RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
RT @aichupanda: 1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15…
1) 한 아이의 미래를 예측하는 것은 얼마나 가능할까요? 정답은 "매우 어렵다"입니다. 2020년에 출판된 한 연구는 전세계 160개 연구팀이 협력을 했으며 12000개의 변수와 최신의 머신러닝 기법이 활용 출생에서 15세까지 아이들의 성장 과정에 대한 예측을 시도합니다. https://t.co/JYoBpt7nBt
Twee dingen nu: 1) voorspellen is iets anders dan validiteit onderzoeken, 2) voorspellen in het sociale domein is extreem moeilijk. Ik wil zo min mogelijk literatuur vermelden, maar op dit punt 2) is dit artikel van belang, Salganik c.s. https://t.co/vhA1
Another sobering read about the significant issues with predictive models and machine-learning: https://t.co/yFvB7QK0d0 (it's a bit nerdy - don't be put off by the algebra, see discussion section).
Great key-note by @FryRsquared at #SCWEC22! “Data-informed, not data-driven” Maybe an unpopular opinion for this audience but one take-away I fully subscribe! I also love you used failure to predict life outcomes with data in your talk. Loved that one! ht
@arthur_spirling I think this is probably common across the social sciences: see https://t.co/uaBwWHRcMY, which bills itself as using data "painstakingly" collected
RT @henryfarrell: @lastpositivist Likely you’ve seen this but in case not https://t.co/eZonvbaMae https://t.co/QxJshXGaVb
@PieterHendrikse @ASMLcompany 'Krachtig selectief vermogen'. Daar kijken we in de selectiepsychologie van op, want zoiets is toch wel heel zeldzaam. Tenminste, als het met kwaliteit gepaard moet gaan. Interessant in dit verband: https://t.co/Ub6MxoA8go
This article is about a scientific study that tried to predict six different life outcomes. The study found that the best predictions were only slightly better than a simple benchmark model. https://t.co/XsaodJnCOo
13| Quellen: Q1: https://t.co/f6TsxM4xkZ Q2: https://t.co/b905GDgtFw Q3: https://t.co/u3C9HexokU Q4: https://t.co/NoERDjhwGK
Tomorrow we welcome Chang-Tai Hsieh from @ChicagoBooth to our #Wednesdayseminar. He will present his paper "Romer or Ricardo?" https://t.co/91Yr0fbhoL #qualitydifferences #productivitydifferences #creativedestruction #innovation #importgrowthrates #researc
@andreastirez @LodeCossaer @wduyck Ken het boek niet, maar als het gaat over sociale mobiliteit, lees ook dit artikel; hoewel op niveau van populatie relatie aangetoond is, is het onmogelijke individuele uitkomsten te voorspellen. Comforting. https://t.co/
RT @msalganik: If hundreds of scientists created predictive algorithms with high-quality data, how well would the best predict life outcome…
@jt_kerwin @yanivbrandvain @JEconometrics This is really interesting. Reminds me of the Fragile Families challenge, too. https://t.co/I9ydqhvx5W
For instance, dozens of teams of highly qualified scientists armed with thousands of covariates were not that well able to predict typical sociological and psychological variables in the Fragile Families Challenge two years ago.(https://t.co/WOOhKFH26q) /3
If you are new to machine learning and social science/human applications, I recommend starting with @msalganik 's Fragile Families challenge: https://t.co/x7YkyLG3DU
@krichard1212 @LATECAT_ Nothing predicts achievement better than past achievement. What does that tell us? In general, though, it is extremely difficult to predict outcomes over periods of several years: (predictability of life outcomes) https://t.co/U
@ericjang11 Interesting, how much potential value do you expect additional work on a model of that kind to have? I’m particularly curious as it relates to attempts at predicting life outcomes. That seems, at a conceptual level, to have useful aspects that
@JimFruchterman @AndrewYNg However, most interesting social problems are really about predicting individuals' life outcomes (e.g. in education, employment, criminal justice, economic policy) for which "big models" have not performed well. (ref the study:
RT @msalganik: If hundreds of scientists created predictive algorithms with high-quality data, how well would the best predict life outcome…
RT @econ_martens: Fascinating paper on the ability of machine learning to predict life events. https://t.co/Ds864f2Tpe 1/n
Fascinating paper on the ability of machine learning to predict life events. https://t.co/Ds864f2Tpe 1/n
'Zonder experiment bevestigen we verschillen die al bestaan.' Is dat erg, dan? Ja, want de toekomst is niet goed voorspelbaar, althans veel en veel minder goed dan men denkt. Zie https://t.co/Ub6MxoA8go [via Harden 'The genetc lottery' note 13 ch 6]
@aggieerin @Savvylewis15 The entirety of the Fragile Families Challenge with this being the big paper from the competition: https://t.co/PWSDaMeBax
@bijlesduits 'Nauwelijks voorspelbaar', maar dat kan toch niet kloppen? Daarom doe ik er deze verwijzing bij: Matthew J. Salganik and many, many others (2020). https://t.co/Ub6MxoA8go Ik had zo'n experiment graag een halve eeuw eerder gezien.
Interesting experiment, ML-researchers tried to predict life out-comes of 15-yo's based on data from birth to age 9. "the best predictions were not very accurate" https://t.co/7yOlDFHrXq
@DidactiefOnline Interessant, heb ik in 2017 gemist?. Jammer dat de auteurs spreken over excellentie en talent alsof het om vaste kenmerken van de betreffende leerlingen zou gaan. Maar dat is dus niet zo: https://t.co/Ub6MxoA8go
This does not sound that different from Astrology https://t.co/mnPkH4iqhb
Voor ieder die denkt dat sociale wetenschappers heel veel kunnen voorspellen, wanneer ze maar goede data hebben [Niet, dus]: Matthew J. Salganik et al. (2013). 'Measuring the predictabilityof life outcomes with a scientific mass collaboration' open access
Leaving a citation in here to show it's not just this handful of talks, in case its helpful: https://t.co/qAvTR6eiGb
RT @NetNezva: de cómo fallan miserablemente los modelos predictivos de conducta humana, en Salganik et al. 'Measuring the predictability of…
RT @NetNezva: de cómo fallan miserablemente los modelos predictivos de conducta humana, en Salganik et al. 'Measuring the predictability of…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @otrasenda_AC: This might be the only social disciplines paper that may actually hold for reproducibility. The papers result: they could…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
This might be the only social disciplines paper that may actually hold for reproducibility. The papers result: they couldn't predict noting. https://t.co/XbtE8KDzFV https://t.co/JuUOreYLUW
RT @NetNezva: de cómo fallan miserablemente los modelos predictivos de conducta humana, en Salganik et al. 'Measuring the predictability of…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
de cómo fallan miserablemente los modelos predictivos de conducta humana, en Salganik et al. 'Measuring the predictability of life outcomes with a scientific mass collaboration' https://t.co/qCsPgSrxNh
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
RT @emollick: Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will…
Given tons of data on at-risk kids at 9, can we tell what will happen to them at age 15, like their GPA or whether they will be jailed? Not really! 160 teams of social scientists tried different machine learning & predictive models, and nothing worked
Wie denkt dat onderwijsloopbanen voor 12 jarigen goed zijn te voorspellen, zie Matthew J Salganik cs (2020). Measuring the predictability of life outcomes with a scientific mass collaboration abstract: https://t.co/vhA1lxPHTM open: https://t.co/QnCmHXSfBs
@OliverFaude @GeoWieland Was meinst du? Es gibt viele Studien, die menschliche Expertenprognosen mit mehr oder weniger techn. anspruchsvollen Modellen vergleichen. Wenn es ein social Outcome ist, sind Menschen tendenziell besser und das ist es ja hier. Bsp
@hypervisible The newspaper article title missed an opportunity to state "fakes predictive policing". This reminded me of the huge study attempting to predict life outcomes where "none of the predictions were very accurate". https://t.co/TjKC2xKSqy
@AlexTISYoung A point of comparison (which I discuss in the book): https://t.co/CQMFr7Jo79
RT @repTeaTokyo: 9/30はYarkoniさんもお勧めするCommon task method(予測コンペ)を、実際に社会科学研究につかってみたよ、という論文。 Salganik et al. (2020). Measuring the predictabili…
RT @repTeaTokyo: 9/30はYarkoniさんもお勧めするCommon task method(予測コンペ)を、実際に社会科学研究につかってみたよ、という論文。 Salganik et al. (2020). Measuring the predictabili…
9/30はYarkoniさんもお勧めするCommon task method(予測コンペ)を、実際に社会科学研究につかってみたよ、という論文。 Salganik et al. (2020). Measuring the predictability of life outcomes with a scientific mass collaboration. PNAS, 117(15), 8398–8403. https://t.co/9weLNF5h2D
@RobertoAnguloS @DANE_Colombia - Del poder para gestionar y analizar millones de datos y correr algoritmos complejos. - Sobre la participación del “pobre” en las mediciones de pobreza. - A propósito de resultados como: https://t.co/oCR5RovCfL En términos
Consideramos literatura reciente sobre la imposibilidad de predecir trayectorias vitales con métodos estadísticos (aka IA). A menudo, cuando este tipo de cosas se predicen es porque la BBDD tiene trampa. https://t.co/hOrQ9hizVM
RT @JakeMGrumbach: Reading this 2020 PNAS sociology piece on the difficulty of predicting life outcomes, even with lots good data and ML te…
I've found this pretty true -- you obviously want a complex, non-linear model to recognize a group of pixels as a "1", "2"... but for typical social science data you usually don't do much better than linear with maybe a couple interactions
RT @JakeMGrumbach: Reading this 2020 PNAS sociology piece on the difficulty of predicting life outcomes, even with lots good data and ML te…
RT @JakeMGrumbach: Reading this 2020 PNAS sociology piece on the difficulty of predicting life outcomes, even with lots good data and ML te…
RT @JakeMGrumbach: Reading this 2020 PNAS sociology piece on the difficulty of predicting life outcomes, even with lots good data and ML te…
Reading this 2020 PNAS sociology piece on the difficulty of predicting life outcomes, even with lots good data and ML techniques. Casts doubt on predictive modeling in criminal justice. Also makes descriptive & causal aggregates look cool in compariso
RT @turkerkilic: İnsan gelişiminin öngörülemezliği konusunda önemli bir yayın. https://t.co/rRYA7KAiJm
Çocuğun merak yolculuğu öğretici olandır. Öğrenmenin tersine döndüğü bu çağda kalıplaşmış erişkinler kendi özlerini arama fırsatı yakalayabilirler. https://t.co/MX7lZqIp6o
RT @turkerkilic: İnsan gelişiminin öngörülemezliği konusunda önemli bir yayın. https://t.co/rRYA7KAiJm
İnsan gelişiminin öngörülemezliği konusunda önemli bir yayın. https://t.co/rRYA7KAiJm
RT @emollick: Here’s the paper: https://t.co/nHrwU1zOST
RT @emollick: Here’s the paper: https://t.co/nHrwU1zOST
Here’s the paper: https://t.co/nHrwU1zOST
@cristobalyoung5 we can't predict things... https://t.co/fyFKvgVhGc
Measuring the predictability of life outcomes with a scientific mass collaboration | PNAS https://t.co/gRsG7LpiKL
RT @msalganik: If hundreds of scientists created predictive algorithms with high-quality data, how well would the best predict life outcome…