Youth Well-being: June 14, 2023 Onsite Research Presentation

Raw text

Transcribed text as recorded, including headers and formatting fragments.

Youth Well-being June 14, 2023 Onsite Research Presentation PLTF EX PLTF EX-01487 PLTF EX-01487 0001 Judy CONTENT (1) History of Well-being at Meta (2) Youth Well-being (3) Problematic Use (4) SSI and ED (5) Anti-bullying (6) Connections (7) What's next? (8) Post-presentation discussion PLTF EX-01487 0002 Judy History of Well-being at Meta PLTF EX-01487 0003 Judy HISTORY OF WELL-BEING AT MΕΤΑ 2015 2016 2017 2018 2019 Protect and Care This product team was the precursor to what we now know as integrity. A small group of individuals in Protect and Care focused on "care and compassion"topics which included: Suicide Prevention Eating Disorder Mental health Social comparison Prevention Legacy and Memorialization Breakup Management Harassment Authenticlidentity (Le. how to compassionately enforce our policies around using an Authentic Identity on Facebook) Private Impersonation Photo Safety Unwanted Friencing Unwanted Messaging Compassion The "care" team splits off into the Compassion product team Focus was on helping people navigate "sensitive life moments" Suicide Prevention Eating Disorder Legacy and Memorialization Breakup Management Mental health Social comparison Well-being In 2017, the Compassion team changes to focus purely on understand work and measurement. Foundational Well-being The Compassion team becomes a research-only team focused on "well-being" topics in 2018. The team is reorg'd to become a strategy team and a temporary product team for WB is spun up and then deprioritized at the end of 2019. Problematic use Loneliness Social comparison Well-being measurement Integrity Topics like suicide pervention and eating disorders are retumed back to "Protect and Care" team which turns into what we will become known as Integrity. Concerns about youth and social media already pronounced by 2015 Instagram Well-being An IG team spins up which is analogous to Facebook's Integrity team Other app well-being teams emerge in this era, mostly integrity focused Instagram Mental Health IG WB spits off a section of its beam to focus on bullying and social comparison PLTF EX-01487 0004 Judy HISTORY OF WELL-BEING AT MΕΤΑ 2019 2020 2021 2022 2023 Foundational Well-being Team sets its prioritization for understand work. Focuses on working with other teams that do well-being adjacent work Loneliness • Well-being topic prioritization for company Integrity Foundational Well-being Team is re-orged into Central social Impact Well-being understand work focusing on company externalization and prioritization of topic areas Suicide pervention, eating disorders, ancianti-bullying are monitored by integrity. Instagram Well-being IG Mental health becomes part of IG support anti-bullying, mental health Kid and teen focused teams emerge in Messenger and Instagram. Topics include child safety. Instagram Youth V-team Youth platform work including child safety andfamily center shifts into Instagram. xMeta Well-being Team is given a mandate to centralize well-being work across Meta and given product XFN Suicide Prevention/Eating Disorder Anti-bullying Problematicuse xMeta Youth All youth well-being related work is centralized into Central Social Impact Core Supporting Narrative #7 Company announces focus on Youth Well-being. PLTF EX-01487 0005 Youth Well-being PLTF EX-01487 0006 YOUTH WELL-BEING External critiques from the media and academics say "social media harms teen mental health" Namative Review Published: 01 September 2018 How Does Social Media Use Relate to Adolescents' Internalizing Symptoms? Conclusions from a Systematic Narrative Review Irene G. Sarmiento Chelseasons, Geckhlong Yes, Y. Anthony Chen, Cataline L. Toma, B. Bradford Brown, Ammy Bellmore & Marie-Louise Marss Adolescent Research Review 6, 381-404 (2000) Cte this article 1595 Accesses 7 Citations 6 Altmetric | Metrics Clinical Psychological Science On Parenting Increased use of social media takes mental health toll on teens 1/ FOX NEWS Instagram is even worse than we thought for kids. What do we do about it? The Washington Post Joumal Home Browse Joumalv Joumal Info Stay Connected Submit Paper Increases in Depressive Symptoms, Suicide-Related Outcomes, and Suicide Rates Among U.S. Adolescents After 2010 and Links to Increased New Media Screen Time Jean M. Twenge, Thomas E. Joiner Megan L. Rogers, more.. Finst Published November 14, 2017 Research Article Article Information Show off authe Social media damages teenagers' mental health, report says BBC Anya By now you are all probably intimately familiar with the narrative that's out there related to teens and mental health and social media. At its core, the critiques accuse social media of damaging teen mental health, causing teens to develop problems with depression, anxiety, body image, self-esteem, suicide, and more. Although this narrative has recently been catapulted into the spotlight, this critique is not new. Often times whenever there is a new technology, society worries about its impact, especially on youth. PLTF EX-01487 0007 Anya YOUTH WELL-BEING The impact of social media on teens is top of mind for many of Meta's critics 1) There is a common external critique and concern among many different stakeholders about the impacts of social media on mental health; also described in the Surgeon General's recent advisory. 1) This figure highlights the current emphasis of the concern around negative impacts on youth, where the perception is that teens are more negatively affected. The positive effects were emphasized for later age groups. Whose well-being do you think is most aflected when they use Instagram? 22%- 215- 40% 201 CSN rentew deck Oder adults (55-6414 Impacted: Positively Negatively Percentage 7.698 PLTF EX-01487 0008 Anya YOUTH WELL-BEING What do we know about teens, well-being, and social media? 1) Adolescence is a time of elevated risk for well-being in general, due to all the important physical, psychological, and social changes teens experience during the transition to adulthood. Some teen mental health problems have increased in certain parts of the world in the last 10 years (see data from the US and the UK), highlighting the importance of understanding what may be contributing. 1) Evidence to date suggests that the role of social media in impacting teen mental health and well-being is mixed, with many small effects. There is a need for more nuanced research and focus on specific pathways and vulnerable cohorts (review). 1) Narrative research has pointed to the importance of mitigating risks and improving positive perceptions for this key user group (CSN7). PLTF EX-01487 0009 YOUTH WELL-BEING Teen development primer 1) Adolescence is a time of rapid social, cognitive, and biological changes. a) There are differences across age segments, but not clear lines between them (e.g., not much different between 17 and 18 year old, a lot of individual variability) 2) Major goals of adolescence: a) Relationship-building (including romantic) b) Exploring your interests c) Establishing and understanding your independent identity 3) Significant cross-cultural variation in expectations of adolescence-levels of responsibility, expectations around differentiation from family (e.g., western societies emphasize more differentiation and less responsibility. 4) Key characteristics of adolescence are: a) The increasing importance/salience of peer relationships b) Biological changes that affect reward sensitivity, impulsivity, and emotion regulation c) More black and white thinking; quick and extreme changes in emotions d) Common timing for mental health challenges (e.g., depression and anxiety) to begin Elena Individual differences PLTF EX-01487 0010 YOUTH WELL-BEING Teen development Biological Tweens (10-12) Mostly prepubescent (puberty M girls 11, boys 12) Socioemotional Straddle both child and teen worlds (eg, teen aspirational but some child interests) Same-sex friendships Early Teens (13-15) Pubescence, rapid biological changes Brain changes mean increased sensitivity to rewards Exploration of identity Emerging Interest in romantic/sexual relationships Heavy influence of peers Late Teens (16-17) Slowing physical growth (esp. giris) Continued sensitivity to rewards Stabilizing of identity Increased individualization More intimate, egalitarian relationships Cognitive Black-white, concrete thinking Emerging abstract/hypothetical thought Future thinking/planning Parent/Child Relationship More accepting of parental oversight and rules Increased desire for independence (with more family conflict) Parent more as adviser (reduced family conflict) More independence Kremer (2021) Chart compiled from various 3rd party child development resources such as CDC, WHO and textbooks (eg, Grusec & Lytton; Siegler). Exact ages of bands may vary slightly by source. Full chart Ellos.docs.google.com/document/d/toeuO8uLsic6P3aYxBsaMkKddLHLVE Sim AkJdJF974/ecal See more detail. Neurobiological development re: reward bere Elena Individual differences PLTF EX-01487_0011 YOUTH WELL-BEING What is well-being, mental health, and how did we prioritize what to focus on? 1) There is no one definition for "mental health" or "well-being." Academics have 2150 definitions of these concepts. a) They range from clinical framings (e.g., depression) to feelings (e.g., happiness), and other life areas like economic security. 1) Our approach to understand priority topical areas in this space involved an intensive six-month understand & prioritization processes in 2020-2021, which included a synthesis of inputs from academic research, company leadership, policy insiders, and the general population about what are the most important topics for us to focus on. a) One of the priority areas was around understanding the emotional impacts of our platforms, which is a research focus to identify relevant product levers. 1) In Q2 2023, we conducted a refresh of this prioritization process to include more recent data inputs. As part of this effort, we identified people problems first within each priority topical areas, then organized by themes that extended across topics (e.g., experiencing crisis, lack of control, content exposure). Finally we prioritized the people problems against criteria: reach, severity, and alignment with industry practices. Elena Note- we provide additional context on this prioritization refresh process later in the presentation Additional context on 2020 prioritization process Defined well-being: Exploring a wide set of potential definitions and measures of well-being, we aligned with leadership on a definition for well-being: Facebook's impact on individuals' lives; well-being is a set of topics that cover the individual level impacts Facebook has socially, emotionally, informationally, and beyond. We ultimately selected this definition as it allows us to consider the widest set of critiques and praises about Facebook. Conceptualized well-being: We then systematically reviewed all the impacts we could find associated with Facebook, Instagram, and WhatsApp. We reviewed 400+ academic articles, 900+ media articles, 150+ internal research articles, which resulted in the identification of more than 1000 potential well-being topics. Prioritized Well-being Topics: We presented the 28 most commonly referenced topics via a MaxDiff exercise to 91 leaders at Facebook, 402 policy insiders, and 12,002 people from the general population in the US, UK, and Australia and they ranked the importance of Facebook focusing on each topic (see here for more details on the MaxDiff survey). PLTF EX-01487_0012 YOUTH WELL-BEING Investment to date (teen-specific; significant platform variation) [WIP perspective] Topic Problematic Use Foundational understanding of the topic and what we should do High Product investment Mid-Low SSI ED Bullying & Harassment Connections Elena -brief - snapshot of eating disorders Low-Mid (emerging) Low. The current research we have is based on Appearance Comparison. Mid-High on IG; More work needed on teen-specific B&H Understand High, distributed Low-Mid (emerging) Low High, IG-focused High; distributed PLTF EX-01487_0013 Elena YOUTH WELL-BEING Well-being is currently one of six youth focus areas Meta identified for supporting Youth. There are five topical areas within it. Age Assurance Safety Youth Focus Areas for Meta Well-being Data Use Age Appropriate Experiences Guardian Supervision 1. Problematic Use 2. SS 3. Eating Disorders 4. Bulying & Harassment 1. Connections This is our team's focuel PLTF EX-01487_0014 Questions so far? PLTF EX-01487_0015 Problematic Use PLTF EX-01487 0016 PROBLEMATIC USE Problematic Use Definition 1. 1. A perceived lack of control over usage of a technology or social media platform and This lack of control is perceived to lead to negative life impacts (e.g., in domains such as relationships, work/school, sleep, etc) Problematic use Focus is here Perceived lack of control over social media use perosived to contribute to Negative life impacts Control over time spent Control over experiences Low control experiences may or may not be perceived in negative Negative Ife impacts must be related to low control to be "problematic use" See 1-pager here Life impacts are more distal from the actual on-platform experience, and the role social media plays in a user's life overall is person- and context-dependent; thus it is outside the scope of what products can directly focus on. PLTF EX-01487_0017 PROBLEMATIC USE Perceived life interference from app use is highest for younger users, and tracks with age Percentage endorsing life interference "very often" Main Tracking Survey, Facebook Feb-March 2023 14 12 E 13-07 18-34 25-36 35-44 45-54 55-64 654 Age bucket Main Tracking Survey and specific question: Does your use of Meta's apps and products prevent you from fulfilling your responsibilities in other areas of your life (for example, work, school, or relationships)? 15 PLTF EX-01487_0018 PROBLEMATIC USE Problematic use primer 1) Our strategic approach to addressing this space is focusing on understanding and addressing areas of the platform or patterns of usage that are associated with perceived low control. a) Success is: "People have the tools that they need and are empowered to control their time and experiences on our platforms." 1) We measure perceived low control in two ways: a) Perceived low attentiveness around platform use (i.e., spending more time than wanted, checking, and consuming content without realizing it) b) Perceived low capability to change platform use (ie., low confidence in ability to reduce time, checking, or interactions) 1) Youth (<18) and adults (18+) report perceived low control (of both types) at generally similar rates; overall rates are highest for youth and young adults (international survey data from 30K youth and 30K adults on each of FB and IG) 1) The behavioral signals (e.g., late night use) and product usage patterns (e.g., video chaining) associated with perceived low control can be used to prioritize when and where to offer control features, or to change underlying platform mechanics. a) It is nuanced how to use behavioral signals for goaling - how to mitigate potential low control behaviors when platform use, especially in a given context for a given user, might also be bringing value? Both perceived low control and negative life impacts are associated with app deletions PLTF EX-01487_0019 PROBLEMATIC USE How can we increase use control, through products? Control Levers: Awareness Attentiveness Leam about and reflect on own behawior Notice behavior in the moment What this might look like Products that give people feedback about their own behavior and experiences on platform. Products that give information in context-tied to specific behavior. How it works Information enables reflection about behavior and informed choices about desired changes. Intentionality & Choice Make explicit decisions about platform use Products that enable behavior change or make change easier. Interrupts and draws attention to automatic behavior, slowing people down, and enabling more deliberate choices. Helps people set goals, create limits, set explicit preferences. Examples Your Time on Facebook/Instagram Feed Pause (FB) Take a Break (IG) Personalized daily time limit PLTF EX-01487 0020 PROBLEMATIC USE Product investments for Youth Industry Practices (example from Facebook) Youth Industry Practice for Well-being (YWBIP) PU-1. Provide tools that control teens total time spent on our platforms. PU-2. Surface insights on usage patterns and durations to teens. PU-3. Offer age-appropriate information to teens on how to manage their time online. Product Translation Translation of how this Industry Practice applies to Facebook's products Provide tools that enable teens to set a daily Rait for total tome spent on Facebook and receive a reminder when they have reached that t Surface insights to teens on dally time spent on Facebook and how long time was spent. Provide teens with information about how to use the tools to manage their screen time, within time spent tools Now. PU-4. Provide options to mute notifications specifi to nighttime Provide settings that mute notifications with a night bime specific range of time. PU-5. Interrupt long periods of watch time with reminders to take a break and stop scrolling, defaulted to on, and/or move to other topics. PU-6. Sendi notifications to beers before their bedtime to remind them to wind down Provide teens with a nodige to take a break, triggered by long usage Provide teens with a reminders to wind down before bedtone, Diggered by late night usage YIP deck PLTF EX-01487 0021 PROBLEMATIC USE By focusing upstream, we can enable shifts in user behavior that can lead to downstream impacts for some people Theory of change 1) Increase perceived control via Overall awareness In-the-moment attentiveness Intentionality and choice 2) More explicit choices People make informed decisions People are more able to change their behavior 3) Negative impact is mitigated The perceived role of social media in contributing to negative life impacts is reduced Puone-page PLTF EX-01487 0022 PROBLEMATIC USE "It was really about, let's figure out how to help him take control of his Facebook, his Instagram, what he does, who he talks to, why he's really on it. So he can get the most out of it, the good stuff - all that support. And not have it feel like it's a thing that's happening to him, or in control of him. We tried to put him back in charge." Qualitative study with mental health clinicians, 2021 I'll leave you with a quote from a psychiatrist in the US, talking about how he treated a young adult male patient with anxiety that was exacerbated by Instagram and TikTok use, and that really exemplifies the findings we're presenting here. He said: [read quote] PLTF EX-01487 0023 SSI & ED PLTF EX-01487_0024 SSI & ED SSI Primer Suicide describes "taking your own life". Self-injury (also referred to as self-harm) describes "the act of deliberately harming your own body, such as cutting or burning yourself". Attempts at suicide, thoughts about suicide, and self-injury are often considered symptoms of larger mental health issues, but can also occur in the absence of known mental health issues, during periods of acute emotional distress. Little robust evidence of global increases in SSI rates, and a dearth of any causal evidence of social media's role in trends.2 Evidence suggests that online bullying and content exposure are mechanisms by which social media in general, including Meta apps, may play a role in SSI, highlighting the interdependence of these two topical areas and the importance of continuing to make progress on both in parallel. SSI content-and user behavior associated with that content-is potentially problematic because it: 0 May signals that a user is, or someone they know, is in distress 0 Displays unhealthy behavior that we do not in any way want to unintentionally normalize, promote, or suggest PLTF EX-01487 0025 SSI & ED SSI Content Definitions Content Policy-Violating Content Borderline Content Definition Any content that encourages, promotes, coordinates or provides instructions for suicide or self-injury, including fictional content such as memes or illustrations and any self-injury content which is graphic, regardless of context. Content that identifies and negatively targets victims or survivors of suicide or self-injury seriously, humorously or rhetorically, as well as real time depictions of suicide or self-injury. Content about recovery of suicide or self-harm that is allowed, but may contain imagery that could be upsetting, such as a healed scar, is placed behind a sensitivity screen. SSI-related content that doesn't rise to the level described above, including: Written/verbal admissions of self-injury Vague, potentially suicidal statements or references (including memes or stock imagery about sad mood or depression) in a suicide or self-injury context. PLTF EX-01487 0026 SSI & ED Approach to SSI Across Meta 1. SSI Resources We currently surface resources across FOA in the following use-cases: Self-helpc 1. 2. 3. 4. User search for or SSI related content (6000+ search terms) User's post or profile was reported for SSI but content is not policy violating We've taken down user's post (policy violating) or When a user expresses suicide Intent on Live Help a friend: Someone who reported SSI content or profile 2. Adding Friction Part of the Remove, Reduce, inform strategy to manage problematic content Product examples: Currently, we are adding friction in the form of search interventions, sensitivity screens, MAS Imark as sensitive), demoting and removing (policy violating) content. 3. Reporting & Moderation SSI is a category users can select when they report content or an actor. The reporting flow is owned by Integrity and is mostly consistent across products/ surfaces. Reporting SSI will also surface resources on how to help a friend. PLTF EX-01487 0027 SSI & ED Approach to SSI Across Meta User Signal Surface Resources Add Friction Remove Content Reduce Content Age Gating Reach Post SSI content X X X X X Report SSI X X X X X content Search for SSI X X content View SSI content X X X X X PLTF EX-01487 0028 SSI & ED XMWB SSI Mission We proactively detect and appropriately address problematic SSI content, and provide access to supportive SSI-related tools and resources when people signal they may be relevant. PLTF EX-01487 0029 SSI & ED H1 Deep Dive: Resources/Tools/Contexts [WIP Key User Needs] Workstream Goal Product Audit Review and compare SSI resources against key competitors and product principles. Funnel Analysis Understand current reach of SSI resources relative to appropriate contexts. Expert Engagement Literature Review Ecosystem Analysis (interim) Solicit feedback from external frontline experts on potential context gaps and recommended changes to SSI resources. Understand what we already know about SSI resources. Identify new opportunities to address teen exposure to SSI content and/or to surface resources/tools. "All workstreams prioritize understanding teens specifically, while being inclusive of gen pop PLTF EX-01487 0030 SSI & ED What did we learn about SSI & Teens? Learning Significant gaps, complexity, and inconsistency in usability, access and content of the resources we offer to users. There are significant differences between resources flow and logging on FB and IG, and key gaps in search inform keyword list. Teens see more SSI content than adults. There are emerging behavioral signals that suggest new opportunities to provide resources/tools (emerging theme: "intentionality") There are additional levers we could pull to address problematic SSI content exposure (e.g.. content dilution) Implication Address key usability gaps, drive consistency across platforms, and remove barriers to access Ensure resource content resonates with teens Fix logging issues so we can learn about who accesses resources and when Improve/automate our search term process Determine drivers and implement solutions to drive exposure down Product Work Underway? In progress (MVP) Not roadmapped In progress Roadmapped for H2 Determine and ship best resources/tools to surface to teens when they send those signals (e.g. saving, repeatedly viewing content). Not roadmapped Explore feasibility and ship experiments that try to prevent individual users from seeing >threshold of SSI content Not roadmapped PLTF EX-01487_0031 SSI & ED Constraints/Limitations Do Don't Use on-platform signals (content, behavior, content x behavior) as signals that resources/tools may be relevant Make inferences or assumptions about people's emotional state Proactively offer resources/tools when we think they Over-pathologize behavior around content (posting may be relevant Lean on internal data and external researchers and clinicians to provide insight into the user journey, key places and ways to show up for users. or consuming) exception: Credible Intent of Suicide Do typical UXR with people who have lived experience with SSI (lots of sensitivities) PLTF EX-01487 0032 Bullying & Harassment PLTF EX-01487 0033 BULLYING & HARASSMENT "Sometimes the comments are really devastating. And even sometimes they'll be like, "What do you think you are doing?" They'll just be making bad comments, making me feel bad sometimes, and it hurts." 16 year old "There was one time where my whole school-it was a troll thing where it's like you report and you hate on other people's accounts that you don't like, just people you don't like, and you just do that. And it would get toxic. And one of my accounts got banned because these group of girls didn't like me." 13 year old Source PLTF EX-01487 0034 BULLYING & HARASSMENT Bullying and Harassment Primer Definition Internally, (on Meta platforms), we define bullying and harassment (B&H) as any unwelcomed interaction that causes emotional or reputation harm to someone, a behavior that takes many different forms from making threats and releasing personally identifiable information to sending threatening messages and making unwanted malicious contact. Types of B&H: 1. 2. One-to-one Harassment Many-to-one (m21h) Harassment 3. Many to Many Harassment Measurement Teens experience much more bullying on platform than adults. Despite critiques, studies from the CDC show that cyberbullying has not increased in the last 20 years. It is also not any more prevalent than it is in-person. Still, we should support teens who are being bullied. Various metrics being used to measure B&H Post-Interaction Behaviors M21H prevalence Holistic Prevalence Negative user interactions Violence and interment Violence Tororom- Sexually explicit language Sexual Sodictation and Prostitution Regulated Goods Privacy Violatio Platform Policy Nudity and Pornography NC8T NC Human Exploitation Health Integrity Hate Orgs Hate Fraud Election Integrity Drugs Dangerous angs Cybersecurity Credible Threats Credible Iriant of Bulide Coordineeing Ha Chat Subject Objectionable Content Child Sexual Exploitation Bullying Age Adult Town 工 Individual 10 PLTF EX-01487 0035 BULLYING & HARASSMENT Bullying and Harassment Primer Anti-bullying strategies from Instagram Support and Well-being teams Least Disruptive to People Tst Line of Darlense Encourage positive interactions 2nd Une of Delemaa Reduce reach of harm Most Disruptive to People 3nd Line of Dealers Support users w/ safety features Identify (measurement and signals) Examples of anti-bullying features Default privacy settings for teens Block and report Comment warnings and nudging to increase empathy Hidden words See all BSH tools available at Meta PLTF EX-01487 0036 BULLYING & HARASSMENT "Some of the comments that I'll get, if I don't really like it, I'll take a keyword out of that comment and put into the filter thingy and then just hoping that it'll filter enough so don't get those comments" 14 year old "And then if there was someone that followed me that didn't like a video of mine, I will remove them from my follow list." 17 year old STOP BULLYING * Source PLTF EX-01487_0037 BULLYING & HARASSMENT Considerations and next steps Teen B&H is highly contextual. Content often isn't mean on its own (e.g., sarcasm) but is perceived as such based on their individual situation. This means that our usual automated interventions (classifiers, downrank, filter, etc.) are not effective and result in false positives and over-enforcement. Teens B&H can often happen in atypical formats. Consider the example of Tea Accounts (internal discussion) or Exclusionary Bullying (eg. people will tag someone in a picture of a party they weren't invited to). Teen B&H is from people teens know in real life. This presents a unique challenge in that B&H can be quite fluid across on and offline scenarios. This is different from the creators who often face harassment mostly from people they don't know and largely online. Our safety tools (block, restrict, limit, mute) may not be effective. These are built to protect people against B&H from people they don't know so may not be effective for teens. These are indexed on solving bulk harassment so may not work for 1:1 or small group bullying that teens face. Social dynamics may mean that teens don't use these e.g. a teen is unlikely to block their classmate who's being mean to them xMeta Well-being had to deprioritize owning active product development in B&H in 2023 H1. Teams working on anti-bullying (Instagram Support, xMeta Integrity, Oculus) have been doing so for years and we should focus on where we can provide a value-add to these efforts. xMeta Well-being should provide guidance, specifically for ensuring YIPs are met and measurement approaches are consolidated. PLTF EX-01487 0038 Teen ecosystem work from SI-X PLTF EX-01487 0039 Ecosystems Teens have worse integrity experiences than adults Integrity Harm Exposure on FB App Aggregate Individual UWOP SWOR 100 150 200 250 300 Violence and incament Vol Tor 881- Sexually explot language- Sexual Servicikation and Prost Regulated Priwary Platform Nudity and Porno 14101 Human Exploitation Health Integrity Hale Orga Fraud Frees Election Integrity Drags Dangerous org Cybersecurity- Credible Threats Credible Intent of Suicide Coordinating Harm Chid Subject Objectionable Content Child Semua Exploitation Bullying 101 Age Adult Tean PLTF EX-01487 0040 Ecosystems Feed Relevance has consistently exacerbated this SWOP A 9457: H2 2022 Feed Pillar Backtest 35911/35000 Order Dy severly weighted 40-44 45-4 30-04 60-64 073-3-48 30 75-79 6156075 80+ ONAL 047+852 9469: H1 2023 Feed Pillar Backtest 30047/35945-experiment and query information 90% CI Onder Dy 25-29 Type to filter by test string 07248.20 048035 8548:10 141014 038-017 843-018 8:276031 -0204050 PLTF EX-01487_0041 Ecosystems One explanation is that Teen ranking is especially driven by engagement terms Engagoment 180 128 Difference in Teenenportance post position explained by tem 18 %post position explained by buckall variablo Conearmption 80 25 Difference is Teenbucket mportance 50 PLTF EX-01487 0042 Connections PLTF EX-01487 0043 CONNECTIONS Connections Primer CAROL D. RYFF AND BURTON H. SINGER actualization (Maslow) Maturity (Allport) executive processes of personality (Neugarten) Set acceptance purpose in The autonomy environmental mastery personal growth positive relationships Individuation Jung) mental healt Jahedal booic Ste tendencies [Bünler) will to meaning Frankl "[A person with positive relationships] has warm, satisfying, trusting relationships with others; is concerned about the welfare of other others; capable of strong empathy, affection, and intimacy; understands give and take of human relationships" PLTF EX-01487 0044 CONNECTIONS 13-17 year olds place a higher value on connecting with friends, family, and their communities than all other age groups. 71% of internal studies mentioned how Facebook positively impacted the relationships they have with others. External studies find that connecting with family, and with friends and acquaintances are two major reasons for using Facebook (35% and 14%, respectively). 90% of clinicians identify that "interpersonal relationships" is often/always playing a role in mental health. With respect to social media, in the US, 15% of clinicians agree/strongly agree that "social media encourages meaningful connections with others" compared to 41% of clinicians in the UK. M W ... designed by freepik PLTF EX-01487 0045 CONNECTIONS How we measure connections at Meta Reciprocal Ties (RT) are a metric created by the Computational Social Science (CSS) team in CSI. It was created to measure Meta's mission "Giving people the power to build community and bring the world closer together". An RT exists between two people if they have interacted in the last 7 days (commenting, messaging, tagging, reacting, etc). RTs are associated with positive well-being measures like higher social support and lower loneliness, as well as higher interpersonal empathy & trust. 2019 2020 2021 2022 2023 Late 2018-Early 2011 Project kickoff-Mark request to measure Meta's progress on the company mission Project moves to Core Data Science. Exploration of mission concepts Community Voice, Closer Together 2020-2021: DSS joins. Prioritize "bringing the world closer together"; define as "Trust, Empathy, Social Capital", Research begins to identify metrics that are behavioral correlates of these concepts. H22021-01 2022: Reciprocal Ties, is found to be robustly associated with the value we expect from connections to Friends & Family-truste empettes social support, lower Janelinees, Andraiser succeer.on EB. and helpful TOPADHANAtions on FB 02-12-2022: RTs available on F and Messenger. AMD metric is created along with a dashboard 1112023: Longitudinal studies and meta-analyses provide supporting evidence that RTs could be used as a metric by product teams. Begin experiment with RTs as ranking inputs to PYMIC PLTF EX-01487 0046 CONNECTIONS Facebook products are facilitating reciprocal ties between teens 2 Ederred age Sep 19 Jan 20 May 20 Sep 20 Jan 21 May 21 Teens (13-17) Sep 21 Young Adults (18-29) Jan 22 May 22 Sep 22 Jan 23 May 23 Adults (30+) Comment bumping 4%- 3% 2% 1% 0% Feed imegrity Comment bumping 0.5% of posts in Feed 0.4% 0.3% increases RTs by up 0.2% to 4% overall (2% 0.1% impact on teens) 0.0% Teens have the most Reciprocal Ties on Facebook compared to other age groups! Having Integrity- related products increases RTs by -0.3% for teens PLTF EX-01487_0047 CONNECTIONS Limitations of current measurement Reciprocal Ties metric currently available on Facebook, Messenger, and Instagram (soon!). It requires extensive validation work to stand up this metric across our platforms. RTs measure interactions between "connected" (friends, followers) individuals. Messaging data is one of the critical components of calculating RTs but EPD and E2E makes it very difficult to use data on platforms like WhatsApp to calculate RTs. It may never be possible. Ideally we'd like to look at inter-platform measures of connection but linking accounts is not simple. Measuring an analogous metric to RT in Metaverse will be labor intense. Future work includes looking at community connections (unconnected individuals) and connections to creators. PLTF EX-01487 0048 What's next? PLTF EX-01487 0049 WHAT'S NEXT Leveraging prioritized issues and clusters from April's strategy sprint Prioritization exercise: Highest priority issues Prevalanse Hy Hy Loe Observations & problem clusters Problematic content exposure: Youth users are frequently viewing problematic content. "Problematic content can be SSI, ED or (witnessing) B&H. Exposure can be either "pushed to users (eg. in-feed recommendations) or "pulled" by user engagement (eg, search). PU [time]: Youth users perceive a lack of control over how much time & what times of day they use our products. SSI crisis support & resources: Youth users are not consistently ① accessing crisis support resources when they are experiencing an acute mental health crisis (while on one of our platforms) Deck PLTF EX-01487 0050 WHAT'S NEXT Methodology: Process for Prioritization The subtopics were derived from the following sources: 1. Youth Industry Practices in 2023 2. Literature reviews 3. SME across functions review For each of the following subtopics a lead SME went through past literature reviews and seminal studies, as well as the youth industry practices, and listed out the people problems/sub problems/areas/issues (aka "subtopics") that had been identified. Once they were all entered, the researcher then went through and clustered them, and identified top-level themes. Additional SMEs across research, policy, product, design, and engineering reviewed the top list of subtopics to make sure no major themes were missing, and the wording was directionally appropriate. With the top themes, the lead SMEs then wrote up descriptions of these suptopics based on the literature review information. The main themes were extracted, and then brought onto the next phase of evaluating in terms of prevalance, severity, etc. For the full list of subtopics identified, and sources, see here. Deck PLTF EX-01487 0051 WHAT'S NEXT Methodology: Process for Prioritization Coding Criteria YIP To evaluate the criteria for this, two subject matter experts read through all of the youth industry practices, and aligned on the following criteriac Yes: This subtopic is related to an already defined YIP. Noc: This subtopic is not related to an already defined YIP. Reach To evaluate the criteria for reach, we pulled as many surveys or log data reports we could find that were demographically or nationally representative of teen or teen girl populations in the United States (e.g., Pew, Common Sense Media) or large-scale global surveys intermally (eg., Hard Life Moments). The data we relied on may contain a directional skew towards US, girls, and Instagram users/experiences. High: Data Input(s) directionally suggest that at least roughly a quarter of teens experience this subtopic monthly on a Meta platform. Medium: Data input(s) directionally suggest between 10-25% of teens experience this subtopic monthly on a Meta platform. Low: Data input(s) directionally suggest less than a 10% of teens experience this subtopic monthly on a Meta platform. If there were multiple data inputs and one suggested high, and the other suggested low, it was scored as medium (unless there was log data available). There were 6 subtopios we could not locate any data. In such instances, three subject matter experts aligned on a theoretical rationale for the coding in contrast to the other categories where we did have data. The goal here was not to provide absolute prevalence, but to try and give an understanding of the relative prevalence among the subtopics. For a deep dive into the data inputs, see here. Intensity To evaluate the criteria for severity, three subject matter experts aligned evaluated each subtopic through the following lens: High: This subtopic is explicitly related to a mental health condition Medium: This subtopic is not related to an explicit mental health condition, but it is either a subtopic that is inherently linked with something involving a frequency/higher volume (eg, a lot of content), tied with a direct interaction (e.g., bullying), or tied with a critical life activity (e.g., sleep). Low: All other subtopics. PLTF EX-01487 0052 WHAT'S NEXT Leveraging or repositioning product levers to apply across youth priorities Problematic use Product lever (P1 anly) Crisis (SSI) Product lever Potentially problematic content (specific topics) Product lever 1a. In-contest nudges 1h. Centralized Controls 1. Easier accessed resources 1. Central settings & controls 2. Education 3. Guardrails around mechanics 2. In-context interventions A. "Show less" B. Alternative topic nudge 2. Improved resources and corrent 3. Ranking & recommendations A Global youth/policy thresholds for SSI,ED & B&H B. Experimental incremental thresholds 4. Measurement infra (targeting/trigger logic) Deck PLTF EX-01487 0053 See example WHAT'S NEXT Emerging/potential levers for underlying platform mechanics across topics Content diversification: changing ranking and recommendations systems so that youth see a broader slate of content with overall more variability Relevant for preventing platform-driven over-exposure of any specific topic; injecting variability within content consumption patterns Relevant for promoting more breadth of information exposure and enabling exploration (important for teen development) Preventing over-indexing recommendations based on past behavior, which may be based on temporary emotions or interests; give teens more ability to reset "Can we reduce the signal to noise ratio here? If that's [SSI-related content] all people are seeing, it becomes really salient. You could try using the dilution effect here using something neutral/positive." - Dr. Craig Bryan Guardrails: ensuring that experiments that impact engagement are not only driving the high usage end of the spectrum (for time, content exposure, etc.) Relevant for shifting experiments to evaluate user impact and potential user value across a range of usage, not only impacting relatively 'excessive' behaviors More youth-specific and well-being-related content monitoring for whether engagement experiments are inadvertently impacting content exposure PLTF EX-01487 0054 Elena WHAT'S NEXT What do we mean by "education"? Digital education is about acquiring knowledge, skills, and dispositions to navigate the online world safely and healthily. Knowledge What someone knows or believes In our products, we need to support teens moving from awareness to knowledge to skill development to mindful reflection. Skills This approach to product education should make teens feel empowered and confident, not afraid or defensive. Dispositions Things someone knows how and is equipped to do Tendencies that guide thinking and behavior Aclapted from Common Sense Education and Project Zero's framework PLTF EX-01487 0055 We heard yesterday about many personal experiences, connection to the work - this can fuel us and bring a lot of passion and motivation to the space. This is like the fire behind the rocket ship. Let's work to steer the rocket ship in the direction of what our learning and the available evidence points us to. This is a sensitive space and can affect us personally at times - something to be mindful of. Reminder to take care of yourself PLTF EX-01487 0056 Group Discussion PLTF EX-01487_0057 See example DISCUSSION What's something surprising you learned today? PLTF EX-01487 0058 See example DISCUSSION What more do you want to learn about? PLTF EX-01487 0059