{"id":4543,"date":"2025-11-22T07:57:12","date_gmt":"2025-11-22T07:57:12","guid":{"rendered":"https:\/\/academicsolidarity.com\/?p=4543"},"modified":"2025-11-22T08:05:29","modified_gmt":"2025-11-22T08:05:29","slug":"yapay-zeka-geri-cekilmis-makaleleri-ayirt-edemiyor","status":"publish","type":"post","link":"https:\/\/academicsolidarity.com\/?p=4543&lang=tr","title":{"rendered":"Yapay Zek\u00e2 Geri \u00c7ekilmi\u015f Makaleleri Ay\u0131rt Edemiyor"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">Bilimsel literat\u00fcrde <strong>geri \u00e7ekilmi\u015f makaleler<\/strong>, ara\u015ft\u0131rma b\u00fct\u00fcnl\u00fc\u011f\u00fcn\u00fc korumak i\u00e7in kullan\u0131lan en sert ve g\u00f6r\u00fcn\u00fcr uyar\u0131 i\u015faretleridir. Ancak Retraction Watch\u2019ta 19 Kas\u0131m 2025\u2019te yay\u0131mlanan yeni bir \u00e7al\u0131\u015fma, h\u0131zla yayg\u0131nla\u015fan yapay zek\u00e2 sohbet botlar\u0131n\u0131n bu kritik uyar\u0131 i\u015faretlerini tan\u0131makta son derece zorland\u0131\u011f\u0131n\u0131 g\u00f6steriyor. Ara\u015ft\u0131rmac\u0131lar, \u00f6zellikle ChatGPT ve benzeri ara\u00e7lara y\u00f6nelen akademisyenlerin, bu modellerin verdi\u011fi yan\u0131tlar\u0131 \u201cotomatik do\u011fruluk filtresi\u201d gibi kullanmalar\u0131 h\u00e2linde ciddi hatalara davetiye \u00e7\u0131kard\u0131klar\u0131 konusunda uyar\u0131yor (<a href=\"https:\/\/retractionwatch.com\/2025\/11\/19\/ai-unreliable-identifying-retracted-research-papers-study\/\">https:\/\/retractionwatch.com\/2025\/11\/19\/ai-unreliable-identifying-retracted-research-papers-study\/<\/a>).<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u00c7al\u0131\u015fmay\u0131 y\u00fcr\u00fcten Campinas Eyalet \u00dcniversitesi\u2019nden Konradin Metze ve ekibi, asl\u0131nda olduk\u00e7a basit bir deney tasarlad\u0131. Anesteziyoloji alan\u0131ndaki b\u00fcy\u00fck bilimsel sahtecilik skandal\u0131yla bilinen Joachim Boldt\u2019un yay\u0131nlar\u0131ndan olu\u015fan bir listeyi 21 farkl\u0131 yapay zekaya sundular. Listenin i\u00e7inde en \u00e7ok at\u0131f alan geri \u00e7ekilmi\u015f Boldt makaleleri, yine en \u00e7ok at\u0131f alan ama geri \u00e7ekilmemi\u015f Boldt yay\u0131nlar\u0131 ve ayr\u0131ca soyad\u0131 Boldt olan ba\u015fka yazarlar\u0131n yazd\u0131\u011f\u0131 makaleler yer al\u0131yordu. Toplam 132 referans\u0131n her biri i\u00e7in botlardan tek bir \u015fey istenmi\u015fti: Bu makale geri \u00e7ekildi mi, \u00e7ekilmedi mi?<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Sonu\u00e7 \u00e7arp\u0131c\u0131yd\u0131. Sohbet botlar\u0131n\u0131n \u00e7o\u011fu, geri \u00e7ekilmi\u015f makalelerin yar\u0131s\u0131ndan az\u0131n\u0131 do\u011fru olarak tan\u0131mlad\u0131. \u00dcstelik yaln\u0131zca \u201cka\u00e7\u0131rmakla\u201d kalmad\u0131lar; geri \u00e7ekilmemi\u015f makalelerin de hat\u0131r\u0131 say\u0131l\u0131r bir b\u00f6l\u00fcm\u00fcn\u00fc yanl\u0131\u015fl\u0131kla geri \u00e7ekilmi\u015f gibi i\u015faretlediler. Bu, hem duyarl\u0131l\u0131k hem de \u00f6zg\u00fcll\u00fck bak\u0131m\u0131ndan ciddi bir zay\u0131fl\u0131k anlam\u0131na geliyor: Yapay zek\u00e2, hem yanl\u0131\u015f g\u00fcvence veriyor hem de sa\u011flam makalelere gereksiz \u015f\u00fcphe d\u00fc\u015f\u00fcr\u00fcyor.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Ara\u015ft\u0131rma ekibi \u00fc\u00e7 ay sonra deneyin bir k\u0131sm\u0131n\u0131 tekrarlad\u0131\u011f\u0131nda daha da ilgin\u00e7 bir tabloyla kar\u015f\u0131la\u015ft\u0131. \u0130lk turda botlar genellikle kesin ifadeler kullan\u0131rken, ikinci turda \u201cmuhtemelen geri \u00e7ekilmi\u015f olabilir\u201d veya \u201cdaha fazla inceleme gerektiriyor\u201d gibi mu\u011flak ve ka\u00e7amak c\u00fcmleler kurmaya ba\u015flad\u0131lar. Ara\u015ft\u0131rmac\u0131lar bu de\u011fi\u015fimi, modellerin \u201cyanl\u0131\u015f bir kesinlik sunmak\u201d ile \u201cbelirsiz ifadelerle kendini kurtarmaya \u00e7al\u0131\u015fmak\u201d aras\u0131nda gidip geldi\u011fi \u015feklinde yorumluyor.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Retraction Watch haberinde, Sheffield \u00dcniversitesi\u2019nden Mike Thelwall\u2019\u0131n k\u0131sa s\u00fcre \u00f6nce yay\u0131mlad\u0131\u011f\u0131 ba\u015fka bir \u00e7al\u0131\u015fma da hat\u0131rlat\u0131l\u0131yor. Thelwall, geri \u00e7ekilmi\u015f ya da hakk\u0131nda ciddi \u015f\u00fcpheler bulunan 217 makaleyi ChatGPT\u2019ye toplam 6510 kez de\u011ferlendirdi. Bu binlerce cevab\u0131n hi\u00e7birinde, ChatGPT makalenin geri \u00e7ekildi\u011fini, hakk\u0131nda soru i\u015fareti oldu\u011funu ya da bilimsel sorun i\u00e7erdi\u011fini belirtmedi. Aksine, baz\u0131 geri \u00e7ekilmi\u015f makaleleri \u201cy\u00fcksek kaliteli \u00e7al\u0131\u015fma\u201d olarak \u00f6vd\u00fc\u011f\u00fc bile g\u00f6r\u00fcld\u00fc. Bu durum, yapay zek\u00e2n\u0131n yaln\u0131zca retraction bilgisini ka\u00e7\u0131rmakla kalmad\u0131\u011f\u0131n\u0131, ayn\u0131 zamanda hatal\u0131 veya sahte bilimsel bulgular\u0131 \u00f6vg\u00fcyle yeniden \u00fcretebildi\u011fini de g\u00f6steriyor (<a href=\"https:\/\/sheffield.ac.uk\/ijc\/news\/new-research-suggests-chatgpt-ignores-article-retractions-and-errors-when-used-inform-literature?utm_source=chatgpt.com\">https:\/\/sheffield.ac.uk\/ijc\/news\/new-research-suggests-chatgpt-ignores-article-retractions-and-errors-when-used-inform-literature?utm_source=chatgpt.com<\/a>).<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Sorun yaln\u0131zca tan\u0131mada de\u011fil. Journal of Advanced Research\u2019ta yay\u0131mlanan bir ba\u015fka \u00e7al\u0131\u015fma, sohbet botlar\u0131n\u0131n verdi\u011fi cevaplarda geri \u00e7ekilmi\u015f makaleleri kaynak olarak kulland\u0131\u011f\u0131n\u0131 ortaya koydu. Bu da yapay zek\u00e2n\u0131n, bilimsel literat\u00fcrde art\u0131k ge\u00e7ersiz say\u0131lan bilgileri yeniden dola\u015f\u0131ma sokabildi\u011fi anlam\u0131na geliyor. Akademik d\u00fcnyada gittik\u00e7e daha fazla ki\u015fi ChatGPT gibi ara\u00e7lar\u0131 h\u0131zl\u0131 \u00f6zet \u00e7\u0131karmak, ara\u015ft\u0131rma fikri geli\u015ftirmek veya literat\u00fcre h\u00e2kim olmak i\u00e7in kullan\u0131rken, geri \u00e7ekilmi\u015f bilgilerin yeniden dola\u015f\u0131ma girmesi giderek b\u00fcy\u00fcyen bir risk haline geliyor.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Bilim sosyolo\u011fu Serge Horbach, bu geli\u015fmeleri \u201ca\u00e7\u0131k bir uyar\u0131\u201d olarak nitelendiriyor: LLM modelleri, geri \u00e7ekilmi\u015f makaleleri ay\u0131klamak i\u00e7in uygun ara\u00e7lar de\u011fil. Yapay zek\u00e2 modellerinin e\u011fitim verisi hem tarihsel olarak gecikmeli hem de retraction bilgilerinin da\u011f\u0131n\u0131k bi\u00e7imde yay\u0131mland\u0131\u011f\u0131 bir sistemden besleniyor. Bir makalenin geri \u00e7ekildi\u011fine dair bilgi yaln\u0131zca dergi sayfas\u0131nda, yaln\u0131zca PubMed\u2019de ya da yaln\u0131zca Retraction Watch veri taban\u0131nda g\u00f6r\u00fcn\u00fcr olabiliyor. Bu par\u00e7al\u0131 yap\u0131y\u0131 g\u00fcvenlikle ve do\u011frulukla taramak, bug\u00fcnk\u00fc sohbet botlar\u0131n\u0131n teknik kapasitesinin olduk\u00e7a \u00f6tesinde.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Academic Solidarity a\u00e7\u0131s\u0131ndan bu bulgular, \u00f6zellikle s\u00fcrg\u00fcndeki veya g\u00fcvencesiz ko\u015fullarda \u00e7al\u0131\u015fan akademisyenler i\u00e7in \u00f6zel bir anlam ta\u015f\u0131yor. Ara\u015ft\u0131rma altyap\u0131s\u0131na eri\u015fimin s\u0131n\u0131rl\u0131 oldu\u011fu durumlarda ChatGPT gibi ara\u00e7lar cazip bir h\u0131z ve kolayl\u0131k sunuyor. Ancak bu kolayl\u0131k, geri \u00e7ekilmi\u015f veya hatal\u0131 bilgilere dayal\u0131 \u00e7al\u0131\u015fmalar\u0131n fark edilmeden yeniden \u00fcretilmesi riskini beraberinde getiriyor. Politik, hukuki veya insan haklar\u0131 alanlar\u0131nda \u00e7al\u0131\u015fan ara\u015ft\u0131rmac\u0131lar i\u00e7in bu risk daha da a\u011f\u0131r olabilir; yanl\u0131\u015f bilgi yaln\u0131zca bilimsel bir hata de\u011fil, ayn\u0131 zamanda politik bir manip\u00fclasyonun kap\u0131s\u0131n\u0131 da aralayabilir.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Bu tablo, yapay zek\u00e2n\u0131n ara\u015ft\u0131rma s\u00fcre\u00e7lerinde tamamen d\u0131\u015flanmas\u0131n\u0131 gerektirmiyor; ancak kritik bir s\u0131n\u0131r\u0131 hat\u0131rlat\u0131yor: ChatGPT ve benzeri modeller, geri \u00e7ekilmi\u015f literat\u00fcr\u00fc tespit etmek i\u00e7in g\u00fcvenilir bir filtre de\u011fil. Bu ara\u00e7lar en fazla not tutmaya, metni sadele\u015ftirmeye, tart\u0131\u015fma fikri \u00fcretmeye yard\u0131mc\u0131 olabilir. Fakat bir makalenin ger\u00e7ekten geri \u00e7ekilip \u00e7ekilmedi\u011fine karar verme i\u015fi, g\u00fcn\u00fcm\u00fczde halen insan ara\u015ft\u0131rmac\u0131n\u0131n sorumlulu\u011funda olmal\u0131. Bilimsel b\u00fct\u00fcnl\u00fc\u011f\u00fcn as\u0131l y\u00fck\u00fcn\u00fc ta\u015f\u0131yan da bu sorumluluk oluyor.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Bilimsel literat\u00fcrde geri \u00e7ekilmi\u015f makaleler, ara\u015ft\u0131rma b\u00fct\u00fcnl\u00fc\u011f\u00fcn\u00fc korumak i\u00e7in kullan\u0131lan en sert ve g\u00f6r\u00fcn\u00fcr uyar\u0131 i\u015faretleridir. Ancak Retraction Watch\u2019ta 19 Kas\u0131m 2025\u2019te yay\u0131mlanan yeni bir \u00e7al\u0131\u015fma, h\u0131zla yayg\u0131nla\u015fan yapay zek\u00e2 sohbet botlar\u0131n\u0131n bu kritik uyar\u0131 i\u015faretlerini tan\u0131makta son derece zorland\u0131\u011f\u0131n\u0131 g\u00f6steriyor. Ara\u015ft\u0131rmac\u0131lar, \u00f6zellikle ChatGPT ve benzeri ara\u00e7lara y\u00f6nelen akademisyenlerin, bu modellerin verdi\u011fi yan\u0131tlar\u0131 \u201cotomatik do\u011fruluk [&hellip;]<\/p>\n","protected":false},"author":3,"featured_media":4548,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[114],"tags":[],"class_list":["post-4543","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-guncel"],"jetpack_featured_media_url":"https:\/\/i0.wp.com\/academicsolidarity.com\/wp-content\/uploads\/2025\/11\/20251122ChatGPT-2.jpg?fit=831%2C423&ssl=1","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/academicsolidarity.com\/index.php?rest_route=\/wp\/v2\/posts\/4543","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/academicsolidarity.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/academicsolidarity.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/academicsolidarity.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/academicsolidarity.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=4543"}],"version-history":[{"count":1,"href":"https:\/\/academicsolidarity.com\/index.php?rest_route=\/wp\/v2\/posts\/4543\/revisions"}],"predecessor-version":[{"id":4544,"href":"https:\/\/academicsolidarity.com\/index.php?rest_route=\/wp\/v2\/posts\/4543\/revisions\/4544"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/academicsolidarity.com\/index.php?rest_route=\/wp\/v2\/media\/4548"}],"wp:attachment":[{"href":"https:\/\/academicsolidarity.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4543"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/academicsolidarity.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4543"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/academicsolidarity.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4543"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}