[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-120731-en":3,"doc-seo-120731-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},120731,549758146520,"Patrick","https://ap-avatar.wpscdn.com/avatar/80002397d8c0411e94?_k=1775819394049821470",8,"Research & Report","Machines Do Not Decide Hate Speech - Machine learning, power, and the intersectional approach","The chapter examines automated hate speech detection in the context of social media, where scale is achieved through machine learning while fairness remains unresolved. It argues that machines do not determine what counts as hate speech because hate speech is embedded in societal norms shaped by power relations. By linking power dynamics to data generation and annotation, the chapter shows how intersectional analysis reveals bias in datasets used to train detection systems, and why ignoring it prevents equitable automation.","[www.ssoar. info](www.ssoar. info)  \nMachines do not decide hate speech: Machine learning, power, and the intersectional approach  \nKim , Jae Yeon  \nErstveröffentlichung / Primary Publication Sammelwerksbeitrag / collection article  \nEmpfohlene Zitierung / Suggested Citation:  \nKim , J. Y. (2023) . Machines do not decide hate speech: Machine learning , power, and the intersectional approach. In C. Strippel , S. Paasch-Colberg , M. Emmer, & J. Trebbe (Eds.) , Challenges and perspectives of hate speech research (pp. 355-369). Berlin [https://doi.org/10.48541/dcr.v12.21](https://doi.org/10.48541/dcr.v12.21)  \nNutzungsbedingungen:  \nDieser Text wird unter einer CC BY Lizenz (Namensnennung) zur Verfügung gestellt. Nähere Auskünfte zu den CC-Lizenzen finden Sie hier:  \n[https://creativecommons.org/licenses/by/4.0/deed.de](https://creativecommons.org/licenses/by/4.0/deed.de)  \nTerms of use:  \nThis document is made available under a CC BY Licence (Attribution). For more Information see:  \n[https://creativecommons.org/licenses/by/4.0](https://creativecommons.org/licenses/by/4.0)  \nDigital Communication [Research.de](Research.de)  \nRecommended citation: Kim, J. Y. (2023) . Machines do not decide hate speech: Machine learning, power, and the intersectional approach. In C. Strippel, S. Paasch-Colberg, M. Emmer, & J. Trebbe (Eds.), Challenges and perspectives of hate speech research (pp. 355–369). Digital Communication Research. [https://doi.org/10.48541/dcr.v12.21](https://doi.org/10.48541/dcr.v12.21)  \nAbstract: The advent of social media has increased digital content—and, with it, hate speech. Advancements in machine learning help detect online hate speech at scale, but scale is only one part of the problem related to moderating it. Machines do not decide what comprises hate speech, which is part of a societal norm. Power relations establish such norms and, thus, determine who can say what comprises hate speech. Without considering this data-generation process, a fair automated hate speech detection system cannot be built. This chapter first examines the relationship between power, hate speech, and machine learning. Then, it examines how the intersectional lens—focusing on power dynamics between and within social groups—helps identify bias in the data sets used to build automated hate speech detection systems.  \nLicense: Creative Commons Attribution 4.0 (CC-BY 4.0)  \nDOI 10.48541/dcr.v12 .21  \nJae Yeon Kim  \nMachines Do Not Decide Hate Speech Machine learning, power, and the intersectional approach1  \n1 Introduction  \nThe advent of social media platforms—such as Twitter, Facebook, and YouTube—has increased digital content. Alongside this change, hate speech—defined as highly negative and often violent speech that targets historically disadvantaged groups (Walker, 1994; Jacobs & Potter, 1998; see also Sponholz in this volume)– has also increased. In response, social media platforms have leveraged machine learning to scale up their efforts to detect and moderate users’ content (Gitari et al., 2015; Agrawal & Awekar, 2018; Watanabe et al., 2018; Koushik et al., 2019; see also Ahmad in this volume). Developing a system that relies less on human inspection and validation is desirable for these firms because this system’s efficiency gains would allow them to grow further and increase profits.  \nUnfortunately, scale is only part of the problem related to hate speech detection and moderation. Marginalized groups and individuals (e.g., ethnic and racial minorities, women, lesbian, gay, bisexual, transgender, and queer  \n1 I thank Thomas R. Davidson, Renata Barreto, two anonymous reviewers, and the editors ofthis volume for their constructive comments on an earlier draft of this chapter.  \nJ. Y. Kim  \n[LGBTQ] people, immigrants, and people with disabilities) are major targets of hate speech, which is one reason why many social media platforms cite potential harm against marginalized people as the main reason to target hate speech (Twitter, 2021; ","cbCaipQOpViWei8o","https://ap.wps.com/l/cbCaipQOpViWei8o","pdf",465973,1,17,"English","en",105,"# Introduction\n## Hate speech and the push for scalable machine learning moderation\n## Why “scale” does not guarantee fairness\n## Power relations as a driver of what counts as hate speech\n## Intersectional analysis and bias in annotation data","[{\"question\":\"Why does scaling up hate speech detection with machine learning not solve fairness problems?\",\"answer\":\"Because what counts as hate speech depends on social norms rather than being determined by technology alone. Power relations shape those norms and affect how speech is labeled.\"},{\"question\":\"How do power relations influence hate speech annotation and automated detection?\",\"answer\":\"Changes in power relations over time and across contexts determine who can say what constitutes hate speech. Since annotation relies on human judgments about norm violations, automated systems inherit these context-dependent patterns.\"},{\"question\":\"How does the intersectional lens help identify bias in hate speech detection systems?\",\"answer\":\"Intersectionality focuses on power dynamics between and within social groups, making it easier to detect systematic biases in the datasets used for training and evaluating automated detection models.\"}]","Machines Do Not Decide Hate Speech - Machine learning, power, and the intersectional approach | PDF",1785731748,43,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"machines-do-not-decide-hate-speech-machine-learning-power-and-the-intersectional-approach","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/machines-do-not-decide-hate-speech-machine-learning-power-and-the-intersectional-approach/120731/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why does scaling up hate speech detection with machine learning not solve fairness problems?","Question",{"text":75,"@type":76},"Because what counts as hate speech depends on social norms rather than being determined by technology alone. Power relations shape those norms and affect how speech is labeled.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How do power relations influence hate speech annotation and automated detection?",{"text":80,"@type":76},"Changes in power relations over time and across contexts determine who can say what constitutes hate speech. Since annotation relies on human judgments about norm violations, automated systems inherit these context-dependent patterns.",{"name":82,"@type":73,"acceptedAnswer":83},"How does the intersectional lens help identify bias in hate speech detection systems?",{"text":84,"@type":76},"Intersectionality focuses on power dynamics between and within social groups, making it easier to detect systematic biases in the datasets used for training and evaluating automated detection models.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]