ntxiv tswv yim

Seem no yog tsim los siv raws li ib tug siv, es tsis yuav tsum tau nyeem raws li ib tug narrative.

• Introduction (Section 2.1)

Ib yam ntawm peb coj raws li hais tias yog tsis muaj nyob rau hauv tshooj no yog ethnography. Yog xav paub ntxiv nyob rau hauv ethnography nyob rau hauv cov chaw saib Boellstorff et al. (2012) , thiab rau ntau rau ethnography nyob rau hauv mixed cov thiab lub cev rau tej qhov chaw saib Lane (2016) .

• Loj cov ntaub ntawv (Section 2.2)

Thaum koj repurposing cov ntaub ntawv, muaj ob puas siab puas ntsws tricks uas yuav pab tau koj to taub cov tau cov teeb meem uas tej zaum koj yuav pom me ntsis. Ua ntej, koj yuav tau sim xav txog tej yam lub zoo tagnrho dataset rau koj qhov teeb meem thiab cov piv tias mus rau qhov dataset uas koj siv. Yuav ua li cas yog lawv zoo sib xws thiab yuav ua li cas no ho sib txawv? Yog hais tias koj tsis tau sau koj cov ntaub ntawv koj tus kheej, muaj tej zaum yuav muaj qhov sib txawv ntawm dab tsi koj xav tau thiab koj muaj dab tsi. Tab sis, koj yuav tsum txiav txim siab yog tias cov kev txawv no yog me los yog loj.

Ob, nco ntsoov hais tias ib tug neeg tsim thiab sau koj cov ntaub ntawv rau ib co yog vim li cas. Koj yuav tsum tau sim kom to taub lawv txoj kev xav. Qhov no zoo rov qab-engineering yuav pab tau koj txheeb xyuas tau cov teeb meem thiab biases nyob rau hauv koj repurposed cov ntaub ntawv.

Yog tsis muaj ib qho kev pom zoo ntsiab txhais ntawm "loj cov ntaub ntawv", tab sis muaj ntau yam txhais cov ntsiab lus zoo li siab rau lub 3 Vs: volume, ntau yam, thiab tshaj tawm (piv txwv li, Japec et al. (2015) ). Ntau dua li ntawm cov yam ntxwv ntawm cov ntaub ntawv, kuv txhais tsom ntau tshaj rau yog vim li cas cov ntaub ntawv yog tsim.

Kuv xam nrog ntawm tsoom fwv kev tswj cov ntaub ntawv hauv cov qeb ntawm loj cov ntaub ntawv yog ib tug me ntsis unusually. Lwm cov neeg uas tau ua cov ntaub ntawv no, xws li Legewie (2015) , Connelly et al. (2016) , thiab Einav and Levin (2014) . Yog xav paub ntxiv txog cov nqi ntawm tsoom fwv kev tswj cov ntaub ntawv rau kev tshawb fawb, saib Card et al. (2010) , Taskforce (2012) , thiab Grusky, Smeeding, and Snipp (2015) .

Rau ib tug saib ntawm kev tshawb fawb los ntawm hauv tsoom fwv statistical system, tshwj xeeb yog cov US Census Bureau, saib Jarmin and O'Hara (2016) . Rau ib phau ntawv ntev kev kho mob ntawm tus thawj tswj cov ntaub ntawv tshawb fawb ntawm Statistics Sweden, saib Wallgren and Wallgren (2007) .

Nyob rau hauv tshooj, kuv luv luv piv ib txhis daim ntawv ntsuam xyuas xws li cov General Social Survey (GSS) mus rau ib tug kev sib raug zoo xov xwm ntaub ntawv qhov twg los xws li Twitter. Rau ib txhij txhua thiab ceev faj kev sib piv ntawm cov tsoos cov ntawv tshawb fawb thiab kev tawm cov ntaub ntawv no, mus saib Schober et al. (2016) .

• Common yam ntxwv ntawm loj cov ntaub ntawv (Section 2.3)

Cov 10 cov yam ntxwv ntawm loj cov ntaub ntawv tau piav nyob rau hauv ib tug ntau yam ntawm txoj kev sib txawv los ntawm ib tug ntau yam sib txawv ntawm cov sau phau ntawv. Sau ntawv hais tias tus kuv xav rau cov teeb meem muaj xws li: Lazer et al. (2009) , Groves (2011) , Howison, Wiggins, and Crowston (2011) , boyd and Crawford (2012) , Taylor (2013) , Mayer-Schönberger and Cukier (2013) , Golder and Macy (2014) , Ruths and Pfeffer (2014) , Tufekci (2014) , Sampson and Small (2015) , Lewis (2015) , Lazer (2015) , Horton and Tambe (2015) , Japec et al. (2015) , thiab Goldstone and Lupyan (2016) .

Thoob plaws hauv tshooj no, kuv twb siv lub sij hawm cov ib co kua nplaum, uas kuv xav tias yog tus nruab nrab. Lwm nrov lub sij hawm rau cov ib co kua nplaum yog cov hneev (Golder and Macy 2014) , tab sis raws li Hal Abelson, Ken Ledeen, thiab Harry Lewis (2008) taw tes, ib tug ntau tsim nyog lub sij hawm no tej zaum cov yiv tes. Thaum koj tsim hneev, koj paub txog dab tsi tshwm sim thiab koj hneev yuav tsis feem ntau yuav twb tsis siv rau koj tus kheej. Tib yam yog tsis muaj tseeb rau koj cov khoom siv. Nyob rau hauv qhov tseeb, koj yuav ncaim ib co kua nplaum rau tag nrho cov sij hawm hais txog tej uas koj muaj heev me ntsis kev paub. Thiab, txawm hais tias cov ib co kua nplaum tsis muaj koj lub npe rau lawv, lawv muaj peev xwm feem ntau yuav txuas rov qab mus rau koj. Nyob rau hauv lwm yam lus, lawv yog neeg zoo li cov yiv tes: pom thiab tus kheej paub.

loj

Yog xav paub ntxiv txog vim li cas loj datasets, kav statistical kev ntsuam xyuas problematic, saib Lin, Lucas, and Shmueli (2013) thiab McFarland and McFarland (2015) . Cov teeb meem yuav tsum ua kev soj ntsuam kom pom tseeb rau cov tswv yim tseem ceeb es tsis statistical tseem ceeb.

Yuav-on

Thaum twg yog xav yeej ib txwm-on ntaub ntawv, nws yog ib qho tseem ceeb yuav tau xav txog seb koj puas muab piv rau cov caij nyoog tib neeg lub sij hawm los yog hais tias koj muab piv ib co kev hloov pab pawg neeg ntawm cov neeg; saib piv txwv li, Diaz et al. (2016) .

Uas tsis yog-reactive

Ib tug classic phau ntawv tsis yog-reactive ntsuas yog Webb et al. (1966) . Cov piv txwv nyob rau hauv phau ntawv pre-hnub cov muaj hnub nyoog, tab sis lawv tseem illuminating. Rau cov piv txwv ntawm cov neeg hloov lawv tus cwj pwm vim hais tias ntawm lub xub ntiag ntawm loj keb soj ntsuam, saib Penney (2016) thiab Brayne (2014) .

uas ua tsis tiav

Yog xav paub ntxiv nyob rau hauv cov ntaub ntawv linkage, saib Dunn (1946) thiab Fellegi and Sunter (1969) (keeb kwm lub) thiab Larsen and Winkler (2014) (niaj hnub). Zoo li cov mus tau kuj tau tsim nyob rau hauv computer science nyob rau hauv lub npe xws li cov ntaub ntawv deduplication, Piv txwv li kev qhia kom paub, lub npe txuam, theej tawm nrhiav kom tau, thiab theej tawm cov ntaub ntawv nrhiav kom tau (Elmagarmid, Ipeirotis, and Verykios 2007) . Tseem muaj kev ceev ntiag tug khaws cia le caag los sau linkage uas tsis yuav tsum tau cov kis ntawm tus kheej yam ntaub ntawv (Schnell 2013) . Facebook kuj tau tsim ib tug npaj mus rau txuas lawv cov ntaub ntawv mus pov ntawv xaiv tsa tus cwj pwm; qhov no yog ua los soj ntsuam ib tug xyaum ua tej yam uas kuv yuav qhia rau koj txog nyob rau hauv Tshooj 4 (Bond et al. 2012; Jones et al. 2013) .

Yog xav paub ntxiv nyob rau hauv dlaim validity, saib Shadish, Cook, and Campbell (2001) , Tshooj 3.

muab tsis cuag

Yog xav paub ntxiv nyob rau hauv lub AOL nrhiav cav debacle, saib Ohm (2010) . Kuv muaj tswv yim hais txog koom tes nrog cov tuam txhab uas muag thiab tsoom fwv nyob rau hauv Tshooj 4 thaum kuv piav thwmsim. Ib tus xov tooj ntawm tus sau phau ntawv muaj qhia kev txhawj xeeb txog kev tshawb fawb uas cia siab rau muab tsis cuag cov ntaub ntawv no, mus saib Huberman (2012) thiab boyd and Crawford (2012) .

Ib txoj kev zoo rau lub tsev kawm ntawv soj ntsuam tau tej ntaub ntawv yog ua hauj lwm rau ib lub tuam txhab raws li ib tug intern los sis mus xyuas soj ntsuam. Nyob rau hauv tas li ntawd mus enabling tej ntaub ntawv, qhov no yuav tseem yuav pab tus kws tshawb fawb xav paub ntau ntxiv txog yuav ua li cas rau cov ntaub ntawv yog tsim, uas yog ib qho tseem ceeb rau kev tsom xam.

Uas tsis yog-tus neeg sawv cev

Uas tsis yog-representativeness yog ib tug loj teeb meem rau kev soj ntsuam thiab tsoom fwv uas xav mus ua nqe lus hais txog ib tug tag nrho cov pejxeem. Qhov no yog tsawg dua ntawm kev txhawj xeeb rau cov tuam txhab uas feem ntau teem rau lawv cov neeg siv. Yog xav paub ntxiv nyob rau hauv yuav ua li cas Statistics Netherlands considers qhov teeb meem uas tsis yog-representativeness ntawm ua lag ua luam loj cov ntaub ntawv, saib Buelens et al. (2014) .

Nyob rau hauv Tshooj 3, Kuv mam li piav qhia txog piv txwv thiab kwv yees nyob rau hauv ntau dua kom meej. Txawm hais tias cov ntaub ntawv no tsis yog-tus neeg sawv cev, nyob rau hauv tej yam kev mob, lawv muaj peev xwm yuav hnyav rau tsim zoo kev kwv yees.

drifting

System drift yog heev nyuaj rau saib los ntawm sab nraum. Txawm li cas los, lub MovieLens project (tham ntau nyob rau hauv Tshooj 4) tau khiav rau ntau tshaj 15 xyoo los ntawm ib tug kev kawm tshawb fawb pab pawg neeg. Yog li ntawd, lawv tau sau thiab qhia cov lus qhia txog txoj kev uas lub system muaj evolved dua lub sij hawm thiab yuav ua li cas qhov no tej zaum kev cuam tshuam tsom xam (Harper and Konstan 2015) .

Ib tug xov tooj ntawm kws tshawb fawb tau teem rau drift nyob rau hauv Twitter: Liu, Kliman-Silver, and Mislove (2014) thiab Tufekci (2014) .

Algorithmically neeg nyias hais nyias

Kuv xub hnov ​​lub sij hawm "algorithmically neeg nyias hais nyias" siv los ntawm Jon Kleinberg nyob rau hauv ib tug sib tham. Lub ntsiab lub tswv yim qab performativity yog tias ib txhia social science theories yog "xyaw tsis koob yees duab" (Mackenzie 2008) . Uas yog, lawv ua tau zoo rau hauv lub ntiaj teb no es cia ntes nws.

Dirty

Tsoom fwv statistical cov koom haum hu cov ntaub ntawv tu, ntaub ntawv editing. De Waal, Puts, and Daas (2014) piav qhia txog tej ntaub ntawv editing hom kev kawm tsim rau daim ntawv ntsuam xyuas cov ntaub ntawv thiab xyuas uas raws li lawv muaj feem xyuam rau loj cov ntaub ntawv qhov chaw, thiab Puts, Daas, and Waal (2015) pib ib co ntawm cov tib lub tswv yim rau ib tug ntau rau sawv daws.

Rau ib co piv txwv ntawm cov kev tshawb fawb teem rau spam nyob rau hauv Twitter, Clark et al. (2016) thiab Chu et al. (2012) . Thaum kawg, Subrahmanian et al. (2016) qhia txog cov kev tshwm sim ntawm lub DARPA Twitter Bot sib tw.

sensitive

Ohm (2015) xyuas ua ntej lawm kev tshawb fawb nyob rau hauv cov tswv yim ntawm rhiab cov ntaub ntawv thiab muaj ib tug multi-tau xeem. Cov plaub yam nws npaj yog: qhov yuav tshwm ntawm kev tsim txom; yuav raug mob; muaj ib tug pub leej twg paub kev sib raug zoo; thiab seb cov kev pheej hmoo muaj kev cuam tshuam majoritarian kev txhawj xeeb.

• Suav tej yam (Section 2.4.1)

Farber txoj kev kawm taxis nyob rau hauv New York tau raws li nyob rau hauv kev ua ntej lawm txoj kev kawm Camerer et al. (1997) uas siv peb sib txawv yooj yim kuaj ntawm daim ntawv txawv tebchaws los xij-daim ntawv ntaub ntawv siv los ntawm cov neeg tsav tsheb rau sau txawv tebchaws pib lub sij hawm, thaum xaus lub sij hawm, thiab nqi. Qhov no ua ntej lawm kev tshawb pom tias cov neeg tsav tsheb ciali mus yuav lub hom phiaj earners: lawv ua hauj lwm tsawg nyob rau hnub uas lawv nyiaj ua hauj lwm tau ntau dua.

Kossinets and Watts (2009) tau teem rau lub hauv paus pib ntawm homophily nyob rau hauv kev tes hauj lwm. Saib Wimmer and Lewis (2010) rau ib tug txawv mus kom ze rau qhov teeb meem qub uas siv cov ntaub ntawv los ntawm Facebook.

Nyob rau hauv tom ntej ua hauj lwm, King thiab lug txhawb cov miv tau ntxiv tshawb hauv internet censorship nyob rau hauv Tuam Tshoj (King, Pan, and Roberts 2014; King, Pan, and Roberts 2016) . Rau ib yam kev mus kom ze rau ntsuas online censorship nyob rau hauv Tuam Tshoj, saib Bamman, O'Connor, and Smith (2012) . Yog xav paub ntxiv nyob rau hauv statistical txoj kev zoo li ib siv nyob rau hauv King, Pan, and Roberts (2013) los laij rau tus sentiment ntawm cov 11 lab posts, saib Hopkins and King (2010) . Yog xav paub ntxiv nyob rau hauv saib xyuas kev kawm, saib James et al. (2013) (tsawg kev) thiab Hastie, Tibshirani, and Friedman (2009) (ntau kev).

• Forecasting (Section 2.4.2)

Forecasting yog ib tug loj ib feem ntawm muaj ntaub ntawv kev kawm (Mayer-Schönberger and Cukier 2013; Provost and Fawcett 2013) . Ib hom ntawm forecasting uas nquag ua los ntawm kev sib raug zoo soj ntsuam yog demographic forecasting, piv txwv li Raftery et al. (2012) .

Google kab mob khaub thuas Tiam sis tsis yog tus Thawj peb tes num rau siv kev nrhiav cov ntaub ntawv rau nowcast kab mob khaub thuas loj heev. Nyob rau hauv qhov tseeb, kev soj ntsuam nyob rau hauv lub tebchaws United States (Polgreen et al. 2008; Ginsberg et al. 2009) thiab Sweden (Hulth, Rydevik, and Linde 2009) tau pom tias tej yam kev tshawb fawb cov ntsiab lus uas (xws li, "mob npaws") kwv yees lub teb chaws pej xeem kev noj qab nyob keb soj ntsuam cov ntaub ntawv ua ntej nws tso tawm. Tom qab ntau, ntau lwm yaam num muaj sim siv cov cim tseg cov ntaub ntawv rau tus kab mob keb soj ntsuam kom paub tias, saib Althouse et al. (2015) rau ib tug saib xyuas.

Nyob rau hauv tas li ntawd mus siv cov cim tseg cov ntaub ntawv twv seb noj qab haus huv tau, muaj kuj tau ib tug lossis loj npaum li cas ua hauj lwm siv Twitter cov ntaub ntawv twv seb kev xaiv tsa ua tau; rau kev txheeb xyuas saib Gayo-Avello (2011) , Gayo-Avello (2013) , Jungherr (2015) (Ch. 7), thiab Huberty (2015) .

Siv kev tshawb fawb cov ntaub ntawv rau kev kwv yees kab mob khaub thuas loj heev thiab siv Twitter cov ntaub ntawv twv seb kev xaiv tsa yog ob qho tib si piv txwv ntawm kev siv ib co zoo ntawm cov cim tseg twv seb tej yam kev tshwm sim nyob rau hauv lub ntiaj teb no. Muaj ib tug enormous tooj ntawm cov kev tshawb fawb uas muaj no general qauv. Rooj 2.5 muaj xws li ob peb lwm cov piv txwv.

• Approximating thwmsim (Section 2.4.3)

Cov phau ntawv journal PS Political Science muaj ib tug symposium rau loj cov ntaub ntawv, causal inference, thiab tsim kev tshawb xav, thiab Clark and Golder (2015) tshab xyuas txhua pab. Cov phau ntawv journal zaug ntawm qhov National Academy ntawm Sciences ntawm lub tebchaws United States of America muaj ib tug symposium nyob causal inference thiab loj cov ntaub ntawv, thiab Shiffrin (2016) tshab xyuas txhua pab.

Nyob rau hauv cov nqe lus ntawm tej yam ntuj tso thwmsim, Dunning (2012) muab ib qho zoo heev phau ntawv ntev kev kho mob. Yog xav paub ntxiv nyob rau kev siv cov Vietnam cua ntsawj ntshab lottery raws li ib tug natural xyaum ua tej yam, saib Berinsky and Chatfield (2015) . Rau lub tshuab kev kawm le caag uas sim tau nrhiav tej yam ntuj tso kev rau hauv loj cov ntaub ntawv qhov chaw, saib Jensen et al. (2008) thiab Sharma, Hofman, and Watts (2015) .

Nyob rau hauv cov nqe lus ntawm txuam, rau ib tug optimistic kev ntsuam xyuas, saib Stuart (2010) , thiab rau ib tug phem heev xyuas saib Sekhon (2009) . Yog xav paub ntxiv nyob rau hauv txoj kev raws li ib yam ntawm cov pruning, saib Ho et al. (2007) . Rau cov phau ntawv uas muab zoo heev kev kho mob ntawm txuam, saib Rosenbaum (2002) , Rosenbaum (2009) , Morgan and Winship (2014) , thiab Imbens and Rubin (2015) .