4.2 Mene ne gwajen?

Yi da ka sarrafawa gwaje-gwajen da hudu main sinadaran: daukar ma'aikata na mahalarta, randomization magani, bayarwa na magani, da kuma ji na sakamakon.

Yi da ka sarrafawa gwajen iya dauka da yawa siffofin da za a iya amfani da su yi nazarin da yawa iri hali. Amma, a da core, yi da ka sarrafawa gwaje-gwajen da hudu main sinadaran: daukar ma'aikata na mahalarta, randomization magani, bayarwa na magani, da kuma ji na sakamakon. The digital shekaru ba ya canja da muhimman hakkokin yanayin experimentation, amma yana yi musu sauki logistically. Alal misali, a baya shi zai kasance da wuya a auna hali na miliyoyin mutane, amma da aka yanzu routinely faruwa a yawancin digital tsarin. Masu bincike wanda zai iya gane yadda za a cin gajiyar wadannan sabon damar da zai iya gudu gwaje-gwajen da suke ba zai yiwu ba a baya.

Don yin wannan dukan wani bit more kankare-biyu abin da ya zauna da wannan kuma abin da ya canza-bari mu yi la'akari da Michael Restivo da Arnout van de Rijt ta (2012) . The masu bincike so ya fahimci sakamakon na yau da kullum tsara lada a kan Editorial gudunmawar Wikipedia. Musamman ma, sun karanta sakamakon barnstars, an kyautar da cewa duk wani Wikipedian iya ba da duk wani Wikipedian amince da aiki da kuma saboda himma. Restivo kuma van de Rijt ba barnstars zuwa 100 cancanci Wikipedians. Sa'an nan kuma, Restivo kuma van de Rijt sa ido da masu karɓa 'm gudunmawar zuwa Wikipedia kan gaba 90 days. Mafi yawan su mamaki, mutane ga wanda suka bayar barnstars kula yi m gyararrakin bayan karbar daya. A wasu kalmomin, da barnstars alama da za a karaya maimakon ƙarfafa taimako.

Abin farin, Restivo kuma van de Rijt aka ba gudu a "perturb kuma suka tsayar" gwaji. da suka kasance sunã gudãna a yi da ka sarrafawa gwaji. Saboda haka, ban da zabar 100 top bayar da gudunmawa ga sama a barnstar, su ma tsince 100 top bayar da gudunmawa ga wanda ba su ba a barnstar. Wadannan da ɗari yi aiki a matsayin mai kula da kungiyar, da kuma wanda ya samu a barnstar da kuma wanda bai da aka ƙaddara da ka. A lokacin da Restivo kuma van de Rijt dube kula da kungiyar sun gano cewa, shi yana da m drop a gudunmawar ma. A karshe, a lõkacin da bincike idan aka kwatanta da mutane a lura kungiyar (ie, karbi barnstars) da kuma mutanen da a kula da kungiyar, sun gano cewa, barnstar sa gyara don taimakawa game da 60% more. Amma, wannan karuwa a taimako da aka faruwa kamar yadda wani ɓangare na wani overall dakushe biyu kungiyoyin.

Kamar yadda wannan bincike ya nuna, da iko kungiyar a gwaje-gwajen da muhimmanci a hanyar da yake da ɗan paradoxical. Domin daidai auna sakamako na barnstars, Restivo da van der Rijt da ake bukata a kalla mutane cewa bai sami barnstars. Mutane da yawa sau bincike suke ba su saba da gwaje-gwajen kasa godiya da m darajar da iko kungiyar. Idan Restivo kuma van de Rijt basu da iko kungiyar, dã sun kõma daidai da ba daidai ba ƙarshe. Control kungiyoyi ne don haka da muhimmanci cewa Shugaba na manyan gidan caca kamfanin ya bayyana cewa, akwai kawai hanyoyi uku da ma'aikata za a iya kora daga company: sata, jima'i dama, da kuma a guje wani gwaji, ba tare da wani iko kungiyar (Schrage 2011) .

Restivo kuma van de Rijt ta binciken ya nuna hudu main sinadaran wani gwaji: daukar ma'aikata, randomization, baki, da kuma sakamakon. Tare, waɗannan hudu sinadaran damar masana kimiyya don motsawa bayan huldodi da auna causal sakamako na jiyya. Musamman, randomization yana nufin cewa a lokacin da ka kwatanta sakamakon domin lura da kuma kula da kungiyoyin da ka samu wani kimanta daga cikin causal sakamako na cewa baki ga wanda ya kafa mahalarta. A wasu kalmomin, tare da yi da ka sarrafawa gwaji za ka iya tabbata da cewa duk wani bambance-bambance a sakamakon suna lalacewa ta hanyar da baki da kuma ba a confounder, da'awar cewa na yi daidai a cikin Technical Shafi ta yin amfani da m sakamakon tsarin.

Bugu da ƙari, kasancewa a nice zane na makanikai da gwaje-gwajen, Restivo kuma van de Rijt ta binciken ya nuna cewa dabaru na digital gwajen iya zama gaba daya daban-daban daga analog gwaje-gwajen. A Restivo kuma van de Rijt gwajin Miller, shi ne mai sauki ba da barnstar ga kowa a duniya kuma shi ne sauki waƙa da sakamako-yawan gyararrakin-kan wani Extended lokaci (saboda edit tarihin ne ta atomatik rubuta da Wikipedia). Wannan ikon isar da jiyya da kuma gwada sakamakon a wani farashi ne qualitatively sabanin gwaje-gwajen a baya. Ko da yake wannan gwaji da hannu da mutane 200, shi ma an gudu tare da 2,000 ko 20,000 mutane. Babban abu da hana masu bincike daga hawa sama da gwaji da wani factor na 100 da aka ba kudin, shi ne xa'a. Wato, Restivo kuma van de Rijt ba ya so ya ba barnstars to undeserving editoci da ba su so su gwaji to rushe Wikipedia al'umma (Restivo and Rijt 2012; Restivo and Rijt 2014) . Saboda haka, ko da yake gwaji na Restivo kuma van de Rijt ne in mun gwada sauki, shi a fili ya nuna cewa wasu abubuwa game da gwaje-gwajen sun zauna guda kuma wasu sun canza. Musamman, cikin muhimman dabaru na experimentation ne guda, amma dabaru sun canza. Next, domin ya more fili ware da damar halitta da wannan canji, zan kwatanta gwaje-gwajen da masu bincike za su iya yi yanzu da iri gwaje-gwajen da aka yi a baya.