{"id":"https://openalex.org/W4396822621","doi":"https://doi.org/10.1145/3630106.3658968","title":"The Dark Side of Dataset Scaling: Evaluating Racial Classification in Multimodal Models","display_name":"The Dark Side of Dataset Scaling: Evaluating Racial Classification in Multimodal Models","publication_year":2024,"publication_date":"2024-06-03","ids":{"openalex":"https://openalex.org/W4396822621","doi":"https://doi.org/10.1145/3630106.3658968"},"language":"en","primary_location":{"id":"doi:10.1145/3630106.3658968","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3630106.3658968","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3630106.3658968","source":null,"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"The 2024 ACM Conference on Fairness, Accountability, and Transparency","raw_type":"proceedings-article"},"type":"preprint","indexed_in":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3630106.3658968","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5038758207","display_name":"Abeba Birhane","orcid":"https://orcid.org/0000-0001-6319-7937"},"institutions":[{"id":"https://openalex.org/I205274468","display_name":"Trinity College Dublin","ror":"https://ror.org/02tyrky19","country_code":"IE","type":"education","lineage":["https://openalex.org/I205274468"]}],"countries":["IE"],"is_corresponding":true,"raw_author_name":"Abeba Birhane","raw_affiliation_strings":["School of Computer Science and Statistics, Mozilla Foundation and Trinity College Dublin, Ireland"],"affiliations":[{"raw_affiliation_string":"School of Computer Science and Statistics, Mozilla Foundation and Trinity College Dublin, Ireland","institution_ids":["https://openalex.org/I205274468"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5056461554","display_name":"Sepehr Dehdashtian","orcid":"https://orcid.org/0000-0002-2512-8815"},"institutions":[{"id":"https://openalex.org/I87216513","display_name":"Michigan State University","ror":"https://ror.org/05hs6h993","country_code":"US","type":"education","lineage":["https://openalex.org/I87216513"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Sepehr Dehdashtian","raw_affiliation_strings":["Department of Computer Science and Engineering, Michigan State University, USA"],"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering, Michigan State University, USA","institution_ids":["https://openalex.org/I87216513"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5086218466","display_name":"Vinay Uday Prabhu","orcid":"https://orcid.org/0000-0001-9602-0134"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Vinay Prabhu","raw_affiliation_strings":["HAL51 Inc, USA"],"affiliations":[{"raw_affiliation_string":"HAL51 Inc, USA","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5031717929","display_name":"Vishnu Naresh Boddeti","orcid":"https://orcid.org/0000-0002-8918-9385"},"institutions":[{"id":"https://openalex.org/I87216513","display_name":"Michigan State University","ror":"https://ror.org/05hs6h993","country_code":"US","type":"education","lineage":["https://openalex.org/I87216513"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Vishnu Boddeti","raw_affiliation_strings":["Computer Science and Engineering, Michigan State University, USA"],"affiliations":[{"raw_affiliation_string":"Computer Science and Engineering, Michigan State University, USA","institution_ids":["https://openalex.org/I87216513"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5038758207"],"corresponding_institution_ids":["https://openalex.org/I205274468"],"apc_list":null,"apc_paid":null,"fwci":4.8609,"has_fulltext":true,"cited_by_count":20,"citation_normalized_percentile":{"value":0.96153001,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":90,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"1229","last_page":"1244"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11448","display_name":"Face recognition and analysis","score":0.9923999905586243,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11448","display_name":"Face recognition and analysis","score":0.9923999905586243,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T13877","display_name":"Law in Society and Culture","score":0.9901000261306763,"subfield":{"id":"https://openalex.org/subfields/3308","display_name":"Law"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.9768000245094299,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6629814505577087},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.5923361778259277},{"id":"https://openalex.org/keywords/offensive","display_name":"Offensive","score":0.5910055637359619},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.5344699025154114},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.5007302761077881},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.44542351365089417},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.11257457733154297},{"id":"https://openalex.org/keywords/geography","display_name":"Geography","score":0.09446179866790771}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6629814505577087},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.5923361778259277},{"id":"https://openalex.org/C176856949","wikidata":"https://www.wikidata.org/wiki/Q2001676","display_name":"Offensive","level":2,"score":0.5910055637359619},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5344699025154114},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.5007302761077881},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.44542351365089417},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.11257457733154297},{"id":"https://openalex.org/C205649164","wikidata":"https://www.wikidata.org/wiki/Q1071","display_name":"Geography","level":0,"score":0.09446179866790771},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.0},{"id":"https://openalex.org/C42475967","wikidata":"https://www.wikidata.org/wiki/Q194292","display_name":"Operations research","level":1,"score":0.0},{"id":"https://openalex.org/C166957645","wikidata":"https://www.wikidata.org/wiki/Q23498","display_name":"Archaeology","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1145/3630106.3658968","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3630106.3658968","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3630106.3658968","source":null,"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"The 2024 ACM Conference on Fairness, Accountability, and Transparency","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2405.04623","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2405.04623","pdf_url":"https://arxiv.org/pdf/2405.04623","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"}],"best_oa_location":{"id":"doi:10.1145/3630106.3658968","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3630106.3658968","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3630106.3658968","source":null,"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"The 2024 ACM Conference on Fairness, Accountability, and Transparency","raw_type":"proceedings-article"},"sustainable_development_goals":[{"score":0.6800000071525574,"id":"https://metadata.un.org/sdg/5","display_name":"Gender equality"}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":false},"content_urls":{"pdf":"https://content.openalex.org/works/W4396822621.pdf"},"referenced_works_count":53,"referenced_works":["https://openalex.org/W202799293","https://openalex.org/W603723718","https://openalex.org/W1810983823","https://openalex.org/W1978779947","https://openalex.org/W1992465642","https://openalex.org/W2280302689","https://openalex.org/W2343234953","https://openalex.org/W2593831809","https://openalex.org/W2604281640","https://openalex.org/W2760089875","https://openalex.org/W2787188261","https://openalex.org/W2790119928","https://openalex.org/W2795908329","https://openalex.org/W2891796941","https://openalex.org/W2899136066","https://openalex.org/W2899928646","https://openalex.org/W2911122801","https://openalex.org/W2948141579","https://openalex.org/W2957654274","https://openalex.org/W2963349562","https://openalex.org/W2972818416","https://openalex.org/W2989168403","https://openalex.org/W2996844929","https://openalex.org/W3007328579","https://openalex.org/W3029264758","https://openalex.org/W3091454361","https://openalex.org/W3091722278","https://openalex.org/W3095351420","https://openalex.org/W3119150429","https://openalex.org/W3119746452","https://openalex.org/W3134970617","https://openalex.org/W3155422163","https://openalex.org/W3155476465","https://openalex.org/W3163396516","https://openalex.org/W3174220540","https://openalex.org/W3184127157","https://openalex.org/W3184144760","https://openalex.org/W3186782927","https://openalex.org/W3199665477","https://openalex.org/W3203942433","https://openalex.org/W4226067429","https://openalex.org/W4283156811","https://openalex.org/W4283163904","https://openalex.org/W4293918026","https://openalex.org/W4297982512","https://openalex.org/W4308239184","https://openalex.org/W4385201870","https://openalex.org/W4386724905","https://openalex.org/W4394865455","https://openalex.org/W6755836617","https://openalex.org/W6801286000","https://openalex.org/W6804645036","https://openalex.org/W6912494966"],"related_works":["https://openalex.org/W2961085424","https://openalex.org/W4306674287","https://openalex.org/W3046775127","https://openalex.org/W3107602296","https://openalex.org/W4394896187","https://openalex.org/W3170094116","https://openalex.org/W4386462264","https://openalex.org/W4364306694","https://openalex.org/W4312192474","https://openalex.org/W4283697347"],"abstract_inverted_index":{"\u2018Scale":[0],"the":[1,4,7,11,15,44,53,61,77,90,102,108,115,119,142,155,162,185,195,201,206,229,240,273,280],"model,":[2],"scale":[3,6],"data,":[5],"GPU":[8],"farms\u2019":[9],"is":[10,40,52,66,187,231],"reigning":[12],"sentiment":[13],"in":[14,43,244,276],"world":[16],"of":[17,46,80,121,154,164,168,208,212,266],"generative":[18],"AI":[19],"today.":[20],"While":[21],"model":[22,35,125,241],"scaling":[23,29,82],"has":[24],"been":[25],"extensively":[26],"studied,":[27],"data":[28,117],"and":[30,58,92,98,137,172,181,216,225,247,254,261,269,279,294],"its":[31],"downstream":[32,78],"impacts":[33],"on":[34,83,89,251],"performance":[36],"remain":[37],"under-explored.":[38],"This":[39,289],"particularly":[41],"important":[42],"context":[45],"multimodal":[47],"datasets":[48,94],"whose":[49],"main":[50],"source":[51],"World":[54],"Wide":[55],"Web,":[56],"condensed":[57],"packaged":[59],"as":[60,107,114,129,134,145,150,176,220],"Common":[62],"Crawl":[63],"dump,":[64],"which":[65],"known":[67],"to":[68,191,235],"exhibit":[69],"numerous":[70],"drawbacks.":[71],"In":[72],"this":[73,277],"paper,":[74],"we":[75,160],"evaluate":[76],"impact":[79],"dataset":[81,186,230,258],"14":[84,156],"visio-linguistic":[85],"models":[86],"(VLMs)":[87],"trained":[88],"LAION400-M":[91],"LAION-2B":[93],"by":[95,179,223],"measuring":[96],"racial":[97],"gender":[99],"bias":[100],"using":[101],"Chicago":[103],"Face":[104],"Dataset":[105],"(CFD)":[106],"probe.":[109],"Our":[110],"results":[111,243],"show":[112],"that":[113],"training":[116],"increased,":[118],"probability":[120,163,207],"a":[122,169,173,213,217,245,264],"pre-trained":[123],"CLIP":[124],"misclassifying":[126,141],"human":[127,146],"images":[128,144],"offensive":[130,147,295],"non-human":[131],"classes":[132,148],"such":[133,149],"chimpanzee,":[135],"gorilla,":[136],"orangutan":[138],"decreased,":[139],"but":[140],"same":[143],"criminal":[151,177,221],"increased.":[152],"Furthermore,":[153],"Vision":[157],"Transformer-based":[158],"VLMs":[159],"evaluated,":[161],"predicting":[165,209],"an":[166,210],"image":[167,211],"Black":[170,214],"man":[171,175,215,219],"Latino":[174,218],"increases":[178],"65%":[180],"69%,":[182],"respectively,":[183,227],"when":[184,228],"scaled":[188,232],"from":[189,233],"400M":[190,234],"2B":[192,236],"samples":[193],"for":[194,200,257],"larger":[196],"ViT-L":[197],"models.":[198],"Conversely,":[199],"smaller":[202],"base":[203],"ViT-B":[204],"models,":[205],"decreases":[222],"20%":[224],"47%,":[226],"samples.":[237],"We":[238],"ground":[239],"audit":[242],"qualitative":[246],"historical":[248],"analysis,":[249],"reflect":[250],"our":[252],"findings":[253],"their":[255],"implications":[256],"curation":[259],"practice,":[260],"close":[262],"with":[263],"summary":[265],"mitigation":[267],"mechanisms":[268],"ways":[270],"forward.":[271],"All":[272],"meta-datasets":[274],"curated":[275],"endeavor":[278],"code":[281],"used":[282],"are":[283],"shared":[284],"at:":[285],"https://github.com/SepehrDehdashtian/the-dark-side-of-dataset-scaling.":[286],"Content":[287],"warning:":[288],"article":[290],"contains":[291],"racially":[292],"dehumanising":[293],"descriptions.":[296]},"counts_by_year":[{"year":2026,"cited_by_count":5},{"year":2025,"cited_by_count":14},{"year":2024,"cited_by_count":1}],"updated_date":"2026-04-14T08:04:32.555800","created_date":"2025-10-10T00:00:00"}