[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-123454-en":3,"doc-seo-123454-105":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},123454,1374391974585,"Genevieve","https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c",8,"Research & Report","Algorithmic Collective Action in Machine Learning - Research Paper on ML Platform Control","A principled study examines how coordinated groups can perform algorithmic collective action on digital platforms that deploy machine learning. The work introduces a theoretical model where a collective pools participants’ data, issues instructions to modify individual data, and thereby steers the firm’s learning problem toward a collective objective. Consequences are analyzed across nonparametric optimal learning, parametric risk minimization, and gradient-based optimization, identifying success criteria tied to collective size. Extensive experiments on a gig-platform resume classification task using a BERT-like model confirm theoretical predictions.","arXiv :2302 .04262v3 [ cs .LG] 7 Aug 2024  \nAlgorithmic Collective Action in Machine Learning ∗ Moritz Hardt 1 , Eric Mazumdar2 , Celestine Mendler-Dünner 1 , and Tijana Zrnic3  \n1 Max Planck Institute for Intelligent Systems, Tübingen, and Tübingen AI Center  \n2 California Institute of Technology  \n3 University of California, Berkeley  \nAbstract  \nWe initiate a principled study of algorithmic collective action on digital platforms that deploy machine learning algorithms. We propose a simple theoretical model of a collective interacting with a firm’s learning algorithm. The collective pools the data of participating individuals and executes an algorithmic strategy by instructing participants how to modify their own data to achieve a collective goal. We investigate the consequences of this model in three fundamental learning-theoretic settings: nonparametric optimal learning, parametric risk minimization, and gradient-based optimization. In each setting, we come up with coordinated algorithmic strategies and characterize natural success criteria as a function of the collective’s size. Complementing our theory, we conduct systematic experiments on a skill classification task involving tens of thousands of resumes from a gig platform for freelancers. Through more than two thousand model training runs of a BERT-like language model, we see a striking correspondence emerge between our empirical observations and the predictions made by our theory. Taken together, our theory and experiments broadly support the conclusion that algorithmic collectives of exceedingly small fractional size can exert significant control over a platform’s learning algorithm.  \n1 Introduction  \nThroughout the gig economy, numerous digital platforms algorithmically profile, control, and discipline workers that offer on-demand services to consumers. Data collection and predictive modeling are critical for a typical platform’s business as machine learning algorithms power ranking, scoring, and classification tasks of various kinds [Woodcock and Graham, 2019 , Gray and Suri, 2019 , Schor, 2021] .  \nTroves of academic scholarship document the emergence and preponderance of precarity in the gig economy. Wood et al. [2019] argue that platform-based algorithmic control can lead to “low pay, social isolation, working unsocial and irregular hours, overwork, sleep deprivation and exhaustion.” This is further exacerbated by “high levels of inter-worker competition with few labor protections and a global oversupply of labor relative to demand.” In response, there have been numerous attempts by gig workers to organize in an effort to reconfigure working conditions. A growing repertoire of strategies, as vast as it is eclectic, uses both physical and digital means towards this goal. Indeed, workers have shown significant ingenuity in creating platform-specific infrastructure, such as their own mobile apps, to organize the labor side of the platform [Chen, 2018 , Rahman, 2021] . Yet,“the upsurge of worker mobilization should not blind us to the difficulties of organizing such a diverse and spatially dispersed labor force.” [Vallas and Schor, 2020]  \nBeyond the gig economy, evidence of consumers seeking to influence the algorithms that power a platform’s business is abundant. Examples include social media users attempting to suppress the algorithmic upvoting of harmful content by sharing screenshots rather than original posts [Burrell et al. , 2019], or individuals creating bots to influence crowd-sourced navigation systems [Sinai et al. , 2014] . The ubiquity of such strategic attempts calls for a principled study of how coordinated groups can wield control over the digital platforms to which they contribute data.  \nIn this work, we study how a collective of individuals can algorithmically strategize against a learning platform. We envision a collective that  pools the data of participating individuals and executes an  \n∗ Authors ordered alphabetically.  \nalgorithmic strate","cbCaibxwzXfnxuw0","https://ap.wps.com/l/cbCaibxwzXfnxuw0","pdf",872902,1,22,"English","en",105,"# Abstract\n# Introduction\n## Our contribution\n# Classification\n## Objective design for the collective\n# Learning-theoretic settings\n## Nonparametric optimal learning\n## Parametric risk minimization\n## Gradient-based optimization\n# Experiments","[{\"question\":\"What model does the document introduce for algorithmic collective action?\",\"answer\":\"It proposes a theoretical framework where a collective pools participants’ data and instructs individuals to modify their own data to redirect the firm’s learning toward a collective goal.\"},{\"question\":\"How is the collective’s influence quantified in the model?\",\"answer\":\"The collective size is represented by α, the fraction of participating individuals drawn from a base distribution, forming the mixture distribution the firm trains on.\"},{\"question\":\"Which learning scenarios are analyzed, and what is the main finding?\",\"answer\":\"The study analyzes nonparametric optimal learning, parametric risk minimization, and gradient-based optimization, showing coordinated strategies with success thresholds dependent on α; even vanishingly small fractional collectives can exert significant control.\"}]","Algorithmic Collective Action in Machine Learning - Research Paper on ML Platform Control | PDF",1785816610,55,{"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},"algorithmic-collective-action-in-machine-learning-research-paper-on-ml-platform-control","",{"@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/algorithmic-collective-action-in-machine-learning-research-paper-on-ml-platform-control/123454/",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","2026-08-04",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},"What model does the document introduce for algorithmic collective action?","Question",{"text":75,"@type":76},"It proposes a theoretical framework where a collective pools participants’ data and instructs individuals to modify their own data to redirect the firm’s learning toward a collective goal.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How is the collective’s influence quantified in the model?",{"text":80,"@type":76},"The collective size is represented by α, the fraction of participating individuals drawn from a base distribution, forming the mixture distribution the firm trains on.",{"name":82,"@type":73,"acceptedAnswer":83},"Which learning scenarios are analyzed, and what is the main finding?",{"text":84,"@type":76},"The study analyzes nonparametric optimal learning, parametric risk minimization, and gradient-based optimization, showing coordinated strategies with success thresholds dependent on α; even vanishingly small fractional collectives can exert significant control.","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"]