Meta & Children's Online SafetyFacebook's Impact on Well-Being: A Deactivation Study [Pilot]
Raw text
Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 1 of 40 AMENDED Exhibit 90 PLAINTIFFS' OMNIBUS OPPOSITION TO DEFENDANTS' MOTIONS FOR SUMMARY JUDGMENT Case No.: 4:22-md-03047-YGR MDL No. 3047 In Re: Social Media Adolescent Addiction/Personal Injury Products Liability Litigation Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 2 of 40 Document Provided in Native Format HIGHLY CONFIDENTIAL (COMPETITOR) META3047MDL-047-00058006 Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 3 of 40 DEPRECATED: Facebook's Impact on Well-Being: A Deactivation Study [Pilot] QUIP: https://fb.quip.com/PU4xAUxokb9r#RcCACARp5TE Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 4 of 40 tl;dr 1 Inan n effort to understand Facebook's impact on people's lives, we ran two 1-week experiments in which some people stopped using Facebook and others used Facebook as normal. We found that those who stopped using Facebook for a week: Reported lower feelings of depression, anxiety, loneliness, and social comparison Reported fewer interactions with community members Reported more time watching tv/movies and spent less time on social media Had lower news knowledge Expectations appear to play a strong role in this study: the effects found are mostly linked to people's prior beliefs on whether or not Facebook is good or bad for them. For example, the reduction in feelings of depression, anxiety, and loneliness appeared only among participants who believed Facebook is bad for them. We explored other areas like political polarization, time spent on other activities, and social interaction quality, and detected no major differences between those who used Facebook and those who didn't. Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 5 of 40 4 Research Design We ran two independent deactivation studies and combined the samples for our analysis. Slightly more than half of participants were invited to deactivate FB for 1 week (N= 3067); the rest continued to use FB as normal. (N= 2487). All participants double-consented to participate in this study. The goal of this study was to provide a framework for a longer and larger study involving both FB and IG. Participants could receive up to $35 dollars to deactivate and up to $11 to complete 3 surveys. The first study was run Oct/Nov 2019 and the second study was run Feb/Mar 2020. We collected survey data, mobile sensor data, and FB/IG log data: Survey data on emotional health (e.g., loneliness, depression, affect, etc.), political health (e.g., affective polarization, issue polarization, news knowledge, etc.), and social health (e.g., time talking to friend, meaningful interactions, etc.) Mobile sensor data (e.g., times spent on news, video, social apps, etc.) FB/IG log data (e.g., Time spent on FB, graph size, etc.) Because this study could not be conducted double-blinded, we attempted a study that could take into account potential expectation effects. We had three conditions in our experiment: Neutral prime (participants were not given information on the purpose of the study) Well-being prime (participants told the study was about well-being; so we explored the political-related outcomes in this condition, but not the social/emotional ones) Polarization prime (participants told the study was about polarization; so we explored the emotional- and social-related outcomes in this condition, but not the political ones) Nielsen ran the study on our behalf and FB branding was revealed at the end; people had the opportunity to pull their data prior to sharing back with Facebook. Pha 31 Screener Survey Phase 2 Baseline Survey Phase 3 Phase 4 1 week deactivation from FB Endline Survey Use FB as normal For background on extemal research already conducted in this space, see here. We received advisory from extemal experts on the design. Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 6 of 40 What happens when people stop using Facebook? Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 7 of 40 People who who stopped using Facebook for a week reported lower feelings of depression, anxiety, loneliness, and social comparison. They also had lower news knowledge scores, reported fewer interactions with community members and more time watching TV and movies. No statistically effects detected on other areas explored including political polarization. Meaningfulness of interactions Decreased Increased Time watching tv/movies" Positive affect Life satisfaction Self-esteem Social support Negative affect Loneliness" Anxiety Depression Social Comparison" -1.0 -0.5 0.0 0.5 Standardized effect of 1-week FB deactivation 1.0 Time with friends & family Time exercising Issue polarization Affective polarization News knowledge" Communication (far-away family & friends) Communication (community members)" *Effect size is scaled by the standard deviation of the outcome under control in the endline period. -1.0 Decreased Increased -0.5 0.0 0.5 Standardized effect of 1-week FB deactivation 1.0 Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Meaningfulness of interactions Positive affect Life satisfaction Self-esteem Social support Negative affect Loneliness Anxiety Depression Social Comparison' -1.0 Decreased Increased -0.5 0.0 0.5 Standardized effect of 1-week FB deactivation 1.0 Time watching tv/movies* Time with friends & family Time exercising Issue polarization Affective polarization News knowledge* Communication (far-away family & friends) Communication (community members)* -1.0 Page 8 of 40 Decreased Increased -0.5 0.0 0.5 Standardized effect of 1-week FB deactivation *Effect size is scaled by the standard deviation of the outcome under control in the endline period. 1.0 Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 9 of 40 What role do expectations play in these findings? Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 10 of 40 W5knew going into this study that the expectations and beliefs people have about Facebook would be a 1 chulenge we needed to consider. In an ideal world when studying impact, people are not able to guess the study's objective as this can directly/indirectly influence their thoughts and behaviors, and thus the findings. In a deactivation study like this, people may consciously/unconsciously garner what the study is trying to get at based on media coverage, prior academic research, etc. As a result, when deactivating their accounts, people may think they should feel more happy, and report so in a survey, even when in actuality their levels of happiness remain the same after they stopped using Facebook. Thus, in this study, to get signal on people's expectations, we asked people about their beliefs about Facebook's impact on their lives. To assess the role of beliefs, we compare the effect of stopping using Facebook across the following groups: People who think FB make them more polarized vs. people who think FB makes them less polarized when examining the impact of FB deactivation on political outcomes. People who think FB worsens their well-being vs. people who think FB improves their well-being when examining the impact of FB deactivation on social and emotional outcomes. The effects we've identified appear to be linked with people's a priori beliefs. For example, we find that the reduction of feeling of depression, anxiety, loneliness, and affect associated with FB happened only among people who thought FB is bad for them prior to participating in the experiment. Similarly, only people who thought Facebook was good for them reported a reduction in communication with community members. Social comparison was the only factor that remained for both groups. Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 11 of 40 The effects we've identified so far are linked with people's beliefs about how good or bad Facebook is for them. For example, the reduction in feelings of depression, anxiety, and loneliness appeared only among participants who believed Facebook is bad for them prior to participating in the experiment. Meaningfulness of interactions Positive affect Life satisfaction Self-esteem Social support Loneliness Depression Anxiety Negative affect Social Comparison Decreased Increased FB is bad FB is good -1.0 -0.5 CO 0.5 1.0 Standardized effect of 1-week FB deactivation Time with friends & family Time exercising Time watching tvimovies Issue polarization Affective polarization News knowledge Communication (far-away family & friends) Communication (community members) Decreased Increased FB is bad FB is good -1.0 -0.5 0.0 0.5 1.0 Standardized effect of 1-week FB deactivation Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Decreased Increased Meaningfulness of interactions Positive affect Life satisfaction Self-esteem Social support Loneliness Depression Anxiety Negative affect Social Comparison • FB is bad • FB is good -1.0 -0.5 0.0 0.5 Standardized effect of 1-week FB deactivation 1.0 Page 12 of 40 Decreased Increased Time with friends & family Time exercising Time watching tv/movies Issue polarization Aflective polarization News knowledge Communication (far-away family & friends) Communication (community members) FB is bad • FB is good -1.0 -0.5 0.0 0.5 Standardized effect of 1-week FB deactivation 1.0 Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 13 of 40 Can we overcome the role that expectations play in a deactivation study? Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 14 of 40 We knew expectations could be a major limitation with this study, so we attempted a study design that would account for this. 6710 W1 used prompts to distract respondents about the goal of the study in effort to prevent expectation effects from bi-zing people's responses on a particular topic. Two-thirds of respondents were shown a prompt summarizing 10 research on FB's impact on "well-being" or political polarization before starting the endline survey. Distract Prime. When participants were told the study was about "well-being," we actually analyzed the questions about political outcomes (and vice versa). Amplify Prime. When participants were told the study was about "polarization," we actually studied the questions about social and emotional outcomes. The premise here was that we hoped these primes would bury the lead further on the study and help overcome expectation effects. However, the primes did not have an impact on the effect of FB deactivation. The treatment effects were on average the same across all prime groups. Why were the primes ineffective? Two main hypotheses: (1) People are able to judge what is good or bad for them or (2) primes were not very strong. The analysis of mobile data provides suggestive evidence consistent with the second hypothesis: people do not change overall time spent on mobile, but report higher amount time spent with friends following deactivation. Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Meaningfulness Decreased Increased of interactions Social Comparison Anxiety Self-esteem Social support Life satisfaction Negative affect Positive affect Loneliness Depression • Distract Prime • Amplify Prime -1.0 -0.5 0.0 0.5 1.0. Standardized effect of 1-week FB deactivation Issue polarization Page 15 of 40 Decreased Increased Affective polarization News knowledge Time with friends & family Time exercising Time watching tv/movies Communication (community members) Communication (far-away family & friends) • Distract Prime Amplify Prime -1.0 -0.5 0.0 0.5 1.0 Standardized effect of 1-week FB deactivation Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 16 of 40 These charts highlight the effects of each factor broken down by primes. Overall, we do not see indication that the primes had an effect. Meaningfulness of interactions Social Comparison Anxiety Self-esteem Social support Life satisfaction Negative affect Positive affect Loneliness Depression Decreased Increased • Distract Prime Amplify Prime -1.0 -0.5 0.0 0.5 1.0 Standardized effect of 1-week FB deactivation Decreased Increased Issue polarization Affective polarization Newe knowledge Time with friends & family Time exercising Time watching tv/movies Communication (community members) Communication (far-away family & friends) • Distract Prime Amplify Prime -1.0 -0.5 0.0 0.5 1.0 Standardized effect of 1-week FB deactivation Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 17 of 40 The majority of respondents think the study is about the impact of FB on well- being and political polarization, even after being exposed to a prime. Percent who believe study was about impact of FB on WB and politics Percent of participants 0.0 0.2 0.4 0.6 0.8 1.0 Neutral Politics SWB Prime Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 18 of 40 What are people doing on their phones when they stop using Facebook? Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 19 of 40 3 People who stop using Facebook, unsurprisingly end up spending less time on Facebook, but also spend less time on social media overall. We don't see evidence that people are using their phones any more or less overall. Time on browser Time watching video Time gaming Messaging and calls Mobile time spent Instagram time Time on Social Media* Decreased Increased Time on Facebook -60 -40 -20 0 20 40 60 Effect of 1-week FB deactivation (in minutes) Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Time on browser Time watching video Time gaming Messaging and calls Mobile time spent Instagram time Time on Social Media* Time on Facebook* Decreased Increased -60 -40 -20 0 20 40 60 Effect of 1-week FB deactivation (in minutes) Page 20 of 40 Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 21 of 40 12 M takeaways Similar to external studies, we find evidence of both positive and negative impacts Facebook may have on people's lives. The most robust finding in our study is that Facebook increases feelings of social comparison (regardless of beliefs). People's perceptions of Facebook seem to impact the findings -- and this is important to note for our future research, and interpretation of others' findings. We haven't cracked a way to account for this yet or understand it fully. Deactivation studies are a promising direction to invest in further. Proposed next steps Explore FB data and mobile sensor data alongside survey outcomes. We have been blocked because of COVID WFH (since the data must remain on campus). Analyses we'd prioritize: How people spend their time on FB in relationship to beliefs A deeper understanding of how people spent their time when they stopped using Facebook Additional studies [w/ academics?] (addressing limitations of the current study) and exploring a longer timeframe Discussion questions Immediate thoughts and reactions? Open questions we can answer with the current data set? Is it worth getting permission to finish the data match and analyses on campus? If so, what is the priority level of this? Are we allowed to share this more widely? Do we want to get this to a place where we'd share this more widely? Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 22 of 40 Appendix Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 23 of 40 This needs cleaning... Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 24 of 40 Limitations 13 Panel Survey a. Sample is composed of members one of Nielsen's survey panel b. Critics may claim that "professional" respondents answer to surveys differently 2. Combining Survey Responses with FB Data a. The strict protocol to combine survey and log data has bugs b. To this date we are able to successfully link record for about 50% of respondents on the sample 3. Generalizability a. Given the nature of the design (impact of deactivation among people who value the app by a certain monetary amount) one cannot generalize effects to FB MAP. 4. Credibility a. While we brought Nielsen on for this very reason, some may still not trust the findings because Facebook was involved. Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 25 of 40 We estimate the impact of FB deactivation on 18 different survey-based outcomes grouped in three broad categories: (1) Well-Being; (2) Social Interactions and; (3) Political Polarization. Well-Being Social Interactions Political Polarization Depression, Anxiety, Loneliness, Meaningful Social Interactions, Time News Knowledge, Issue, and Positive and Negative Affective, Life Satisfaction, Self-Esteem, Social Comparisons, Social Support with Family and Friends, Time Exercising, Time Watching TV, Affective Polarization Communication with Community, Friends, and Family Members We estimate the effect of FB deactivation for the people who deactivated FB because they were encouraged to do so. In our context, this corresponds to the difference in means between treatment and control groups in a given outcomes divided by the proportion of people who deactivated FB in the treatment group. We find FB deactivation led to a decrease in depression, anxiety, and social comparisons. The magnitude of the effect ranged from X to Y standard deviation of the outcome in the control group. Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 26 of 40 5 T14 the extent there are effects, these are driven by beliefs people have about FB prior to deactivation FB is bad (good) defined as bottom (top) two responses to: In your opinion, to what extent does using Facebook make you more or less extreme in your political views? In your opinion, to what extent does using Facebook improve or hurt your well-being? Time with friends & family Time exercising Time watching movies Issue polarization News knowledge Communication far-away family & friend Communication -1.0 Decressed Increased FB bad FB & good 05 00 05 Standardized effect of 1-wook FB deactivation 1.0 Meaningfulness of interactions Posithe affect Life satisfaction Sed Social support Loneliness Depression Avtalety Negathe affect Deceased Increased FB bad FB is good -1.0 -05 0.0 0.5 1.0 Standardized effect of 1-wock FB deactivation Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Data* Survey data Mobile sensor data FB/IG log data Page 27 of 40 Emotional health (e.g., News apps loneliness, positive affect) Video apps Photo apps Time spent on FB Time spent on messenger issue) Polarization (e.g., affective, Games apps Self-reported activities and interactions(e.g., news consumption, time with friends and family, advice received) Social apps other Search apps Shopping apps Reading apps Music/Podcasts apps Other (search engines, docs, etc.) Messaging & call apps Time spent on IG Email apps Time spent watching video Graph size Times spent in Groups Interactions on FB Interactions on IG Why is our study unique? Panel quality Meter quality Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 16 Variables & Operationalization - Emotions Page 28 of 40 1. Depression 6. Social Support a. Below are some ways you might have felt during the last 7 days. Please tell us how often during the last 7 days you felt depressed. a. How much do you agree or disagree with the following statements? i. 1 item; 5-point: Never (1) - Very often (5) b. 2. Loneliness a. b. During the last 7 days, how often did you feel that you lacked companionship? During the last 7 days, how often did you feel left out? C. C. During the last 7 days, how often did you feel isolated from others? d. i. 3 items (simple average); 5-point: Never (1) - Always (5) During the last 7 days, there were people I could depend on to help me. During the last 7 days, there were people who gave me support and encouragement. During the last 7 days, there were people who appreciated me as a person. 3. Negative Affect i. a. b. How much do you agree or disagree with the following statements? During the last 7 days, I felt negative most of the time. 3 items (simple average); 5-point: Strongly disagree (1) - Strongly agree (5) C. During the last 7 days, I experienced unhappy feelings most of the time. 7. a. d. During the last 7 days, I felt bad most of the time. Self-Esteem How much do you agree or disagree with the following statement? i. 3 items (simple average); 5-point: Strongly disagree (1) - Strongly agree (5) b. During the last 7 days, I had high self-esteem. 4. Positive Affect i. a. How much do you agree or disagree with the following statements? 1 item; 5-point: Strongly disagree (1) - Strongly agree (5) b. During the last 7 days, I felt positive most of the time. 8. Anxiety C. During the last 7 days, I felt happy most of the time. a. d. During the last 7 days, I felt good most of the time. i. 3 items (simple average); 5-point: Strongly disagree (1) - Strongly agree (5) Below are some ways you might have felt during the last 7 days. Please tell us how often during the last 7 days you felt anxious. 5. Life Satisfaction i. 1 item; 5-point: Never (1) - Very often (5) a. How much do you agree or disagree with the following statements? 9. b. During the last 7 days, in most ways my life has been close to my ideal. a. C. During the last 7 days, I have been satisfied with my life. d. During the last 7 days, my life has been going well. b. i. 3 items (simple average); 5-point: Strongly disagree (1) - Strongly agree (5) Social Comparison During the last 7 days, how often did you compare your own accomplishments to those of other people? During the last 7 days, how often did you feel worse about yourself after comparing yourself to someone else? i. 2 items (simple average); 5-point: Never (1) - Very often (5) Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 29 of 40 Variables & Operationalization - Social Interactions Communication: Distant Friends/Family 1. a. During the last 7 days, how often did you communicate with people from the following groups? By "communicate," we mean having interactions that are in person, online, or through calling or messaging. Friends and family who live far away b. i. 2. Binary coding; 5-point: Not at all in the last 7 days (0), Only once in the last 7 days, A few times in the last 7 days, Once per day, A few times per day (1) Communication: Community/Group Members a. During the last 7 days, how often did you communicate with people from the following groups? By "communicate," we mean having interactions that are in person, online, or through calling or messaging. Community members/Group members b. i. 3. 4. 5. Binary coding; 5-point: Not at all in the last 7 days (0), Only once in the last 7 days, A few times in the last 7 days, Once per day, A few times per day (1) Relative Activity: Television a. b. During the last 7 days, did you spend more or less time than you usually do doing the following activities? Watching TV shows or movies i. Trinary coding; 5-point: I never do this, A lot less time, A little less time (-1), Neither more nor less time (0), A little more time, A lot more time (1) Relative Activity: Exercise During the last 7 days, did you spend more or less time than you usually do doing the following activities? Exercising, working out, or playing sports a. b. i. Trinary coding; 5-point: I never do this, A lot less time, A little less time (-1), Neither more nor less time (0), A little more time, A lot more time (1) Relative Activity: Friend/Family Hangout During the last 7 days, did you spend more or less time than you usually do doing the following activities? Hanging out with friends and family i. a. b. Trinary coding; 5-point: I never do this, A lot less time, A little less time (-1), Neither more nor less time (0), A little more time, A lot more time (1) 6. a. During the last 7 days, how meaningful were your interactions with people? i. 1 item; 5-point: Not at all meaningful (1) - Extremely meaningful (5) Meaningful Interactions Removed: 1. Playing video games or board games 2. Reading books or listening to podcasts/audiobooks 3. Running errands or doing household chores 4. Working or studying 5. Talking to People I haven't talked to in a long time 6. Talking to Strangers 7. Find out about a job opportunity 8. Learn about a major life event of someone you care about (e.g., wedding, birth, new job, new house, etc.) Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 30 of 40 Attend an event (e.g., a party, concert, game, conference, etc.) 10. Buy, sell, trade, or give away something to someone in your community 1. Volunteer or donate to a cause Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 31 of 40 Variables & Operationalization - Polarization & News 1. Affective Polarization a. How do you feel about people who support the Democratic Party? b. How do you feel about people who support the Republican Party? i. Difference of 2 items; 5-point: very negative (1) - very positive (5); scaled according to party affiliation/lean 2. Ideological Polarization a. b. C. d. There has been some discussion about abortion during recent years. Which one of the opinions below best agrees with your view? In general, do you support or oppose same-sex marriage? Do you think the federal government should make it more difficult for people to buy a gun than it is now, make it easier for people to buy a gun, or keep these rules about the same as they are now? Which is more important: reducing the federal budget deficit, even if the unemployment rate remains high, or reducing the unemployment rate, even if the federal budget deficit remains high? e. Do you think the number of immigrants from foreign countries who are permitted to come to the United States to live should be <decreased/increased>? i. 5 items (abortion, marriage, guns, employment, immigration): custom 4-point or 5-point scales, coded to liberal or conservative position answers; average of absolute value difference between liberal positions and conservative positions 3. News Knowledge a. News knowledge questions (only baseline examples shown below) i. President Trump recently pulled US troops from which of the following countries? ii. Abhijit Banerjee, Esther Duflo, and Michael Kremer received the 2020 Nobel Prize in iii. economics for their contributions to combating which of the following issues? iv. Last week, the CEO of which of the following companies passed away? V. The White House announced that the 2020 meeting of the G-7 countries will take place at which venue? vi. NASA made history last week for which of the following? vii. viii. ix. Rick Perry announced his plans to resign as secretary from which of the following departments? Which of the following musicians fell off stage while performing last week? 1. 7 items (baseline: troops, Nobel, CEO, G7, NASA, Rick Perry, Lady Gaga; endline: justice prize, fires, quantum, impeachment, prime minister, affair, Uluru): 4 unique answers; dichotomize correct/incorrect answers and sum Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Primes Politics Prime: Well-Being Prime: Facebook reduces political polarization Facebook amplifies political polarization Some research finds Facebook users are exposed to diverse political views, which makes people less extreme in their political views and more understanding of others. Some research finds Facebook users are only exposed to their own political views, which makes people more extreme in their political views and less understanding of others. Facebook is good for well-being Some research finds Facebook users have more meaningful interactions with people, which makes them feel more happy and more connected to others. Facebook is bad for well-being Some research finds Facebook users have less meaningful interactions with people, which makes them feel less happy and less connected to others. The order of the columns was randomized in the survey. Page 32 of 40 Proportion of FB Active Accounts Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 33 of 40 About 70% of people who were invited to deactivate their account did in fact do so their during the 1-week period of the study. Monday (n=825) 1.00- 0.75- 0.50- 0.25- 0.00- -10 0 10 20 -10 0 Tuesday (n=428) 10 20 Day before/after beginning of FB deactivation period Wednesday (n = 148) -10 0 10 20 - Treatment Group Control Treatment Weighted Population Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 34 of 40 Participants in the treatment group experienced a significant decrease in FB time spent during the 1-week deactivation period. Average Time Spent on FB (minutes) 8 60 40 Monday -10 10 20 -20 -10 Tuesday 10 20 Day before/after beginning of FB deactivation period Wednesday -10 0 10 20 Treatment Group 0 1 Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 35 of 40 CUT (keeping for now just in case) Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 36 of 40 CUT.... We want to understand our impact on society and be net positive on people's lives. The most feasible way to understand Facebook's causal impact is through an experiment, allowing us to understand how people feel when it is gone, how they substitute their time, as well as the core values of our product to help inspire future directions, and areas of prioritization. 1. Add title to slides 2. % beliefs breakdown 3. Evidence of weak primes plots 4. Push compliance plots to the appendix 5. Add mobile data evidence with 6. Use same format in all plots Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 37 of 40 stats to add Compliance 67% compliance rate across both waves, conditional on completing the endline survey (68% for W1, 67% for W2) Among users who think FB is bad for their well-being, the compliance rate is 74%. Among users who think FB is good for their well-being, the compliance rate is 58%. Beliefs 1267 respondents (23%) believe FB hurts their well-being a little or a lot 1430 respondents (26%) believe that FB improves their well-being a little or a lot 2637 (47%) believe that FB neither hurts nor improves their well-being 524 respondents (9%) believe that FB makes them a lot or a little less extreme 3922 respondents (71%) believe that FB makes them neither more nor less extreme 1108 respondents (20%) believe that FB makes them more extreme Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 38 of 40 Primes did influence the proportion of people who believe the study was only about WB or politics, but it was a relatively small subset regardless Percent who believe study was about impact of FB on WB only Percent who believe study was about impact of FB on Politics only Percent of respondents 0.0 0.2 04 0.6 0.8 1.0 Neutral Politics SWB Prime Percent of respondents 0.0 0.2 04 0.6 0.8 1.0 Neutral Politics SWB Prime Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 39 of 40 Survey primes to overcome the role of beliefs in assessing the impact of FB 20 We use prompts to distract respondents about the goal of the study, so that they do not inject their beliefs when responding to survey questions on a particular topic. Two thirds of respondents were shown a prompt summarizing research on FB's impact on a well- being or political polarization before starting the endline survey. Distraction Prime. When studying the impact of FB deactivation on well-being (polarization) we expect the most accurate answers to be those of respondents who were told the study is about political polarization (well-being). Amplifying Prime. When studying the impact of FB deactivation on well-being (polarization) we expect that the least accurate answers to be those of respondents who were told the study is about well-being (polarization). We find that primes do not have an impact on the effect of FB deactivation. Treatment effects are on average the same across all prime groups. Why are primes ineffective? Two possibilities: (1) People are able to judge what is good or bad for them or (2) primes are not very strong. The analysis of mobile data provides suggestive evidence consistent with the second alternative: people do not change overall time spent on mobile, but report higher amount time spent with friends following deactivation. • • Hawthorne effects pose a challenge in experimental research. In medical RCTs, researchers overcome them by using placebos, so that study participants do not know whether they are in the treatment or control group. In social science research, people know the treatment group they are part of, which lets their beliefs inform how they should respond to survey questions. Case 4:22-md-03047-YGR Document 2872-2 Filed 03/25/26 Page 40 of 40 tl;dr 9 I121ın effort to understand Facebook's impact on people's lives, we ran two 1-week consented experiments in which some people stopped using Facebook, and others continued to use Facebook. We found that those who stopped using Facebook for a week reported lower feelings of depression, anxiety, loneliness, and social comparison. We also find no effects of FB deactivation on political polarization. Finally, we find that the beliefs people have about FB's impact on them predict the observed effects we find in the data.