4.6.2 Sauya, tata, kuma Rage

Make your gwaji mafi m da maye gurbin gwaje-gwajen da wadanda ba na gwaji karatu, refining da jiyya, da kuma rage yawan mahalarta.

Na biyu yanki na shawara cewa Ina so in bayar game da zayyana digital gwajen shafi xa'a. Kamar yadda Restivo kuma van de Rijt gwaji a kan barnstars a Wikipedia shows, rage halin kaka yana nufin cewa xa'a zai zama da samun muhimmanci na bincike zane. Baya ga da'a frameworks shiryarwa mutum batutuwa bincike cewa zan bayyana a Babi na 6, masu bincike zayyana digital gwaje-gwajen kuma iya zana a kan da'a ideas daga daban-daban source: da da'a ka'idojin raya ya shiryar da gwaje-gwajen shafe dabbobi. Musamman ma, a cikin landmark littafin ka'idoji na m gwajin dabara, Russell and Burch (1959) samarwa uku ka'idojin da ya kamata shiryar da dabba bincike: Sauya, tata, kuma Rage. Ina so in ba da shawara cewa wadannan uku R ta kuma za a iya amfani da-in wani dan kadan modified form-to shiryar da zane na 'yan Adam gwaje-gwajen. Musamman,

• Sauya: Sauya gwaje-gwajen da kasa masu cin zali hanyoyin idan zai yiwu
• Sake tace: tata lura su sa shi a matsayin m kamar yadda zai yiwu
• Rage: Rage yawan mahalarta a cikin gwaji kamar yadda zai yiwu

Domin yin wadannan uku R ta kankare da nuna yadda za su iya yiwuwar kai ga mafi alhẽri kuma mafi m gwaji zane, Zan bayyana wani online filin gwajin da generated da'a muhawara. Sa'an nan zan bayyana yadda uku R ta bayar da shawarar kankare da m canje-canje ga zane na gwaji.

Daya daga cikin mafi ethically muhawara digital filin gwajen ne "Wani tunanin Contagion," wanda aka gudanar da Adam Kramer, Jamie Gillroy, kuma Jeffrey Hancock (2014) . The gwajin ya faru a kan Facebook, aka m da wani mix na kimiyya da m tambayoyi. A wannan lokacin, da rinjaye hanyar da masu amfani mu'amala da Facebook ya News Feed, an algorithmically curated sa na Facebook status updates daga wani mai amfani da Facebook friends. Wasu masu sukar na Facebook ya nuna cewa saboda News Feed yana kagaggun tabbatacce posts-abokai nuna kashe su latest jam'iyyar-shi zai iya sa masu amfani don jin bakin ciki saboda rayukansu ze m m a kwatanta. A daya hannun, watakila da sakamako ne daidai akasin. watakila ganin abokinka da ciwon mai kyau lokaci zai sa ka ji farin ciki? Domin magance wadannan gasar jarrabawa-da ci gaba da fahimtar yadda wani mutum motsin zuciyarmu suna tasiri da ta friends 'motsin zuciyarmu-Kramer da kuma abokan aiki gudu wani gwaji. The masu bincike sanya game da 700.000 masu amfani a cikin hudu kungiyoyin daya mako: a "negativity rage" kungiyar, ga wanda posts da mummunan magana (misali, m) da aka da ka katange daga bayyana News Feed, a "positivity rage" kungiyar ga wanda posts da m kalmomi (misali, m) da aka da ka katange. da biyu iko kungiyoyin. A kula da kungiyar ga "negativity rage" kungiyar, posts aka da ka katange a wannan kudi a matsayin "negativity rage" kungiyar amma ba tare da game da wani tunanin content. The iko kungiyar ga "positivity rage" kungiyar da aka gina a cikin wani layi daya fashion. The zane na wannan gwaji ya nuna cewa ya dace iko kungiyar ba ko da yaushe daya ba tare da wani canje-canje. Maimakon haka, wani lokacin kula da kungiyar sami wani magani domin ya haifar da daidai kwatanta cewa wani bincike tambaya na bukatar. A cikin dukkan lokuta, da posts da aka katange daga News Feed kasance har yanzu akwai ga masu amfani, ta hanyar wasu sassa na Facebook website.

Kramer da kuma abokan aiki gano cewa ga mahalarta a cikin positivity rage yanayin, da yawan m magana a status updates rage da yawan mummunan kalmomi karu. A daya hannun, domin mahalarta a cikin negativity rage yanayin, da yawan m kalmomi karu da kaso na korau kalmomi rage (Figure 4.23). Duk da haka, wadannan effects kasance quite kananan: bambanci a tabbatacce kuma korau kalmomi tsakanin jiyya da kuma controls ya game da 1 a 1,000 kalmomi.

Figure 4.23: Evidence na tunanin contagion (Kramer, Guillory, and Hancock 2014) . Kashi na tabbatacce kalmomi da mummunan magana da gwaji yanayin. Bars wakilci kiyasta misali kurakurai.

Na sa a tattaunawa na kimiyya al'amurran da wannan gwaji a kara karatu sashe a karshen sura, amma abin takaici, wannan gwaji ne mafi sani ga samar da da'a muhawara. Just kwanaki bayan wannan takarda da aka buga a gabatarwar na National Academy of Sciences, akwai wani babban outcry daga duka masu bincike da kuma latsa. Ƙeta doka a kusa da takarda mayar da hankali a kan biyu main maki: 1) mahalarta bai samar da wani amsa bayan da misali Facebook sharuddan-of-service a lura cewa wasu tunani zai sa cutar da mahalarta da kuma 2) da binciken ya ba halartar uku-jam'iyyar da'a review (Grimmelmann 2015) . The da'a tambayoyi tashe a cikin wannan muhawara ta haifar da mujallar da sauri buga wani m "Editorial magana da damuwa" game da xa'a da da'a review tsari ga bincike (Verma 2014) . A m shekaru, da gwajin ya ci gaba da zama tushen m muhawara da bambancin ra'ayi, kuma wannan sabani iya yi da niyya ba sakamakon tuki cikin inuwa wasu gwaje-gwajen da ake yi da kamfanonin (Meyer 2014) .

Ganin cewa baya game Wani tunanin Contagion, Ina yanzu so a nuna cewa 3 R ta iya bayar da shawarar kankare, m inganta for real karatu (abin da za ka iya kaina tunani a game da xa'a da wannan musamman gwaji). Na farko R ne Sauya: bincike ya nemi ya maye gurbin gwaje-gwajen da kasa masu cin zali da m dabaru, idan zai yiwu. Alal misali, maimakon a guje wani gwaji, da masu bincike sun iya maida su ba kome a halitta gwaji. Kamar yadda aka bayyana a Babi na 2, halitta gwajen ne yanayi inda wani abu ya faru a duniya wanda approximates bazuwar aiki na jiyya (misali, irin caca hukunci wanda za a tsara a cikin soja). A amfani da wata halitta gwaji ne, bincike ba dole ba ne ya sadar jiyya. muhalli ya aikata wancan a gare ku. A wasu kalmomin, tare da wata halitta gwaji, masu bincike dã ba su da ake bukata don gwaje-gwaje yi amfani da mutane News sadarwa.

A gaskiya, kusan bi-biye da wani tunanin Contagion gwaji, Coviello et al. (2014) da aka tozarta abin da za a iya kira wani tunanin Contagion halitta gwaji. Su m, wanda yayi amfani da wata dabara da ake kira instrumental canji, shi ne a bit wuya idan ka taba gani da shi da. Saboda haka, domin a bayyana dalilin da ya sa aka bukata, bari mu gina shi. Na farko ra'ayin cewa wasu masu bincike domin su yi nazarin tunanin contagion zai zama kwatanta your posts on days inda News Feed sosai m to your posts on days inda News Feed ne ƙwarai korau. Wannan m zai zama m idan makasudin shi kawai don hango ko hasashen wani tunanin abun ciki na your posts, amma wannan m matsala idan manufa shi ne nazarin causal sakamako na News Feed on your posts. Don ganin matsalar da wannan zane, yi la'akari da Thanksgiving. A Amurka, tabbatacce posts karu da korau posts plummet on Thanksgiving. Saboda haka, a kan Thanksgiving, masu bincike zai iya ganin cewa your News Feed sosai tabbatacce, kuma dõmin ku posted tabbatacce abubuwa da. Amma, your m posts iya fary ta Thanksgiving ba da abun ciki na your News Feed. A maimakon haka, domin a kimanta da causal sakamako masu bincike bukatar wani abu da ya canza da abun ciki na your News Feed, ba tare da kai tsaye canza your motsin zuciyarmu. Abin farin, akwai wani abu kamar cewa faruwa a duk tsawon lokacin: da weather.

Coviello da kuma abokan aiki gano cewa a ruwa rana a wani birnin za, a kan talakawan, rage rabo daga posts da cewa su ne m by game da 1 yawan maki da kuma ƙara da rabo daga posts da suke korau da game da 1 yawan maki. Sa'an nan kuma, Coviello da kuma abokan aiki maida su ba kome da wannan hujja don nazarin tunanin contagion ba tare da bukatar gwaje-gwaje amfani kowa News Feed. A ainihi abin da suka aikata shi ne ma'auni yadda za ka posts aka yi tasiri da yanayin a cikin birane inda your friends rayu. Don ganin dalilin da ya sa wannan ya sa hankali, tunanin cewa ka rayu a New York City da kuma kana da wani aboki wanda yake zaune a Seattle. Yanzu tunanin cewa wata rana ta fara Ana ruwan sama a Seattle. Wannan ruwan sama a Seattle ba zai kai tsaye shafi your yanayi, amma zai sa ka News Feed zama m tabbatacce kuma mafi korau saboda aboki posts. Saboda haka, da ruwan sama a Seattle da ka manipulates your News Feed. Kunna wannan diraya cikin wani m ilimin kididdiga hanya da wahala (da kuma ainihin m amfani da Coviello da kuma abokan aiki ne mai bit wadanda ba misali) don haka sai na yi sa a mafi cikakken tattaunawa a kara karatu sashe. Abu mafi muhimmanci a tuna game Coviello kuma abokin aiki ta m shi ne cewa shi ya sa su su yi nazarin tunanin contagion ba tare da bukatar gudanar da wani gwaji da zai iya yiwuwar cutar mahalarta, kuma yana iya zama haka al'amarin da cewa a sauran saituna za ka iya maye gurbin gwaje-gwajen da sauran dabaru.

Na biyu a cikin 3 Rs ne tata: bincike ya kamata nẽman tata su jiyya domin sa karami cũta yiwu. Alal misali, maimakon tarewa abun ciki wanda ya ko dai tabbatacce ko korau, da masu bincike sun iya kara abun ciki wanda ke tabbatacce ko korau. Wannan makon zane dã canza tunanin abun ciki na mahalarta News ciyarwa, amma zai yi jawabi daya daga cikin damuwa cewa masu sukar bayyana: cewa gwaje-gwajen da zai iya sa mahalarta su miss muhimmanci bayanai a cikin News Feed. Tare da zane amfani da Kramer da kuma abokan aiki, a sakon da yake da muhimmanci shi a matsayin wata ila za a katange matsayin daya, da yake ba. Duk da haka, tare da inganta zane, da saƙonnin da za a rasa muhallinsu zai zama mãsu m muhimmanci.

A karshe, na uku R ne Rage: bincike ya nemi a rage yawan mahalarta a cikin gwaji, idan zai yiwu. A baya, wannan rage faru halitta saboda m kudin analog gwaje-gwajen da aka high, wanda karfafa bincike don inganta su zane da kuma bincike. Duk da haka, a lokacin da akwai kome m kudin data, masu bincike ba su fuskanci wata cost tauyewa a kan girman da gwaji, kuma wannan yana da m don haifar da unnecessarily manyan gwaje-gwajen.

Alal misali, abokan aiki Kramer kuma zai iya yi amfani da pre-jiyya bayanai game da mahalarta-kamar pre-jiyya Posting hali-yin su analysis fi dacewa. More musamman, maimakon kwatanta da rabo daga m kalmomi a magani da kuma kula da yanayi, da kuma abokan aiki Kramer iya idan aka kwatanta da canji a cikin rabo daga m kalmomi tsakanin yanayi; wani m sau da yawa ake kira bambanci-in-bambance-bambance da kuma wanda aka hankali alaka da gauraye zane da na bayyana a baya a cikin sura ta (Figure 4.5). Wannan ne, ga kowane ɗan takara, da masu bincike zai iya halitta wani canji ci (post-jiyya hali - pre-jiyya hali) sa'an nan kuma idan aka kwatanta da canjin scores mahalarta a lura da kuma kula da yanayi. Wannan bambanci-in-bambance-bambance m ne mafi m ilimin kididdiga, wanda ke nufin cewa masu bincike za su iya cimma wannan ilimin kididdiga amincewa ta yin amfani da yawa karami samfurori. A wasu kalmomin, da ba zalunta mahalarta kamar "Widgets", masu bincike na iya sau da yawa sa more daidai kimomi.

Ba tare da ciwon da raw data yana da wuya a san daidai nawa ya fi dacewa wani bambanci-in-bambance-bambance m dã ya kasance a cikin wannan harka. Amma, Deng et al. (2013) ya ruwaito cewa, a uku online gwaje-gwajen a kan Bing search engine sun kasance iya rage sãɓã wa jũna da kimomi da game da 50%, da kuma irin sakamakon da aka ruwaito ga wasu gwaje-gwajen online at Netflix (Xie and Aurisset 2016) . Wannan 50% sãɓã wa jũna rage yana nufin cewa wani tunanin Contagion bincike zai kasance iya yanyanke samfurin a cikin rabin idan suka yi amfani da wani dan kadan daban-daban analysis hanyoyin. A wasu kalmomin, tare da wani kankanin canji a cikin bincike, 350.000 mutane yiwuwa an kare hannu a cikin gwaji.

A wannan aya ka iya yi mamaki dalilin da ya sa masu bincike ya kamata ya kula idan 350.000 mutane sun kasance a Wani tunanin Contagion unnecessarily. Akwai biyu musamman siffofin Wani tunanin Contagion cewa yin damuwa da matsanancin size dace, da kuma wadannan siffofin da ake yadawa mutane da yawa digital filin gwajen: 1) akwai rashin tabbas game da ko gwajin zai sa cutar da akalla wasu mahalarta da kuma 2) hallara ba da son rai. A gwaje-gwajen da wadannan biyu halaye alama bu mai kyau a ci gaba da gwaje-gwajen a matsayin karamin wuri.

A ƙarshe, da uku-R's Sauya, sake tace, da kuma rage-samar da ka'idojin da za su iya taimaka masu bincike gina xa'a a cikin gwaji kayayyaki. Hakika, kowane daga cikin wadannan m canje-canje ga Wani tunanin Contagion gabatar da cinikayya-offs. Alal misali, bayyananna daga halitta gwaje-gwajen ba ko da yaushe kamar yadda mai tsabta a matsayin shaida daga yi da ka gwaje-gwajen da gonakin zai kasance mafi logistically wuya a yi fiye da block. Saboda haka, manufar bayar da shawara wadannan canje-canje ba su biyu-tsammani yanke shawara da sauran masu bincike. Maimakon haka, shi ne domin ya nuna yadda uku R ta iya amfani da wani mai idon basira halin da ake ciki.