Chat: we really need to take a very close look at what sessions (metric) is doing to vulnerable sub populations

Raw text

Transcribed text as recorded, including headers and formatting fragments.

Case 4:22-md-03047-YGR Document 2889-40 Filed 03/27/26 Page 1 of 4 AMENDED Exhibit 247 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 2889-40 Filed 03/27/26 Page 2 of 4 Message From: Sent: To: Subject: Miki Rothschild [/O=THEFACEBOOK/OU=EXTERNAL (FYDIBOHF25SPDLT)/CN=RECIPIENTS/CN= 9/28/2020 6:03:41 AM Miki Rothschild Message summary [{"otherUserFbId":100008503654679,"threadFbId":null}] Attachments: 120183674_635015003699293_5334994313740145970_n.png Michael Yehuda Rothschild (9/27/2020 19:30:35 PDT): >#quiet just got around to it. Great note! Do you know if there's any role models for this type of work across the industry? Which company's AI strategy is most aligned with human value and how. I have my own opinion but curious to hear your POV (9/27/2020 22:50:18 PDT): >manks! I am a part of an external working group called Partnership on AI from RAI is on the board or some such leadership position for PAI. So AI value alignment is a wett researched topic among external AI experts (the panel I am on is along with Google AI engineers, AI ethics profs from UC SF etc. it's a good group). My 2 cents is we are a bit behind. I reached out to folks in our AI org and RAI org but there doesn't seem to be a systematic set up within the company. Several leads across the company are interested in this space though ( for example) but I am wondering how we might make this more systematic. This is not like a 6m or 1 year project but a mental shift and like a 2-5yr investment to evolve AI to better human outcomes. (9/27/2020 22:50:51 PDT): >understandably, it is too long term to easily generate a ton of excitement easily among engg. (9/27/2020 22:51:53 PDT): >Jonn is pretty supportive of this (as you can see by his comments).. (9/27/2020 22:52:09 PDT): >But hard, long term initiative... Michael Yehuda Rothschild (9/27/2020 22:52:13 PDT): >Totally agree. FWIW, I think we are systematically in a much worse place here compared to for eg Google Search (9/27/2020 22:52:22 PDT): >yes, tend to agree (9/27/2020 22:52:33 PDT): >BUL we are so short term focused on sessions, MSI and so on... Michael Yehuda Rothschild (9/27/2020 22:52:54 PDT): >The reason being that most of our AI systems are using behavioral user implcit signals vs Google which uses objective human ratings (9/27/2020 22:53:27 PDT): >I recently talked to an ex Googler, Google Travel apparently had 2 top line metrics: sentiment (user satisfaction thru' surveys) and DAP. Michael Yehuda Rothschild (9/27/2020 22:53:29 PDT): >So while we think we are optimizing for signals that correlate with good things we often don't actually know all the patterns that take place (9/27/2020 22:54:07 PDT): >yes...I think surveys are not perfect but closer to the ground truth than behavioral signals but a ton of challenges when it comes to survey based modeling and measurements Michael Yehuda Rothschild (9/27/2020 22:54:46 PDT): >I'm a pretty big fan of what Ads team on IG do Michael Yehuda Rothschild (9/27/2020 22:54:52 PDT): >I'll send you a deck (9/27/2020 22:54:59 PDT): >so folks probably don't want to hear this as the solution but feel like we need some sort of internal AI ethics panel...that pushes the boundary on value alignment and AI safety Michael Yehuda Rothschild (9/27/2020 22:55:05 PDT): >OOPS if you've heard about that CONFIDENTIAL META3047MDL-014-00350817 Case 4:22-md-03047-YGR Document 2889-40 Filed 03/27/26 Page 3 of 4 (9/27/2020 22:55:14 PDT): >mm no, I don't think am familiar (9/27/2020 22:55:50 PDT): >prease share. Michael Yehuda Rothschild (9/27/2020 22:55:57 PDT): >(y) (9/27/2020 22:56:08 PDT): >(also it's a modest plan but we finally are aligning on this feed fairness plan: https://fb.quip.com/BkeXALXAMDYL) (9/27/2020 22:56:24 PDT): >II you have further ideas on AI value alignment, pls do share. I shared my note with fwiw. Michael Yehuda Rothschild (9/27/2020 22:56:39 PDT): >(y) Michael Yehuda Rothschild (9/27/2020 22:58:07 PDT): >I just watched the social dilema yesterday and it made me sad b/c while I don't agree with a lot of what they say there's some truth to it. Have you watched it? (9/27/2020 22:58:17 PDT): >Yes the day it came out. (9/27/2020 22:58:51 PDT): >so the 3 men in the suits and the human puppet is a bit much but I agree with the core points (not the way it is sensationalized): like there seems to be a problem with our business model and definitely a problem with our metrics. (9/27/2020 22:59:17 PDT): >we really need to take a very close look at what sessions (metric) is doing to vulnerable sub populations for example: teens, people with mental health issues, sensitive geographies. Michael Yehuda Rothschild (9/27/2020 22:59:33 PDT): >Yes (9/27/2020 22:59:47 PDT): >Within feed, we are attempting to operationalize a sentiment based metric... (9/27/2020 23:00:10 PDT): >put survey based metrics don't have a level playing ground with behavioral metrics like sessions and MSI. Michael Yehuda Rothschild (9/27/2020 23:00:12 PDT): >And into "rabbit holes" caused by our collaborative filters loops (9/27/2020 23:00:49 PDT): >our metrics are partly to blame and this ties to value alignment. You give the AI system a metric. It assumes that IS the ground truth and intent and will keep on optimizing until you build in strong guardrails. Michael Yehuda Rothschild (9/27/2020 23:01:04 PDT): >I once got enticed by some snake video in watch, now all my recommendations occasionally have snake videos show up. It's terrible (9/27/2020 23:01:08 PDT): >So if a person vulnerable to conspiracy theories will watch more of those videos and comment more and share more, we will keep showing them more (9/27/2020 23:01:16 PDT): >ugn (9/27/2020 23:01:28 PDT): >pecause it gives us more MSI and sessions. Michael Yehuda Rothschild (9/27/2020 23:01:50 PDT): >50, >BTW, I think it might be simpler to rely more on product design and controls, and reduce complexity of our AI systems (9/27/2020 23:01:58 PDT): >and Humane tech have a podcast, don't know if you are familiar... Michael Yehuda Rothschild (9/27/2020 23:02:18 PDT): >Yeah, I am (9/27/2020 23:02:27 PDT): CONFIDENTIAL META3047MDL-014-00350818 Case 4:22-md-03047-YGR Document 2889-40 Filed 03/27/26 Page 4 of 4 >I once obsessively heard all the podcasts and compiled notes here: https://fb.quip.com/aSTeAQkYam8f (9/27/2020 23:02:27 PDT): >It's bad. (9/27/2020 23:02:38 PDT): >(we hosted for a workshop, wasn't great) (9/27/2020 23:03:39 PDT): >imis is my tldr; one slide compilation (9/27/2020 23:03:41 PDT): shared: 120183674_635015003699293_5334994313740145970_n.png CONFIDENTIAL META3047MDL-014-00350819