████: "Currently on FB, when a user attempts to change their age to under 13, a checkbox to confirm the age is shown on www (not on mobile), an error is generated, and the action is logged in the table 'edit_birthday' (userid, new age, old age, and the error message are stored for 35 days). No enforcement action is taken. Checkpointing these users could result in ~126k/year in the US (~3.4M/year globally) MAP impact."
Andrew Bocking: "Why are we planning to limit to US in this proposal?"
████: "The highest legal risk is in the US given COPPA and FTC. The idea was to strike middle ground between mitigating legal risk asap and containing MAP impact."
████: "Checkbox launched globally to 95% on 11/26 – observed 30% drop in u13 DoB change attempts, which translates into 2.3-3M global annual MAP impact (vs ~3.2-4M earlier)... Working team's consensus is to: Launch checkpoint globally this week, pending remaining product/eng work and ops readiness (the capacity now is 2.6k/day, will need to go to 12-16k/day + likely 35 day backlog of 591K accounts)"
████: "Wanted to give an update on our recent age fallback tests and provide Acquisition's recommendation on path forward.
TL;DR: Acquisition wants to launch our existing Age Management solution: defaulting to today's date on the registration birthday step & age fallback together for a net impact of -12.5M MAP/year across all interfaces on January 15th. We have now tested falling back to age buckets and a binary over/under 13 option. Neither experience was better than the age fallback experience already built in terms of retention.
Context:
As previously discussed, Acquisition needs to change the way we currently ask for age in registration. Currently, we have a default date set that is at least 25 years before today's date. There are policy concerns that this pushes people under 13 to enter an incorrect birthday. For the last few months we've been testing solutions, and landed on one which includes two major changes:
1. Defaulting to today's date on the registration birthday step. This will be an estimated -174k WAP@14 hit (18.4M MAP/year).
2. The age fallback product across all interfaces. This will be an estimated +54k WAP@14 win (5.9M MAP/year) on top of the change to default to today's date.
We net out at a -12.5M MAP/year hit with these two changes. This is -4.5M India MAP/year, -1.5M Indonesia MAP year and -210k US MAP/year. 65% of the impact comes from FB4A, 12% from Lite, 11% from mTouch and 10% from Kite.
New test results:
Per feedback from Brian and Alex, over the last month we've tested a new solution - falling back to age buckets or over/under 13 instead of falling back to an age entry screen. These tests were not better than our existing age fallback experience in terms of retention. Both the age buckets & o/u 13 get significantly more people through the registration than age fallback, however the same or more people select an under 13 age and are therefore not able to successfully register. So we net out neutral to negative on/conf. The main test groups are below and you can find a more detailed description of the tests and results can here."
"Recommendation:
The Acquisition team proposes to launch the two above changes to the birthday step in registration across all platforms on 1/15. We plan to continue to do some iterations on the birthday step in Q1, like error messaging & UI changes, but do not have any other wins that can be achieved quickly that will recover a significant amount of the impact and thus believe we should ship and iterate."
████: "I'm a bit confused as to why they want the over/under 13 buttons when they say it didn't work better than the existing age fallback experience (which is what again?) -- maybe I'm reading this wrong but in the very next sentence they say 'both the age buckets and o/u 13 get significantly more people through the registration then age fallback.' Is the distinction here that the tests didn't impact RETENTION but it got more NEW users? Which is also confusing b/c why would they be going through an age flow if they're old users we're retaining? Or perhaps retention means whether or not they complete the flow? Full paragraph I'm referencing is here and highlighted in blue below:
Per feedback from Brian and Alex, over the last month we've tested a new solution - falling back to age buckets or over/under 13 instead of falling back to an age entry screen. These tests were not better than our existing age fallback experience in terms of retention. Both the age buckets & o/u 13 get significantly more people through the registration than age fallback, however the same or more people select an under 13 age and are therefore not able to successfully register. So we net out neutral to negative on/conf. The main test groups are below and you can find a more detailed description of the tests and results can here."
"Unlike communication with close friends and family, virality is something new we have introduced to many ecosystems (so new that many spell-checkers don't recognize the word) and it occurs because we intentionally encourage it for business reasons."
"Unfortunately, research has shown how outrage and misinformation are more likely to be viral, and recent experiments that deprecate these models indicate that removing these models does positively impact metrics for misinformation and hate. Not all content that benefits from our choices here is bad, but as long as we continue to optimize for overall engagement and not solely what we believe individual users will value, we have an obligation to consider what the effect of optimizing for business outcomes has on the societies we engage in."
"We prioritize content based on engagement, but engagement doesn't necessarily mean that a user actually wants to see more of something. One of our biggest signals we use to provide more similar content is comments and as this Buzzfeed article points out (image from article below), a comment that you hate a thing can be seen as a positive signal leading content to get outsized distribution."
"People game this in various ways, posting ever more outrageous things to get comments and reactions that our algorithms interpret as signs we should allow things to go viral. Publishers and campaigns tell us that they are incentivized to do this and we can see in our data that the above tactic of inciting users to negatively comment works to get more distribution (analysis by Delia Mocanu in graph below)."
"At a minimum, we should stop promoting content based on ambiguous signals, as it not only promotes bad experiences (in some cases), but also makes it easier for bad actors to game our algorithms by creating content (e.g. clickbait, sensationalism, outrage-bait) that our algorithms think are wildly popular, but are bad user experiences."
"Indeed, even Mark himself has suggested that the anger reaction is a reasonable way to express that you don't like a piece of content (link here), even as we currently count it as 4x as important as a less ambiguous 'like' for giving you more such content."
"Our Well-being team has been focused this half on 'defensive' strategies; ways to mitigate the negative effects of Facebook on people's well-being. Problematic Use* is more prevalent than previously thought (12.5 % of people affected). Another 11% of people report negative life impact 'very often' but only 'sometimes' lacking control."
"Sharing or viewing filtered selfies in Stories made people feel worse"
"In Stories, sharing or viewing more filtered selfies (i.e., that had a color filter or a camera AR effect) was associated with increased negative comparison. The effects of filters are weaker for Feed, likely because Feed camera filters are less selfie-centric."
"More generally, sharing selfies (both filtered or not; and to Feed or Stories) is also associated with increased negative comparison, but this effect is weaker."
"Method: We measured how social comparison differed based on proportion of either viewed or shared photos that were taken using the front-facing camera and/or had a filter applied in a regression controlling for gender, teen/non-teen, follower count, following count, whether account is private, and country."
"Seeing these camera effects can lead to more negative comparison: Why? These may focus on beautification; heart eyes may be interpreted as showing off a relationship."
"Seeing these camera effects can lead to more positive comparison: Why? These tend to be more 'fun', focusing on additional objects or taking part in a challenge."
[Speaker notes]:
"Some possible reasons:
More negative;
Heart eyes may indicate being in love, and we know from qual research that seeing others in relationships makes people envious
The other two also seem to be more of the skin-smoothing/beautifying-type
More positive:
These filters are more 'fun' - adding a halo, sunglasses, or doing a challenge"
"Stated Age - age collected at registration is primary signal for enforcing age policies & special protections... IG does not have stated age for ~35% of users."
"Limited Sharing & Interactions w/ Strangers:... Unconnected adults can't initiate messages to u18s on Msgr... IG Direct does not currently include this restriction."
"But market dynamics are affecting Instagram use - saturation with users 35 years and younger will prevent significant MAU growth in the US."
"Given that we are close to saturation with <35, and do not have product market fit with adults, we shouldn't expect significant MAU growth in the US beyond the ~4M teens that start using the internet each year."
"The US DAU investigation will focus in part on growing the teen DAU."
"Recent data science findings suggest a DAU opportunity with teens."
"H2 research efforts will address this by sizing competition for low I-ness teens and young adults."
"TL;DR Identity, Main Account/Secondary Account
• Many parents joined Instagram to learn about it so they could help and follow their teen.
• They believed that knowing what their teen did on Instagram helped them protect their teen.
• Most did not become aware of teen finstas/spam accounts, until long after the teen had created it.
• They learned about their teen's second account through conversations with their kids or other kids.
• Many were worried, curious, and conflicted about their teen's finsta/spam account and wanted to spy.
• Most teens learned about these accounts from older siblings or friends, after getting acclimated to IG.
• Teens did not mind when parents discovered the account's existence. Parents did not stop the teen from using it.
• Teens said that their parents and often their younger siblings would not understand what they shared in their finsta."
"Parents had a strong need to protect their kids and believed that seeing what their child did on Instagram enabled this. When they discovered there were parts of their teens' Instagram they couldn't see, they felt powerless. Limited visibility is a strong pain point for parents."
"They worried about what their child might be hiding and feared the worst, like cussing, vaping, breaking rules. Some knew other parents who 'thought their kids were angels' but didn't know the real/whole story."
Jennifer Guadagno: "Social comparison is important to loneliness: Social comparison is the most important driver of loneliness. Given our large Inc. wide Loneliness focus that was just approved by Mark, social comparison becomes an important area to focus on as it relates to loneliness. See Chart 1 in data doc.
People perceive FB and IG as having negative impact on social comparison: 21% of people report that Facebook has a negative impact on social comparison (3rd worst area that we measured). 19% of people report that Instagram has a negative impact on social comparison (6th worst area that we measured). See Chart 3 & 4 in data doc.
Best way for FB to significantly contribute to reducing social comparison is through Like counts: From Moira et al's CHI paper seeing high like counts was one of the (changeable) factors most strongly associated with frequency of social comparison on Facebook.
Taken together, as we develop the strategic plan to approach well-being across FB (which is in the works), social comparison is one of, if not THE, most important areas to focus on given associations to well-being and public perception of us harming. Removing like counts is one of, if the THE, most significant levers FB has to reduce social comparison. From a well-being perspective, a ship decision for both IG and FB is recommended – based on above data and survey data from tests related to social comparison/well-being (see Moira's summary). If we decide not to ship on FB, we should be prepared for well-being hits for Facebook in the press and deepening the narrative that FB is falling behind IG on supporting well-being."
Moira Burke: "Hiding Like counts may seem like an unimportant change for well-being. But as Jennifer outlines below, like counts are a key driver of social comparison, and social comparison is a key driver of well-being. There are many other things we could do to reduce social comparison (e.g., downranking self-aggrandizing content, positive emotion, or certain topics in News Feed, or identifying people who may be vulnerable and providing special tools for them). However, those approaches require complex (infeasible) machine learning, the triggers are different for everyone, may do more harm than good (seeing friends' happy moments is beneficial to most people), may reinvigorate the emotional contagion press, and may be inconsistent with GDPR (e.g., inferring mental health status, even with a goal of helping). By comparison, Daisy will have real (if moderate) impact on people's lives, and our tests have been well received by the press. This google doc has more detail on Daisy tradeoffs and relationships to past social comparison research.
We understand that there are many factors to consider in this decision, including sessions, revenue, finding popular content, and the comms environment. I hope that the well-being implications can be factored into the decision, as well."
"I get that some people, especially teens, may feel self-conscious about posting content that doesn't get enough likes. I'm not sure if the tests showed this was actually helpful at all for those use cases, but that's not really relevant to my own use... In the realm of intermediate steps, one more thing we may want to consider is publishing the research now and trying to correct the misperception that this change has any impact on well-being. If this had the tenor of us simply evaluating a product change rather than addressing a social issue -- and, to be clear, the former is what we're doing -- then we might feel more flexibility in exploring different outcomes here."
"Repeatedly over the past year, across topics and workstreams, we've heard teens complain bitterly about Instagram's non-chronological feed... Although the non-chronological feed fuels increased time spent on the site, teen users say that it makes it much more difficult for them to find relevant info about their closer friends."
"The best external research indicates that Facebook's impact on people's well-being* is negative. FB can have a negative effect when people feel a lack of control over the experience or when use undermines other aspects of life."
"We've identified 3 drivers that might be contributing to negative well-being**: Problematic Use, Loneliness, and Social Comparison"
"Problematic use affects about 12.5% of people on Facebook."
"People perceive the impact. In a comparative survey with competitors, people perceived lower well-being and higher problematic use on Facebook compared to any other service. (Twitter, YouTube, IG, WA, Snapchat, Reddit, World of Warcraft)"
"Loss of productivity - (1) Giving up. The person won't do the task/ activity outside FB. (2) Postponing the beginning. The person will take more time to start the task/ activity off FB. (3) Extending the duration. The person will take more time to complete the task because will stop to check FB.
Sleep disruption – (1) Delaying/reducing sleep hours due to loss of time control; (2) Waking up and checking FB prolonging a return to sleep; and (3) Sleep loss due to disturbing content, like politics or violence.
Relationship impacts - People together but separately on their phones. And a false sense of closeness created by seeing updates on FB, but fewer offline visits.
Parents neglecting their kids – Parents focused more on FB than caring for or bonding with their children. Or refrained from reprimanding kids for risk of later "retaliation" on their own use.
Safety risks – Loss of situational awareness around environmental dangers (moving vehicles, pickpockets).
Regretful purchases – Spending money on things they didn't need."
"We heard about 10+ triggers contributing to PU habits."
"Notifications – Getting too many minor/irrelevant notifications. Try to only look at important ones but get sucked into longer sessions.
Entertaining content - Many said they'd open FB with a clear intent (like checking the news, a specific group, or work-related posts), then get distracted by something entertaining
Videos – Easy to get immersed (especially before bedtime). Auto-play exacerbates the issue.
Groups – Get exponentially more notifications, engaged in chat threads
Fear of missing out (FOMO) – Worry about missing important world news or updates in their social circles.
Mystery of the algorithm – Uncertainty over if they will see posts from those they want; if they can find a post again later.
Marketplace / Ads – Vigilance of buyers/sellers, sales, lower resistance to purchasing at night
Recently post/comment – Higher curiosity to see responses
Ephemeral content (e.g. Stories, birthdays) - Catch it before it's gone
Boredom/free time – Desire to fill downtime or "time pass""
"Red dots are toxic on the home screen."
"People liking things can be addictive. I feel compelled to see who liked it. I think it's a bad habit because [I'm] always checking."
"I turn off notifications. I don't want them to be poppin' up on me. If I commented on your picture, then I get a notification from people commenting who aren't even my friends."
"1. Protect teens from unwanted DMs asap to keep them safe where bullying happens most
2. Protect teens' safety across additional interaction types while maintaining a simple privacy model"
"Why not just do one?
- We know from research that DMs is necessary for safety, though it's not expected.
- We know from research that mentions and tags are expected. Matching user expectations keeps the experience simple.
- We know bullying happens through mentions and tags and that private teens are more likely to be bullied this way. It's not a safe experience if we don't include mention and tags.
- We'll have a frankenstein privacy model if we only include DMs, that will not match users expectations and therefore will not be simple."
[6. Default all new teens into private + restricted interactions via registration, A. Default all new teens into private + restricted interactions via registration flow]:
• Pros: "Keep more people safe from unwanted DMs, tags, and mentions"
• Cons / Risks: "More risk to engagement metrics", "Could bloat graph", "People may not know a follow requester"
• Mitigation: "Nudge people to follow someone they want to interact with", "Allow messages when requesting to follow someone"
[6. Default all new teens into private + restricted interactions via registration, B. Upsell all teens via opt-in QP]:
• Pros: "Less risk to engagement metrics"
• Cons / Risks: "Teens may not understand the upsells as new users"
• Mitigation: "Show after X number of days"
[6. Default all new teens into private + restricted interactions via registration, C. Upsell all teens via opt-out QP]:
• Pros: "Less risk to engagement metrics"
• Cons / Risks: "Teens may not understand the upsells as new users PXFN may block"
• Mitigation: "Show after X number of days"
Margaret Stewart: "Some of the things that Schrep wanted to understand:... Any analysis of effects on the business (growth/revenue) that the current restrictions are having"
Pratiti Raychoudhury: "Age-gate: While we have already called out the risks, there are a couple of other things to keep in mind –
Because we are seeing some of our well-being related efforts are effecting teens more disproportionately, we keep recommending this as a potential path. The problem is that soon we'll have several things that we will recommend (as a company) to age-gate and it if we do all of them we'll have a very stripped down version for Teens"
Karina Newton: "On the age points below will add two notes for IG that may become relevant/want to ensure there is awareness of — first is that we have very limited age information on IG (we only started collecting age in December at registration) so age gating solutions are less desirable. Secondly, teens are some of the biggest users of these filters (and teen girls in particular) on IG so giving them an incentive to lie about their age to get access is very problematic from an integrity and policy standpoint."
Wendy Gross: "The feedback, essentially, is that (1) teens feel addicted to IG and feel a pressure to be present, (2) like addicts, they feel that they are unable to stop themselves from being on IG, and (3) the tools we currently have aren't effective at limiting their time on the ap[p]"
████: "does this sound OK to you? 'Time Spent tools are crucial proof points when speaking to consumers, the press, and stakeholders about our efforts to combat social media addiction. We've also recently heard from teens that the amount of time they spend on IG is one of the worst aspects of their relationship with the app.'"
Wendy Gross: "Do we want to call it addiction?"
████: "Maybe not
Actually yah I think that's the term that press and stakeholders use
(This is internal only)"
Wendy Gross: "Internal only makes it better. I'm just a little cautious about calling it an addiction"
████: "Totally agree, we would never want to say that!"
"My biggest take away from last night was the change over time piece. It absolutely confounded them that we could be working on an issue so long, yet we are where we are. Rather than it seeming responsible, they took it as 'so you've known about this all this time and you still haven't fixed it'."
"Thanks for the quick Community Review sync on Friday. Decisions we aligned on:
1. Default teens to private in June.
2. Interactions and private/public settings (visibility) are separate constructs.
3. Build a new audience control that allows your followers to engage with you.
4. Test and launch mentions, dms, and tag settings for everyone.
5. Don't default teens to restricted interaction settings, but make it easy to find these settings when people need them, such as by making them available on the profile.
6. Do more research (such as a diary study) to understand whether the majority of interactions teens receive from outside their network are unwanted."
"We will not create any new default interaction settings for private accounts since we decided visibility, discoverability and interactions are three different constructs and private = visibility."
"1) Your data pretty succinctly shows that taking away unwanted interactions via private default settings is likeley to lead to a potentially untenable problem with engagement and growth. 2)This is supported by qualitative data that teens dont want to miss out on discovering, connectin, and engaging.
3) The data also shows that we would fix a lot o the unwanted interaction problem by making DM's private. At what cost. to paraphrase ████ how do we increase safety without decreasing value? incedentally, his original wording 'What are the other ways we could increase safety without decreasing value?' is indicative of him wanting to keep the defaults."
"[G]ive us a story that frames our decision that we've already made around defaulting (safety vs value frame) 2) add any other insights that come up regarding new teens DM interactions"
Sources include internal documents, public filings and press reports. “Alleged by plaintiffs” denotes the plaintiffs’ characterization, not Meta’s own words. Source labels identify provenance, not a finding of fact.
Meta & Children's Online Safety