`
`Nos. 22-277 & 22-555
`
`IN THE
`Supreme Court of the United States
`
`ASHLEY MOODY, ATTORNEY GENERAL OF FLORIDA, et al.,
`Petitioners,
`
`v.
`
`NETCHOICE, LLC, DBA NETCHOICE, et al.,
`Respondents.
`
`
`NETCHOICE, LLC, DBA NETCHOICE, et al.,
`Petitioners,
`
`v.
`
`KEN PAXTON, ATTORNEY GENERAL OF TEXAS,
`Respondent.
`
`
`ON WRITS OF CERTIORARI TO THE
`UNITED STATES COURTS OF APPEALS
`FOR THE FIFTH AND ELEVENTH CIRCUITS
`
`BRIEF FOR DEVELOPERS ALLIANCE AND
`SOFTWARE & INFORMATION INDUSTRY
`ASSOCIATION AS AMICI CURIAE IN
`SUPPORT OF NETCHOICE, LLC AND CCIA
`
`
`SONAL N. MEHTA
`WILMER CUTLER PICKERING
` HALE AND DORR LLP
`2600 El Camino Real, Ste. 400
`Palo Alto, CA 94306
`
`HANNAH E. GELBORT
`WILL C. DARIO
`WILMER CUTLER PICKERING
` HALE AND DORR LLP
`60 State Street
`Boston, MA 02109
`
`ARI HOLTZBLATT
` Counsel of Record
`ALLISON M. SCHULTZ
`WILMER CUTLER PICKERING
` HALE AND DORR LLP
`2100 Pennsylvania Ave., NW
`Washington, DC 20037
`(202) 663-6000
`ari.holtzblatt@wilmerhale.com
`
`
`
`
`
`
`
`
`
`TABLE OF CONTENTS
`
`Page
`TABLE OF AUTHORITIES ......................................... iii
`INTEREST OF AMICI CURIAE................................... 1
`INTRODUCTION AND SUMMARY OF
`ARGUMENT ................................................................ 2
`ARGUMENT ....................................................................... 6
`I. SOFTWARE
`DEVELOPERS
`DESIGN
`ALGORITHMS AS TOOLS TO EXECUTE AND
`AMPLIFY A WEBSITE’S EXPRESSION BY
`CURATING AND MODERATING CONTENT
`CREATED BY OTHERS .................................................. 6
`A. Algorithms Operationalize A Website’s
`Content Policies And Priorities .......................... 6
`B. Websites Create Many Different
`Algorithms To Implement Myriad
`Content-Related Policies ................................... 11
`1. Websites deploy moderation
`algorithms to detect and remove
`content that violates their
`expressive standards ................................... 11
`2. Websites deploy prioritization and
`ranking algorithms to match
`content to interested users ......................... 19
`3. Algorithms complement and
`facilitate websites’ human
`moderation and curation efforts ................ 23
`4. Developers refine websites’
`content prioritization, ranking, and
`moderation strategies and
`
`
`
`
`
`ii
`
`TABLE OF CONTENTS—Continued
`
`Page
`
`algorithms to maintain the desired
`messaging ...................................................... 25
`II. THE FACT THAT WEBSITES IMPLEMENT
`THEIR EDITORIAL AND CURATORIAL
`DECISION MAKING THROUGH ALGORITHMS
`DOES NOT STRIP THEM OF FIRST
`AMENDMENT PROTECTION ........................................ 28
`A. The First Amendment Protects
`Editorial And Curatorial Decision
`Making .................................................................. 28
`B. The Same Protections Apply To Coded
`Algorithms Designed To Curate And
`Organize Content Created By Others ............. 30
`CONCLUSION ................................................................. 35
`
`
`
`
`
`iii
`
`TABLE OF AUTHORITIES
`CASES
`
`Page(s)
`Bartnicki v. Vopper, 532 U.S. 514 (2001) ....................... 28
`Hurley v. Irish-American Gay, Lesbian, &
`Bisexual Group of Boston,
`515 U.S. 557 (1995)............. 2, 5, 6, 23, 29, 30, 32, 34
`Los Angeles v. Preferred Communications,
`Inc., 476 U.S. 488 (1986)........................................ 29
`Manhattan Community Access Corp. v. Hal-
`leck, 139 S. Ct. 1921 (2019) ................................... 29
`Miami Herald Publishing Co. v. Tornillo,
`418 U.S. 241 (1974)........................... 5, 28, 29, 30, 32
`Pacific Gas & Electric Co. v. Public Utilities
`Commission of California, 475 U.S. 1
`(1986) ....................................................................... 29
`Rumsfeld v. Forum for Academic & Institu-
`tional Rights, Inc., 547 U.S. 47 (2006) .......... 32, 34
`Seattle Times Co. v. Rhinehart, 467 U.S. 20
`(1984) ....................................................................... 31
`Sorrell v. IMS Health Inc., 564 U.S. 552
`(2011) ................................................................. 28, 31
`Turner Broadcasting System, Inc. v. FCC,
`512 U.S. 622 (1994)................................................. 29
`Twitter, Inc. v. Taamneh, 598 U.S. 471 (2023) ........ 32, 33
`
`STATUTORY PROVISIONS
`Fla. Stat. §501.2041 ..................................................... 15, 25
`Tex. Bus. & Com. Code §120.051 ..................................... 25
`
`
`
`
`
`iv
`
`TABLE OF AUTHORITIES—Continued
`
`Page
`Tex. Civ. Prac. & Rem. Code §143A.002 ........................ 15
`
`OTHER AUTHORITIES
`About, Etsy, https://www.etsy.com/about
`(visited Dec. 6, 2023) ............................................ 20, 21
`Adult Nudity and Sexual Activity: Policy
`Details, Meta Transparency Center,
`https://transparency.fb.com/en-gb/poli-
`cies/community-standards/adult-nudity-
`sexual-activity/ (visited Dec. 6, 2023) ................ 14, 15
`Anti-Defamation League, Sliding Through:
`Spreading Antisemitism on TikTok by
`Exploiting Moderation Gaps (Nov. 20,
`2023), https://www.adl.org/resources/
`blog/sliding-through-spreading-antisemi-
`tism-tiktok-exploiting-moderation-gaps ................. 27
`BBC, Inputs and Outputs, https://www.bbc
`.co.uk/bitesize/topics/zs7s4wx/arti-
`cles/z7wckty#z9cybqt (visited Dec. 6, 2023) ............. 8
`Bernal, Natasha, Facebook’s Content Modera-
`tors Are Fighting Back, Wired (Nov. 6,
`2021), https://www.wired.co.uk/article/face-
`book-content-moderators-ireland ............................ 24
`Blazina, Carrie & Galen Stocking, Key Facts
`About Parler, Pew Research Center (Oct.
`20, 2022), https://www.pewre-
`search.org/short-reads/2022/10/20/fast-
`facts-about-parler-as-kanye-west-report-
`edly-plans-acquisition-of-site/ ..................................... 7
`
`
`
`
`
`v
`
`TABLE OF AUTHORITIES—Continued
`
`Page
`
`Buckley, Nicole & Joseph S. Schafer, “Censor-
`ship-Free” Platforms: Evaluating Content
`Moderation Policies and Practice of
`Alternative Social Media, 4 For(e)Dialogue
`1 (Feb. 3, 2022), https://foredialogue.pub-
`pub.org/pub/bsh5uhll/release/1 ................................ 16
`Carlson, Nicholas, Pinterest CEO: Here’s How
`We Became the Web’s Next Big Thing,
`Business Insider (Apr. 24, 2012),
`https://www.businessinsider.com/pinterest-
`founding-story-2012-4 .................................................. 7
`Chowdhury, Nafia, Automated Content Moder-
`ation: A Primer, Stanford Cyber Policy
`Center Program on Platform Regulation
`(Mar. 19, 2022), https://cyber.fsi.stan-
`ford.edu/news/automated-content-modera-
`tion-primer .................................................................. 17
`Darbinyan, Rem, The Growing Role of AI
`in Content Moderation, Forbes
`(June 14, 2022), https://www.forbes.com/
`sites/forbestechcouncil/2022/06/14/the-
`growing-role-of-ai-in-content-modera-
`tion/?sh=40ccbb474a17 ....................................... 8, 9, 24
`Desjardins, Jeff, How Much Data Is Generated
`Each Day?, World Economic Forum
`(Apr. 17, 2019), https://www.wefo-
`rum.org/agenda/2019/04/how-much-data-is-
`generated-each-day-cf4bddf29f/ ................................. 9
`
`
`
`
`
`vi
`
`TABLE OF AUTHORITIES—Continued
`
`Page
`
`Facebook Community Standards, Meta
`Transparency Center, https://transpar-
`ency.fb.com/policies/community-standards/
`(visited Dec. 6, 2023) .................................................. 10
`Farid, Hany, An Overview of Perceptual
`Hashing, J. Online Trust & Safety (2021),
`https://tsjournal.org/index.php/jots/arti
`cle/view/24/14 .............................................................. 12
`GIFCT’s Hash-Sharing Database, Global
`Internet Forum to Counter Terrorism,
`https://gifct.org/hsdb/ (visited Dec. 6, 2023) ........... 12
`Godtube, About Us, https://www.godtube.com/
`about-godtube.html (visited Dec. 6, 2023) ................ 3
`Godtube, Terms Of Use, https://www.god
`tube.com/terms-of-use.html (visited Dec. 6,
`2023) ............................................................................... 4
`Goodrow, Cristos, On YouTube’s Recommen-
`dation System, YouTube Official Blog
`(Sept. 15, 2021), https://blog.youtube/in
`side-youtube/on-youtubes-recommenda
`tion-system/ ........................................................... 20, 33
`Google, Search Quality Rater Guidelines:
`An Overview, Google, https://services
`.google.com/fh/files/misc/hsw-sqrg.pdf (vis-
`ited Dec. 6, 2023) ......................................................... 24
`
`
`
`
`
`vii
`
`TABLE OF AUTHORITIES—Continued
`
`Page
`
`Gorwa, Robert, et al., Algorithmic Content
`Moderation: Technical and Political
`Challenges in the Automation of
`Platform Governance, 7 Big Data &
`Soc’y 1 (2020) ......................................... 4, 11, 12, 17, 33
`Hall, Mark, Facebook, Britannica, https://
`www.britannica.com/topic/Facebook (up-
`dated Dec. 6, 2023)........................................................ 7
`Heilweil, Rebecca, Parler, the “Free Speech”
`Social Network, Explained, Vox, https://
`www.vox.com/recode/2020/11/24/21579357/
`parler-app-trump-twitter-facebook-censor
`ship (updated Jan. 11, 2021) ........................................ 7
`Hirose, Alyssa, 2023 Instagram Algorithm
`Solved: How to Get Your Content Seen,
`Hootsuite Blog (Apr. 12, 2023), https://
`blog.hootsuite.com/instagram-algorithm/ ......... 26, 27
`Hootsuite, https://www.hootsuite.com/
`(visited Dec. 6, 2023) .................................................. 26
`How Etsy Search Works, Etsy Help Center,
`https://help.etsy.com/hc/en-us/articles/
`115015745428-How-Etsy-Search-Works
`?segment=selling (visited Dec. 6, 2023) .................. 21
`Keller, Daphne, What the Supreme Court Says
`Platforms Do, Lawfare (Sept. 14, 2023),
`https://www.lawfaremedia.org/article/
`what-the-supreme-court-says-platforms-do .......... 25
`
`
`
`
`
`viii
`
`TABLE OF AUTHORITIES—Continued
`
`Page
`
`Merriam-Webster, https://www.merriam-web-
`ster.com/dictionary/algorithm (visited Dec.
`6, 2023) ........................................................................... 8
`Meta, Case On Breast Cancer Symptoms And
`Nudity, https://transparency.fb.com/over
`sight/oversight-board-cases/breast-cancer-
`symptoms-nudity (updated June 12, 2023) ............. 15
`Meta, Facebook Community Standards,
`https://transparency.fb.com/policies/com
`munity-standards/ (visited Dec. 6, 2023)................... 3
`Meta’s Third-Party Fact-Checking Program,
`Meta, https://www.facebook.com/formedia/
`mjp/programs/third-party-fact-checking/
`(visited Dec. 6, 2023) ............................................ 10, 24
`Mosseri, Adam, New Features, Instagram
`(Apr. 20, 2022), https://www.instagram.
`com/reel/Cck9Do_gOPH/?utm_source=
`ig_embed&utm_campaign=loading ......................... 22
`Newberry, Christina, 2023 TikTok Algorithm
`Explained + Tips to Go Viral, Hootsuite
`(Feb. 8, 2023), https://blog.hootsuite.com/
`tiktok-algorithm/ ........................................................ 26
`No, Your News Feed Is Not Limited to Posts
`From 26 Friends, Meta (Feb. 6, 2019),
`https://about.fb.com/news/2019/02/inside-
`feed-facebook-26-friends-algorithm-myth/ ............. 22
`
`
`
`
`
`ix
`
`TABLE OF AUTHORITIES—Continued
`
`Page
`
`Notices on X and what they mean, X Help
`Center, https://help.twitter.com/en/rules-
`and-policies/notices-on-x (visited Dec. 6,
`2023) ............................................................................. 10
`O’Connor, Jennifer F. & Emily Moxley, Our
`Approach to Responsible AI Innovation,
`YouTube Official Blog (Nov. 14, 2023),
`https://blog.youtube/inside-youtube/our-ap
`proach-to-responsible-ai-innovation/ ....................... 18
`Olafson, Karin, How to Use Pinterest for Busi-
`ness: 8 Strategies You Need To Know,
`Hootsuite (July 29, 2021), https://blog.hoot
`suite.com/how-to-use-pinterest-for-busi
`ness/ ................................................................................ 7
`OpenAI Says AI Tools Can Be Effective in
`Content Moderation, Reuters (Aug. 15,
`2023), https://www.reuters.com/technol
`ogy/openai-says-ai-tools-can-be-effective-
`content-moderation-2023-08-15/ ......................... 23, 24
`Our Approach to Facebook Feed Ranking,
`Meta, https://transparency.fb.com/fea
`tures/ranking-and-content/ (updated
`Nov. 28, 2023) .................................................. 10, 19, 21
`Our Approach to Policy Development and En-
`forcement Philosophy, X, https://help.twit
`ter.com/en/rules-and-policies/enforcement-
`philosophy (visited Dec. 6, 2023) .............................. 17
`Our Commitment, BitChute, https://support.
`bitchute.com/policy/our-commitment/ (vis-
`ited Dec. 6, 2023) ......................................................... 17
`
`
`
`
`
`x
`
`TABLE OF AUTHORITIES—Continued
`
`Page
`
`Proactive Rate, Meta Transparency Center,
`https://transparency.fb.com/policies/im
`proving/proactive-rate-metric/ (updated
`Feb. 22, 2023) .............................................................. 14
`RoBERTa: An optimized method for pretrain-
`ing self-supervised NLP systems, Meta
`(July 29, 2019), https://ai.meta.com/blog/
`roberta-an-optimized-method-for-pretrain
`ing-self-supervised-nlp-systems/ ............................. 18
`Samanta, Priyanka & Shweta Jain, Analysis of
`Perceptual Hashing Algorithms in
`Image Manipulation Detection,
`185 Procedia Comp. Sci. 203 (2021) .......................... 11
`Search, Google, https://www.google.com/
`search/howsearchworks/how-search-
`works/ranking-results/ .............................................. 10
`Singhal, Mohit, et al., SoK: Content Moderation
`in Social Media, from Guidelines to
`Enforcement, and Research to Practice
`(Mar. 2023), https://arxiv.org/pdf/2206.
`14855.pdf ...................................................... 9, 10, 11, 16
`Steinberg, Kayla, People Are Using Coded
`Language To Avoid Social Media
`Moderation. Is It Working?, PolitiFact
`(Nov. 4 2021), https://www.politifact.com/
`article/2021/nov/04/people-are-using-coded-
`language-avoid-social-media/ .................................... 27
`
`
`
`
`
`xi
`
`TABLE OF AUTHORITIES—Continued
`
`Page
`
`Tariq, Muhammad U., et al., A Review of the
`Gaps and Opportunities of Nudity and
`Skin Detection Algorithmic Research for
`the Purpose of Combating Adolescent Sex-
`ting Behaviors, in Human-Computer Inter-
`action; Design Practice in Contemporary
`Societies, vol. 11568 (M. Kurosu ed., June
`27, 2019), https://link.springer.com/chap
`ter/10.1007/978-3-030-22636-7_6 ............................... 14
`Tomasik, Emily & Galen Stocking, Key Facts
`About BitChute, Pew Research Center
`(Feb. 17, 2023), https://www.pewresearch.
`org/short-reads/2023/02/17/key-facts-about-
`bitchute/ ....................................................................... 17
`Transparency Report, Pinterest, https://policy.
`pinterest.com/en/transparency-report (vis-
`ited Dec. 6, 2023) ......................................................... 13
`Truth, Community Guidelines, https://help.
`truthsocial.com/community-guidelines-
`page/ (visited Dec. 6, 2023) .......................................... 3
`Types of Content We Demote, Meta, https://
`transparency.fb.com/features/approach-to-
`ranking/types-of-content-we-demote (up-
`dated Oct. 16, 2023) .................................................... 10
`Using AI to Detect COVID-19 Misinformation
`and Exploitative Content, Meta (May 12,
`2020), https://ai.meta.com/blog/using-ai-to-
`detect-covid-19-misinformation-and-exploi
`tative-content/............................................................. 16
`
`
`
`
`
`xii
`
`TABLE OF AUTHORITIES—Continued
`
`Page
`
`Weimann, Gabriel & Natalie Masri,
`TikTok’s Spiral of Antisemitism,
`2 Journalism & Media 697 (2021) .............................. 27
`What are Convolutional Neural Networks?,
`IBM, https://www.ibm.com/topics/convolu
`tional-neural-networks (visited Dec. 6,
`2023) ............................................................................. 16
`What is the Hash Sharing Database?, Global
`Internet Forum to Counter Terrorism,
`https://gifct.org/?faqs=what-is-the-hash-
`sharing-database (visited Dec. 6, 2023) ................... 13
`
`
`
`
`
`
`
`
`
`IN THE
`Supreme Court of the United States
`
`No. 22-277
`ASHLEY MOODY, ATTORNEY GENERAL OF FLORIDA, et al.,
`Petitioners,
`
`v.
`NETCHOICE, LLC, DBA NETCHOICE, et al.,
`Respondents.
`
`
`
`NO. 22-555
`NETCHOICE, LLC, DBA NETCHOICE, et al.,
`Petitioners,
`
`v.
`KEN PAXTON, ATTORNEY GENERAL OF TEXAS
`Respondent.
`
`
`ON WRITS OF CERTIORARI TO THE
`UNITED STATES COURTS OF APPEALS
`FOR THE FIFTH AND ELEVENTH CIRCUITS
`
`BRIEF FOR DEVELOPERS ALLIANCE AND
`SOFTWARE &INFORMATION INDUSTRY
`ASSOCIATION AS AMICI CURIAE IN
`SUPPORT OF NETCHOICE, LLC AND CCIA
`
`
`INTEREST OF AMICI CURIAE1
`The Developers Alliance is a non-profit corporation
`that advocates for software developers. Its corporate
`
`
`1 No counsel for a party authored this brief in whole or in part,
`and no entity or person, other than amicus curiae, its members, and
`
`
`
`
`
`2
`
`mission is to “[a]dvocate on behalf of developers and the
`companies that depend on them, support the industry’s
`continued growth, and promote innovation.” As a group
`created by and for software developers, Developers Al-
`liance recognizes that software created by developers
`reflects the judgments, decision making, and creative
`thinking of those developers. Alliance members include
`industry leaders in consumer, enterprise, industrial, and
`emerging software, and a global network of more than
`75,000 developers.
`The Software and Information Industry Association
`(“SIIA”) is the principal trade association for those in
`the business of information. SIIA’s membership in-
`cludes more than 500 software companies, platforms,
`data and analytics firms, and digital publishers that
`serve nearly every segment of society, including busi-
`ness, education, government, healthcare, and consum-
`ers. It is dedicated to creating a healthy environment
`for the creation, dissemination, and productive use of in-
`formation. SIIA protects the rights of its members to
`use software as a tool for the dissemination of infor-
`mation.
`
`INTRODUCTION AND SUMMARY OF ARGUMENT
`Publishing or disseminating speech created by oth-
`ers inherently requires deciding what content is “worthy
`of presentation” and to whom, Hurley v. Irish-American
`Gay, Lesbian & Bisexual Group of Boston, 515 U.S. 557,
`575 (1995). Newspapers highlight stories they think are
`interesting by placing them on the front page and bury
`less interesting stories by putting them later in the pa-
`per or leaving them on the cutting room floor. A Catholic
`
`its counsel, made a monetary contribution intended to fund the
`preparation or submission of this brief.
`
`
`
`
`
`3
`
`bookstore might promote religious-oriented books in its
`front window, while a children’s bookstore typically car-
`ries only books appropriate for minors. And a sports-
`loving newsstand usually allocates prime location to
`sports magazines or newspapers devoted to local teams.
`Each of these choices expresses a view about the nature
`of the newspaper, bookstore, or newsstand.
`Websites likewise express themselves through their
`choices about what content to display and how—and that
`remains true, even when they use algorithms, rather
`than direct human review, to carry out those decisions.
`When these websites disseminate speech to their users,
`they convey a message about the type of speech they
`find acceptable and about the kind of community they
`hope to foster. To shape those messages, each website
`has various content-related rules and policies that re-
`flect each website’s expressive vision. Truth Social pro-
`hibits content that could “interfere with [its] goal of
`providing a welcoming platform,” including content that
`violates a person’s privacy rights.2 And to foster a “safe
`place” for “authentic” communication, Facebook prohib-
`its hate speech, violent content, harassment, and misin-
`formation.3 Godtube, a “Christian video sharing web-
`site,”4 prohibits “name calling,” “vulgar terms,” or “sex-
`ually suggestive” content in order to create “the safest
`
`
`
`2 Truth,
`https://help.truth-
`Guidelines,
`Community
`social.com/community-guidelines-page/ (visited Dec. 6, 2023).
`3 Meta, Facebook Community Standards, https://transpar-
`ency.fb.com/policies/community-standards/ (visited Dec. 6, 2023).
`4 Godtube, About Us,
`https://www.godtube.com/about-
`godtube.html (visited Dec. 6, 2023).
`
`
`
`
`
`4
`
`and most Christian values sensitive community environ-
`ment possible.”5
`Over the years, as websites have grown, the pro-
`cesses through which they enforce their community
`standards have changed, but the resulting expressions
`of each website’s values and priorities have not. In the
`early days of the Internet, simplistic message boards
`were monitored by human administrators who reviewed
`each post added to the board.6 Posts that failed to com-
`ply with the board’s community standards—because, for
`example, they contained inappropriate language or con-
`cerned topics other than those designated for the
`board—were removed. Administrators thus culled and
`curated speech created by others to ensure an overall
`message consistent with the community’s standards and
`priorities.
`Today, those simplistic message boards have be-
`come the global websites now used by nearly 5 billion
`people worldwide. Each website still has its own rules
`and community standards, but the incredible volume of
`content shared each day makes human review of each
`new post impossible. To implement their rules and
`standards at the scale demanded by the Internet today,
`the people who administer those websites have created
`complex algorithms to translate their policies into code
`that can be executed by computers. By determining in
`advance how to identify content that violates website
`terms and should be removed or demoted, or content
`
`
`5 Godtube, Terms Of Use, https://www.godtube.com/terms-of-
`use.html (visited Dec. 6, 2023).
`6 See Gorwa et al., Algorithmic Content Moderation: Technical
`and Political Challenges in the Automation of Platform Govern-
`ance, 7 Big Data & Soc’y 1, 2 (2020).
`
`
`
`
`
`5
`
`that the website wants to highlight for particular users,
`software developers operationalize technology-neutral
`human editorial decisions. Algorithms thus incorporate
`the rules and decision making needed to implement hu-
`man content judgments.
`Texas and Florida contend that the curation and dis-
`semination of speech online is not expressive and there-
`fore does not trigger First Amendment protections. No.
`22-277 Pet. 18-22; No. 22-555 Opp. 18-20. Specifically, in
`its supplemental brief concerning certiorari, Texas con-
`tends that such content curation and moderation cannot
`be expressive because websites use algorithms, rather
`than human reviewers, to implement their editorial
`choices. No. 22-555 Texas Supp. Opp. 5. But the use of
`algorithms to operationalize website rules and standards
`does not displace human decision making. Rather, algo-
`rithms are tools for implementing human decisions.
`Computers can only do that which they are instructed,
`and so any content-moderation action taken by a com-
`puter reflects the editorial objectives of its human ad-
`ministrator, whether that is by removing certain specif-
`ically identifiable content or looking for patterns to sta-
`tistically identify content likely violative of the websites’
`rules. The fact that content curation and moderation de-
`cisions are now implemented by algorithms does not sap
`those decisions of their expressive nature.
`As this Court has long recognized, the “exercise of
`editorial control and judgment” over the speech or ex-
`pression of others is protected by the First Amendment.
`Miami Herald Publ’g Co. v. Tornillo, 418 U.S. 241, 258
`(1974). And that is true even where the resulting “edited
`compilation of speech generated by other persons” does
`not yield “a narrow, succinctly articulable message.”
`Hurley, 515 U.S. at 569-570. It is simply the “deci[sion]
`to exclude” or include particular messages that triggers
`
`
`
`
`
`6
`
`“the right [of] a private speaker to shape its expression.”
`Id. at 574. As tools for implementing human editorial
`judgments at scale, the fact that algorithms implement
`most content moderation actions does not detract from
`the First Amendment protections afforded to the edito-
`rial and curatorial judgments reflected in and carried out
`by those algorithms.
`Algorithms are the modern equivalent of an editor,
`a red pen, and a layout mockup all rolled into one. They
`implement decisions regarding what content should be
`distributed, what content should not be distributed, and
`how the content should be displayed to users. To enable
`content moderation and curation at the scale required by
`social media, those decisions are made in advance and
`applied automatically as new content is created and
`shared, but the decisions themselves are no less human
`than the red strikethrough on a newspaper mockup.
`Thus, laws—like the Texas and Florida laws here at is-
`sue—that restrict whether and how websites may mod-
`erate content implicate the websites’ First Amendment
`rights, regardless of whether those moderation decisions
`are generally carried out through algorithms rather than
`hands-on human review.
`
`ARGUMENT
`I. SOFTWARE DEVELOPERS DESIGN ALGORITHMS AS
`TOOLS TO EXECUTE AND AMPLIFY A WEBSITE’S EX-
`PRESSION BY CURATING AND MODERATING CONTENT
`CREATED BY OTHERS
`A. Algorithms Operationalize A Website’s Con-
`tent Policies And Priorities
`Websites are created for different expressive pur-
`poses and are geared toward different audiences. Face-
`book, for example, launched in 2004 as a social network
`
`
`
`
`
`7
`
`to connect friends within discrete college communities;
`by 2021 it had nearly 3 billion users worldwide.7 Pinter-
`est entered the scene in 2010 as a site to share beautiful
`images, recipes, and creative inspiration.8
` Parler
`launched in 2018 as an alternative to more traditional
`websites like Twitter and Facebook and promoted itself
`as a bastion of free speech.9
`As purveyors of speech created by others, websites
`communicate these expressive visions—regarding what
`speech they find acceptable and the kinds of communi-
`ties they hope to foster—through their curation of third-
`party content. Each website establishes its own content
`policies and priorities to guide the dissemination of
`speech. These policies and priorities are as diverse as
`the websites that created them and often evolve as the
`websites do. Some “alternative” websites’ expressive
`identities turn on the fact that their content policies tol-
`erate significantly more content than more traditional
`websites.10
` Others prohibit or restrict violent,
`
`
`7 Hall, Facebook, Britannica, https://www.britannica.com/
`topic/Facebook (updated Dec. 6, 2023).
`8 Olafson, How to Use Pinterest for Business: 8 Strategies You
`Need To Know, Hootsuite (July 29, 2021), https://blog.hoot-
`suite.com/how-to-use-pinterest-for-business/; Carlson, Pinterest
`CEO: Here’s How We Became the Web’s Next Big Thing, Business
`Insider (Apr. 24, 2012), https://www.businessinsider.com/pinterest-
`founding-story-2012-4.
`9 Heilweil, Parler,
`the “Free Speech” Social Network,
`Explained, Vox, https://www.vox.com/recode/2020/11/24/21579357/
`parler-app-trump-twitter-facebook-censorship (updated Jan. 11,
`2021).
`10 Blazina & Stocking, Key Facts About Parler, Pew Research
`Center (Oct. 20, 2022), https://www.pewresearch.org/short-reads/
`
`
`
`
`
`8
`
`harassing, false, or misleading information in order to
`create environments for safe expression.
`Websites use algorithms to implement their con-
`tent-related policies. An algorithm is a “step-by-step
`procedure for solving a problem or accomplishing some
`end.”11 Coded algorithms must be incredibly precise,
`telling computers exactly what to do, in what order, in
`response to very specifically defined inputs. That is a
`herculean task that requires significant iteration and im-
`provement. Every action that is taken by a computer,
`including every response to any input that is provided,
`is built into the computer’s code during programming
`and is a function of the instructions that are embodied in
`that code.12
`Given the volume of content posted daily to social-
`media websites, algorithms are the only feasible way for
`websites to implement their editorial judgments. In
`2022, online communications websites collectively had
`4.62 billion active users worldwide, a 10% increase from
`the previous year.13 By 2025, “463 exabytes of data will
`be created each day globally”—the “equivalent of
`
`
`2022/10/20/fast-facts-about-parler-as-kanye-west-reportedly-plans-
`acquisition-of-site/.
`11 Merriam-Webster, Algorithm, https://www.merriam-web-
`ster.com/dictionary/algorithm (visited Dec. 6, 2023).
`12 BBC, Inputs and Outputs, https://www.bbc.co.uk/bitesize/
`topics/zs7s4wx/articles/z7wckty#z9cybqt (visited Dec. 6, 2023).
`13 Darbinyan, The Growing Role of AI in Content Moderation,
`Forbes (June 14, 2022), https://www.forbes.com/sites/forbestech-
`council/2022/06/14/the-growing-role-of-ai-in-content-modera-
`tion/?sh=40ccbb474a17.
`
`
`
`
`
`9
`
`212,765,957 DVDs per day.”14 As of February 2020, us-
`ers uploaded around 500 hours of video to YouTube each
`minute.15 It would be impossible for human moderators
`alone to review, assess, and curate the “huge surge of
`user-generated content” that has emerged.16
`When the operators of a website decide what kinds
`of content to promote, demote, remove, or factcheck,
`they turn to their software developers to make that hap-
`pen. That process is shaped by the creativity of the de-
`veloper. Any number of developers given the same
`problem to address will create the same number of di-
`verse solutions; each might achieve similar results, but
`will do so in different ways, using different inputs and
`different sequences of operations.
`Software developers have numerous tools at their
`disposal to shape the websites’ content through pre-pro-
`gramed methods of moderation and curation. Develop-
`ers create “automated algorithms, which include heuris-
`tic-based and rule-based techniques as well as sophisti-
`cated machine learning-based models.”17 These algo-
`rithms and models operate based on inputs that are se-
`lected by developers and assigned relative weights
`
`
`14 Desjardins, How Much Data Is Generated Each Day?,
`World Economic Forum (Apr. 17, 2019), https://www.weforum.org/
`agenda/2019/04/how-much-data-is-generated-each-day-cf4bddf29f/.
`15 Darbinyan, The Growing Role of AI in Content Moderation,
`supra n.13.
`16 Id.
`17 Singhal et al., SoK: Content Moderation in Social Media,
`from Guidelines to Enforcement, and Research to Practice 2 (Mar.
`2023), https://arxiv.org/pdf/2206.14855.pdf.
`
`
`
`
`
`10
`
`(degrees of influence on the result), depending on the
`context and the human judgments they are implement-
`ing.18
`For example, algorithms implement decisions re-
`garding which content should be (1) featured promi-
`nently, (2) demoted so its distribution is reduced, (3) re-
`stricted to certain users based on their age, (4) flagged
`for further review by human moderators, (5) tagged
`with certain flags or warnings to other users, (6) imme-
`diately removed from the website, or (7) followed by a
`warning to the user who generated the content or his
`suspension or expulsion from the website.19 Each of
`these methods of moderation shapes the content availa-
`ble and accessible on a website, selecting and prioritizing
`the content most relevant to particular users and enforc-
`ing a website’s expressive value judgments around top-
`ics including safety, violence, nudity, or harassment.
`
`
`18 See Search, Google,
`https://www.google.com/search/
`howsearchworks/how-search-works/ranking-results/.
`19 Types of Content We Demote, Meta, https://transparency.
`fb.com/features/approach-to-ranking/types-of-content-we-demote
`(updated Oct. 16, 2023); Our Approach to Facebook Feed Ranking,
`Meta,
`https://transparency.fb.com/features/ranking-and-content/
`(updated Nov. 28, 2023); Meta’s Third-Party Fact-Checking Pro-
`gram, Meta, https://www.facebook.com/formedia/mjp/programs/
`third-party-fact-checking/ (visited Dec. 6, 2023); Facebook Commu-
`nity Standards, Meta Transparency Center, https://transpar-
`ency.fb.com/policies/community-standards/ (visited Dec. 6, 2023);
`Notices on X and what they mean, X Help Center, https://help.twit-
`ter.com/en/rules-and-policies/notices-on-x (visited Dec. 6, 2023);
`Singhal et al., SoK: Content Moderation in Social Media, supra
`n.17, at 3 (discussing “hard” versus “soft” moderation approaches).
`
`