skylight.social — a video app built upon the same open network as Bluesky. You can login with your Bluesky account, and all of your followers seamlessly come with you.
Try out @skylight.social — a video app built upon the same open network as Bluesky. You can login with your Bluesky account, and all of your followers seamlessly come with you.
Here's a good example of auto-reporting synergizing with manual review.
We know there's russian nonsense circulating about Zelensky buying a platinum mine.
The new rule
(Zelensky AND (Northam OR platinum))
got flagged on this account, already auto-flagged for MAGA disinfo posts.
Account now labeled
We know there's russian nonsense circulating about Zelensky buying a platinum mine.
The new rule
(Zelensky AND (Northam OR platinum))
got flagged on this account, already auto-flagged for MAGA disinfo posts.
Account now labeled
April 4, 2025 at 10:31 PM
Here's a good example of auto-reporting synergizing with manual review.
We know there's russian nonsense circulating about Zelensky buying a platinum mine.
The new rule
(Zelensky AND (Northam OR platinum))
got flagged on this account, already auto-flagged for MAGA disinfo posts.
Account now labeled
We know there's russian nonsense circulating about Zelensky buying a platinum mine.
The new rule
(Zelensky AND (Northam OR platinum))
got flagged on this account, already auto-flagged for MAGA disinfo posts.
Account now labeled
Here is the new visual. Clear, non-triggering.
April 3, 2025 at 10:02 PM
Here is the new visual. Clear, non-triggering.
Lack of clarity in labeling, an example epic fail.
This is driving urgent work to revise the labels and reprocess all labeled content so far.
This makes it look like @noupside.bsky.social is spreading disinfo. That is *totally* wrong.
Annoying for the labeled account, confusing for the reader.
This is driving urgent work to revise the labels and reprocess all labeled content so far.
This makes it look like @noupside.bsky.social is spreading disinfo. That is *totally* wrong.
Annoying for the labeled account, confusing for the reader.
April 3, 2025 at 9:50 PM
Lack of clarity in labeling, an example epic fail.
This is driving urgent work to revise the labels and reprocess all labeled content so far.
This makes it look like @noupside.bsky.social is spreading disinfo. That is *totally* wrong.
Annoying for the labeled account, confusing for the reader.
This is driving urgent work to revise the labels and reprocess all labeled content so far.
This makes it look like @noupside.bsky.social is spreading disinfo. That is *totally* wrong.
Annoying for the labeled account, confusing for the reader.
PSA: new labels defined.
First step towards clarifying labeling reasons. Inference should not be needed.
Labeling more abuse is a baby step to encourage civility in atproto-based social media.
Got a sync-compliant firehose? This will work for you.
First step towards clarifying labeling reasons. Inference should not be needed.
Labeling more abuse is a baby step to encourage civility in atproto-based social media.
Got a sync-compliant firehose? This will work for you.
April 3, 2025 at 8:07 PM
PSA: new labels defined.
First step towards clarifying labeling reasons. Inference should not be needed.
Labeling more abuse is a baby step to encourage civility in atproto-based social media.
Got a sync-compliant firehose? This will work for you.
First step towards clarifying labeling reasons. Inference should not be needed.
Labeling more abuse is a baby step to encourage civility in atproto-based social media.
Got a sync-compliant firehose? This will work for you.
PSA regarding recent expansion of 'abusive' labeling.
I don't care about subscriber attrition over this.
I have no way to know who subscribes, and if only my feed is freed up from in-scope slurs, that's a win for me.
There are multiple appeal channels.
Redacted moderation history can be requested.
I don't care about subscriber attrition over this.
I have no way to know who subscribes, and if only my feed is freed up from in-scope slurs, that's a win for me.
There are multiple appeal channels.
Redacted moderation history can be requested.
April 3, 2025 at 2:17 PM
PSA regarding recent expansion of 'abusive' labeling.
I don't care about subscriber attrition over this.
I have no way to know who subscribes, and if only my feed is freed up from in-scope slurs, that's a win for me.
There are multiple appeal channels.
Redacted moderation history can be requested.
I don't care about subscriber attrition over this.
I have no way to know who subscribes, and if only my feed is freed up from in-scope slurs, that's a win for me.
There are multiple appeal channels.
Redacted moderation history can be requested.
Review of informal appeal on behalf of @disqualified.bsky.social
'disinfo' was originally applied under the now-defunct policy of account labeling links to in-scope offsite content. On that basis, perhaps the relevant posts should get 'misinformation" instead. A full review is merited.
'disinfo' was originally applied under the now-defunct policy of account labeling links to in-scope offsite content. On that basis, perhaps the relevant posts should get 'misinformation" instead. A full review is merited.
@stevetownsend0.bsky.social hey hun.. What's going on with labels on this account? He's harmless and just hates British fascists.
April 3, 2025 at 11:00 AM
Review of informal appeal on behalf of @disqualified.bsky.social
'disinfo' was originally applied under the now-defunct policy of account labeling links to in-scope offsite content. On that basis, perhaps the relevant posts should get 'misinformation" instead. A full review is merited.
'disinfo' was originally applied under the now-defunct policy of account labeling links to in-scope offsite content. On that basis, perhaps the relevant posts should get 'misinformation" instead. A full review is merited.
PSA
Rolled out new labels to automatically identify objectively high usage of "facets" below.
Facets covered:
- offsite links (URLs)
- user-mentions (@blablabal.myfriends.xyz)
- hashtags
- the sum of the above in any post/profile
Rolled out new labels to automatically identify objectively high usage of "facets" below.
Facets covered:
- offsite links (URLs)
- user-mentions (@blablabal.myfriends.xyz)
- hashtags
- the sum of the above in any post/profile
March 31, 2025 at 1:33 PM
PSA
Rolled out new labels to automatically identify objectively high usage of "facets" below.
Facets covered:
- offsite links (URLs)
- user-mentions (@blablabal.myfriends.xyz)
- hashtags
- the sum of the above in any post/profile
Rolled out new labels to automatically identify objectively high usage of "facets" below.
Facets covered:
- offsite links (URLs)
- user-mentions (@blablabal.myfriends.xyz)
- hashtags
- the sum of the above in any post/profile