2.3.2.5 Algorithmically sunkuyar

Halayyar a samu data ba halitta, an kore ta injiniya a raga na tsarin.

Ko da yake mutane da yawa samu data kafofin ne wadanda ba mai amsawa saboda mutane ba su sani su data ana rubuta (Sashe 2.3.1.3), masu bincike ya kamata su yi la'akari da hali a cikin wadannan online tsarin ya zama "halitta abin da ke faruwa" ko "m." A gaskiya, da digital tsarin da rikodin hali sosai aikin injiniya zuwa sa takamaiman halayyar kamar danna kan talla ko Posting abun ciki. The hanyoyin da cewa a raga da tsarin zanen kaya iya gabatar da alamu cikin bayanai da aka kira algorithmic confounding. Algorithmic confounding ne in mun gwada unknown to zamantakewa masana kimiyya, amma shi ne mai babbar damuwa daga m data masana kimiyya. Kuma, sabanin wasu daga cikin sauran matsaloli tare da digital burbushi, algorithmic confounding ne sun fi mayar ganuwa.

A gwada m misali na algorithmic confounding shi ne gaskiya cewa on Facebook akwai wani anomalously high yawan masu amfani da kamar 20 friends (Ugander et al. 2011) . Masana kimiyya da nazarin da wannan labari ba tare da wani fahimtar yadda Facebook aiki zai iya doubtlessly samar da dama da labaru game da yadda 20 ne wasu irin sihiri zamantakewa lamba. Duk da haka, Ugander da abokan aiki da wani gwaji fahimtar da tsari da generated da bayanai, da kuma sun san cewa Facebook karfafa mutane da 'yan sadarwa on Facebook yi more friends, har suka isa 20 friends. Ko da yake Ugander da kuma abokan aiki kada ku ce da wannan a cikin takarda, wannan manufar da aka yiwuwa halitta Facebook domin karfafa sabon masu amfani don zama mafi m. Ba tare da sanin game da wanzuwar wannan siyasa, duk da haka, shi ne mai sauki su kusantar da ba daidai ba ƙarshe daga bayanai. A wasu kalmomin, da mamaki high yawan mutanen da game da 20 abokai gaya mana game da Facebook fiye da mutum hali.

More pernicious daga wannan previous misali inda algorithmic confounding samar da quirky sakamakon cewa a hankali masu bincike zai gudanar da bincike kara, akwai wani ko trickier version of algorithmic confounding cewa faruwa a lokacin designers na online tsarin suna sane da zamantakewa theories, sa'an nan kuma gasa wadannan theories cikin aiki da tsarin. Social masana kimiyya kira wannan performativity: a lõkacin da theories canja duniya a cikin irin wannan hanyar da suka zo duniya more cikin layi tare da ka'idar. A cikin lokuta na performative algorithmic confounding, da tozarta yanayin da data ne m ganuwa.

Daya misali da wani abin kwaikwaya halitta performativity ne transitivity a online social networks. A cikin 1970s kuma 1980s, masu bincike akai-akai gano cewa, idan ka kasance abokai da Alice da kake abokai da Bob, to, Bob da Alice ne mafi kusantar su zama abokai da juna fiye da biyu da ka zaba mutane. Kuma, wannan sosai guda juna da aka samu a cikin zamantakewa jadawali on Facebook (Ugander et al. 2011) . Saboda haka, wanda zai kammala da cewa alamu na aminci on Facebook rubanya alamu na offline abota, a kalla a cikin sharuddan transitivity. Duk da haka, da girma da transitivity a Facebook social jadawali ne partially kore ta algorithmic confounding. Wancan ne, data masana kimiyya a Facebook san na empirical da ka'idojin bincike game transitivity sa'an nan kuma gasa shi a cikin yadda Facebook aiki. Facebook yana da "People ka iya sani" alama cewa da shawara new friends, kuma daya hanyar da Facebook yanke shawarar wanda ya bayar da shawarar a gare ku ne transitivity. Wancan ne, Facebook ne mafi kusantar su bayar da shawarar cewa ka zama abokai tare da abokai na your friends. Wannan alama ta haka ne yana da sakamako na kara transitivity a Facebook social jadawali. a cikin wasu kalmomi, ka'idar transitivity kawo duniya cikin layi tare da tsinkaya ka'idar (Healy 2015) . Saboda haka, a lokacin da babban data kafofin bayyana haifa tsinkaya zamantakewa ka'idar, dole ne mu tabbata cewa ka'idar kanta ba gasa a cikin yadda tsarin aiki.

Maimakon tunanin babban data kafofin yadda lura mutane a wata halitta wuri, a more dace misãli aka lura mutane a cikin wani gidan caca. Gidajen caca sosai aikin injiniya muhallin tsara don sa wani halayyar, da kuma masu bincike za su taba sa ran cewa hali a cikin wani gidan caca zai samar da wani unfettered taga a cikin mutum hali. Hakika, za mu iya koyon wani abu game da mutum hali nazarin mutane a gidajen caca-in gaskiya a gidan caca zai yi wani manufa wuri domin nazarin dangantaka tsakanin barasa amfani da kuma hadarin da zaɓin-amma idan muka yi watsi da cewa data ake halitta a cikin wani gidan caca mu cikakken mulki zana wasu miyagun karshe.

Abin baƙin ciki, da ake rubutu algorithmic confounding ne musamman wuya, domin da yawa siffofin online tsarin ne mallakar tajirai, talauci rubuce, kuma kullum canja. Alal misali, kamar yadda zan bayyana daga baya a wannan babi, algorithmic confounding kasance daya m bayani ga hankali hutu-saukar da Google Mura Trends (Sashe 2.4.2), amma wannan da'awar shi ne da wuya a tantance saboda ciki gudanar na Google search algorithm ne mallakar tajirai. The tsauri yanayin algorithmic confounding yana daya nau'i na tsarin yin gantali. Algorithmic confounding yana nufin cewa ya kamata mu kasance m game da duk wani da'awar ga mutum hali da ya zo daga rai guda digital tsarin, ko ta yaya babba.