Sign in Subscribe
4 min read

'Error rates are too high', OpenAI's red teaming research and Nighat's story

The week in content moderation - edition #273

Hello and welcome to Everything in Moderation's Week in Review, your need-to-know news and analysis about platform policy, content moderation and internet regulation. It's written by me, Ben Whitelaw and supported by members like you.

Can you believe it's that time of year again? No, not the festive period but EiM's 2024 Audience Survey, in which I ask you — loyal subscribers of the newsletter — to share your feedback on how EiM is doing and how it could be more useful to your work and life.

This year, I'm teaming up with Katie Harbath from the brilliant Anchor Change to understand what topics you care about so that together we can better serve T&S practitioners and those working in close proximity. Please take a 3-4 minutes to fill it in.

Last year's audience survey led to the creation of Ctrl-Alt-Speech and T&S Insider, the Monday newsletter written by Alice Hunsberger. So do take a moment to tell me what you like, need and are keen to see improved — I read every response.

Today's edition is a lighter than usual due to the slow wifi on my United flight back to London. But I hope it does the trick at least until this week's podcast comes out later today. Thanks for reading — BW

Today's edition is in partnership with Concentrix, the technology and services leader driving trust, safety, and content moderation globally

In November, the European T&S community came together in Amsterdam for the T&S Festival, where Concentrix hosted its European T&S meet up.

For this third edition, the T&S community was invited to collaborate on three pressing points: improving moderators’ wellbeing, evaluating the T&S ROI, and the fight against CSAM. Attendees also had the opportunity to connect with their peers and develop their personal network in the industry.

If you missed T&S Festival, don't worry — follow us on LinkedIn to be informed of future networking events.

Be first to know about future events

Policies

New and emerging internet policy and online speech regulation

It wouldn’t be the first time that Meta had overdone it with its enforcement of its policies but its interesting to hear that Sir Nick Clegg admit that the company sometimes went too far when it sought to “remove or restrict innocuous or innocent content” during the Covid-19 pandemic. In a briefing reported by the FT, Meta’s president of global affairs spoke about Mark Zuckerberg's desire to play an active role shaping tech policy under Donald Trump and the fact that “[moderation] error rates are still too high”.

Has he forgotten? Over-enforcement on Threads — in part due to the difficulty of automated moderation — has been one of the reason why users have departed the platform in recent weeks (EiM #271). The irony is that Clegg has also been seen giving prizes out at glitzy AI hackathons - I wonder if he has clocked that the two could be connected?

Also in this section...

Products

Features, functionality and technology shaping online speech

New ways to automate the red teaming of AI models and an army of expert human testers are two new ways that OpenAI has been playing with to emphasise its approach to safety. In what has been an embarassing week for the company, it published two papers this week explaining how it stress tests models, in an attempt to convince the pubic and regulators it is capable of mitigating harm. However, as the MIT Technology Review piece points out, the science of evaluation is lagging behind model development, which could be a concern if not unaddressed.

Platforms

Social networks and the application of content guidelines

An “important moment” for Telegram and for the child safety community: the Dubai-based messaging app announced on Tuesday that it would partner with the Internet Watch Foundation to address CSAM on its platform. The agreement gives Telegram access to the database of known abuse material and tools to help remove it in a timely manner. Telegram’s CEO Pavel Durov remains on bail since his arrest in France in August (EiM #261).

Also in this section...

People

Those impacting the future of online safety and moderation

Many folks working in trust and safety didn’t grow up wanting to keep the internet safer. Some wanted to be in law enforcement — like Meta’s Hannah Byrne in this Intercept profile — or in human rights or product management. 

Others, like Nighat Dad, did law before turning her attention to empowering women online. The Pakistani internet activist channeled her personal struggles with domestic abuse and digital surveillance into a mission to empower women online. She founded the Digital Rights Foundation in 2012 to combat online harassment and advocate for women's digital rights and was a founding member of the Oversight Board.

Her story is captured in a profile in the new edition of Wired and reminds me of the often vocational nature of roles in online speech and internet regulation, where “facing those restrictions” — as Nighat says —  helps inform an understanding of its importance in the world. 

Posts of note

Handpicked posts that caught my eye this week

  • “Excited to finally share my TED Talk! 📺 - The real-world danger of online myths.” - Moonshot’s Vidhya Ramalingam with something to watch this weekend. 
  • “Had an insightful, generative conversation with Laura Higgins of Roblox and Jordan Shapiro of Sesame Workshop on topics including digital self-harm and other complexities of youth behavior in online environments” - Sameer Hinduja, professor at Florida Atlantic University, caters to the podcast-crowd.
  • “Wonderful to be in Abu Dhabi at the WeProtect Global Alliance Summit- convening so many, doing such crucial work, across child safety globally.” - Resolver's Henry ‘H’ Adams on the conference trail.

Published by:

Ben Whitelaw

Member discussion