Meta & Children's Online SafetyEmail: Fwd: Pre-read for our conversation tomorrow
Raw text
Message From: Sent: To: Subject: Arturo Bejar [/O=THEFACEBOOK/OU=EXCHANGE ADMINISTRATIVE GROUP (FYDIBOHF23SPDLT)/CN=RECIPIENTS/CN= 10/14/2021 11:56:05 PM Alicia @fb.com]; David Gillis @fb.com] Fwd: Pre-read for our conversation tomorrow Hi Alicia, Dave, Sharing with you the pre-read of my conversation with Adam tomorrow, I will keep you posted. Arturo Begin forwarded message: From: Arturo Bejar @fb.com> Subject: Pre-read for our conversation tomorrow Date: October 14, 2021 at 4:40:38 PM PDT To: Adam Mosseri Hi Adam, @fb.com> In order to make the best use of our time tomorrow I put together a short pre-read that I've vetted with the team in well-being. Data points (last 7 days/more than 7 days) Have you ever received unwanted sexual advances on Instagram? 13-15 year olds: 13%/27% Have you ever seen anyone discriminating against people on Instagram because their gender, religion, race, sexual orientation, or another part of their identity? 13-15 year olds: 26%/31% Has anyone done any of these things to you on Instagram? Insulted or disrespected you, contacted you in an in appropriate way, damaged your reputation, threatened you, excluded you or left you out. 13-15 year olds: 11%/25% Have you ever felt worse about yourself because of other peoples' posts on Instagram? 13-15 year olds: 21%/23% Questions: What should be the goal/number of 13-15 year olds on each of the BEEF categories? HIGHLY CONFIDENTIAL HIGHLY CONFIDENTIAL (COMPETITOR) HIGHLY CONFIDENTIAL (ATTORNEY'S EYES ONLY) PLTF EX PLTF EX-00956 FTC-META-003754779 ΜΕΤΑ3047MDL-031-00118481 METANMAG-002-01698104 PLTF EX-00956_0001 Are users able to express these experiences to us in the product? (e.g. for unwanted sexual advances, or negative comparison you can't) What would we build if >90% of the content which drives these experiences is not policy violating or borderline? If you'd like to look at the data directly, here is the data by age: https://docs.google.com/spreadsheets/d/10rR5hbK4v1W-2QmUUMSLGiUFfljgm9ngf- 3gZqVJQEU/edit#gid=951724808 I also find it helpful to put the data in the context of the questions (which convey better than our labels what people are experiencing): https://docs.google.com/spreadsheets/d/1gQGpXK7UkC_Z7_C41Z8SLqb4Puom4yv4KfX0Zs6- wdg/edit#gid=2045546404 Recommendations 1. For Instagram to set goals based on TRIPS/BEEF, use people's experience as the north star for the work: 1. What would you build if the goal was to get to 1% unwanted sexual advances? Or 3% witnessed hate? Or 2% target of bullying? 2. Change the use of the word 'perceived' to 'experienced' - people don't perceive being harassed. 2. Provide features that help us understand the issues and content that people are experiencing so that we may develop interventions/features that help them and improve the community over time. 1. Secondary actions to block/delete where we get user experience data. This has been difficult to date because the team has been running into XFN limitations on understand efforts. 2. Make the reporting flow, or add experiences at the beginning to make people feel heard and supported with what they are experiencing, as well as generate insights on the issues they are having. 3. Invest in features that help us learn how to develop and maintain social norms, and actor feedback. Can we shift the conversation into one of hope and leadership? Everyone in the industry has the same problems right now. Prevalence-based measures, while necessary, don't create a safe and supportive community, you're always behind the latest harmful thing. We have few visible features that help create a safe and supportive environment for everyone. There is a great product opportunity in figuring out the features that make a community feel safe and supportive. It is possible and important to work these issues in partnership with other industry leaders and academics. We have much learn about each of these issues. I believe is possible to help create public conversation on these topics for good. A point which might be good for you to know (which I did not put in the document reviewed by the team) is that many employees I've spoken who are doing this work (and are of different levels) are distraught about how the last few weeks have unfolded. These are people who love FB/IG, and are heart/mission driven to the work. HIGHLY CONFIDENTIAL HIGHLY CONFIDENTIAL (COMPETITOR) HIGHLY CONFIDENTIAL (ATTORNEY'S EYES ONLY) FTC-META-003754780 META3047MDL-031-00118482 METANMAG-002-01698105 PLTF EX-00956_0002 Thanks, Arturo HIGHLY CONFIDENTIAL HIGHLY CONFIDENTIAL (COMPETITOR) HIGHLY CONFIDENTIAL (ATTORNEY'S EYES ONLY) FTC-META-003754781 META3047MDL-031-00118483 METANMAG-002-01698106 PLTF EX-00956_0003 Message From: Sent: To: Subject: Arturo Bejar [/O=THEFACEBOOK/OU=EXCHANGE ADMINISTRATIVE GROUP (FYDIBOHF23SPDLT)/CN=RECIPIENTS/CN= 10/14/2021 11:46:06 PM Yoav Shapira @fb.com]; Samir Fwd: Pre-read for our conversation tomorrow @fb.com]; Miki Rothschild @fb.com] Yoav, Samir, Miki - sent Adam the pre-read, I added a paragraph at the bottom about how I believe this is also a retention question. The bigger the gap gets with how the world sees these issues, I believe it will become harder to retain great people. My hope is for a bigger mission than reducing certain harm by some amount, but how could this team lead the industry on this space. I also added the negative comparison data point, the question asked is powerful, and the numbers are meaningful. Let me know any questions or feedback, will report back after my chat tomorrow, if you'd like to talk before that I'm at Thanks, Arturo Begin forwarded message: From: Arturo Bejar r@fb.com> Subject: Pre-read for our conversation tomorrow Date: October 14, 2021 at 4:40:38 PM PDT To: Adam Mosseri @fb.com> Hi Adam, In order to make the best use of our time tomorrow I put together a short pre-read that I've vetted with the team in well-being. Data points (last 7 days/more than 7 days) Have you ever received unwanted sexual advances on Instagram? 13-15 year olds: 13%/27% Have you ever seen anyone discriminating against people on Instagram because their gender, religion, race, sexual orientation, or another part of their identity? 13-15 year olds: 26%/31% HIGHLY CONFIDENTIAL HIGHLY CONFIDENTIAL (COMPETITOR) HIGHLY CONFIDENTIAL (ATTORNEY'S EYES ONLY) FTC-META-003754797 META3047MDL-031-00118484 METANMAG-002-01698107 PLTF EX-00956_0004 Has anyone done any of these things to you on Instagram? Insulted or disrespected you, contacted you in an in appropriate way, damaged your reputation, threatened you, excluded you or left you out. 13-15 year olds: 11%/25% Have you ever felt worse about yourself because of other peoples' posts on Instagram? 13-15 year olds: 21%/23% Questions: What should be the goal/number of 13-15 year olds on each of the BEEF categories? Are users able to express these experiences to us in the product? (e.g. for unwanted sexual advances, or negative comparison you can't) What would we build if >90% of the content which drives these experiences is not policy violating or borderline? If you'd like to look at the data directly, here is the data by age: https://docs.google.com/spreadsheets/d/10rR5hbK4v1W-2QmUUMSLGiUFfljgm9ng 3gZqVJQEU/edit#gid=951724808 I also find it helpful to put the data in the context of the questions (which convey better than our labels what people are experiencing): https://docs.google.com/spreadsheets/d/1gQGpXK7UkC Z7 C41Z8SLqb4Puom4yv4KfX0Zs6- wdg/edit#gid=2045546404 Recommendations 1. For Instagram to set goals based on TRIPS/BEEF, use people's experience as the north star for the work: 1. What would you build if the goal was to get to 1% unwanted sexual advances? Or 3% witnessed hate? Or 2% target of bullying? 2. Change the use of the word 'perceived' to 'experienced' - people don't perceive being harassed. 2. Provide features that help us understand the issues and content that people are experiencing so that we may develop interventions/features that help them and improve the community over time. 1. Secondary actions to block/delete where we get user experience data. This has been difficult to date because the team has been running into XFN limitations on understand efforts. 2. Make the reporting flow, or add experiences at the beginning to make people feel heard and supported with what they are experiencing, as well as generate insights on the issues they are having. 3. Invest in features that help us learn how to develop and maintain social norms, and actor feedback. Can we shift the conversation into one of hope and leadership? Everyone in the industry has the same problems right now. HIGHLY CONFIDENTIAL HIGHLY CONFIDENTIAL (COMPETITOR) HIGHLY CONFIDENTIAL (ATTORNEY'S EYES ONLY) FTC-META-003754798 META3047MDL-031-00118485 METANMAG-002-01698108 PLTF EX-00956_0005 Prevalence-based measures, while necessary, don't create a safe and supportive community, you're always behind the latest harmful thing. We have few visible features that help create a safe and supportive environment for everyone. There is a great product opportunity in figuring out the features that make a community feel safe and supportive. It is possible and important to work these issues in partnership with other industry leaders and academics. We have much learn about each of these issues. I believe is possible to help create public conversation on these topics for good. A point which might be good for you to know (which I did not put in the document reviewed by the team) is that many employees I've spoken who are doing this work (and are of different levels) are distraught about how the last few weeks have unfolded. These are people who love FB/IG, and are heart/mission driven to the work. Thanks, Arturo HIGHLY CONFIDENTIAL HIGHLY CONFIDENTIAL (COMPETITOR) HIGHLY CONFIDENTIAL (ATTORNEY'S EYES ONLY) FTC-META-003754799 META3047MDL-031-00118486 METANMAG-002-01698109 PLTF EX-00956_0006