RT @random_walker: This is an unusual position. Here's how it came to be. Last year I gave a talk on AI snake oil. Meanwhile Matt led a mas…
This is an unusual position. Here's how it came to be. Last year I gave a talk on AI snake oil. Meanwhile Matt led a mass collaboration that showed the limits of machine learning for predicting kids’ life outcomes. Paper in PNAS: https://t.co/0DirniQjqu ht
So what about interventions that are robust to multiple income shocks? Unfortunately it turns to be a known hard problem in general https://t.co/CeZiclRtWn
Machine-learning researchers drew on a vast dataset painstakingly collected by social scientists over 15y to predict life outcomes, such as grade point average and whether a family would be evicted. However, none made accurate predictions. https://t.co/aW
@math_rachel At Princeton, we use the Fragile Families dataset: a sociological dataset which tracks children from birth to age 15 in hopes of determining factors which keep them in poverty, etc: https://t.co/dCkmb1OgqQ Countless researchers, poor predicti
"Overall, these results suggest practical limits to the predictability of life outcomes in some settings" Measuring the predictability of life outcomes with a scientific mass collaboration https://t.co/8VAo5EZ3O7 Expected @tpq__ @cecilejanssens ? &HT
“if one measures our degree of understanding by our ability to predict, then the results of the Fragile Families Challenge suggest that our understanding of child development and the life course is actually quite poor. “
RT @kph3k: (2) Salganik et al (2020) https://t.co/CQMFr7rNfB
RT @kph3k: (2) Salganik et al (2020) https://t.co/CQMFr7rNfB
(2) Salganik et al (2020) https://t.co/CQMFr7rNfB
My take here is you use the appropriate method for your research.
Yep. I'm reminded of the Fragile Families project: https://t.co/9NspPTj0zu. Chances are, an OLS or logit model will do the job just fine (or even better).
RT @msalganik: If hundreds of scientists created predictive algorithms with high-quality data, how well would the best predict life outcome…
@BeccaFarnum @StephCPotts This is a bonkers story. I would bet small churches DON'T know what the data is really used for, especially because the things in the sales pitch (we predict divorce!) are...really really hard to predict: https://t.co/csHBocylDr
This week in the Boston Globe: “Algorithms may never really figure us out — thank goodness”, based on PNAS paper from the #FragileFamiliesChallenge https://t.co/UB9iIEhLAo
@elonmusk Socio-economic outcomes at t+1 https://t.co/9I8uVPASQa
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
A bunch of researchers tried to predict kids' life outcomes from incredibly large datasets. Accuracy was way bad. https://t.co/Hujy7rbM8W Also, tellingly, using fancy machine learning was just slightly better than a simple regression.
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
I'd also like to link another article that is marginally related to the algorithm part above, more specifically how predictive algorithms can fail. I've only skimmed through it (will read it later), so please note that I may be wrong. 2/2 https://t.co/XA
Life is quite unpredictable. Here is a study looking at the "problem". This is the reason why many promises in biomedicine, i.e. in predictive #genomics is hype. This was a prime argument in my PhD.
i share it in sp 😊
i share it in sp 😊
RT @MohsenJoshanloo: Predictive models of life outcomes were shown to be worthless in a massive study done by 160 teams that independently…
Predictive models of life outcomes were shown to be worthless in a massive study done by 160 teams that independently built predictive models for six life outcomes. The results were "only slightly better than those from a simple benchmark model". https://
“the best submissions, which often used complex machine-learning methods and had access to thousands of predictor variables, were only somewhat better than the results from a simple benchmark model that used linear or logistic regression with four... https
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
Individual life outcomes are unpredictable, no matter how much data you throw at it. I love this https://t.co/BEvRGkE6nB
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
Quelle: https://t.co/IHJwz3rQTi
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
(5) Die Studie heißt "Measuring the predictability of life outcomes with a scientific mass collaboration" und ist frei zugänglich unter https://t.co/C0nFWE8x6Z
Prob not relevant for "precision" medicine/epi/public health, right? Measuring the predictability of life outcomes with a scientific mass collaboration https://t.co/zaid2oD5YC https://t.co/U8f9qupATC
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
sofar no gattacca in real life :)
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
Léanse este estudio. El transcurso de la vida de las personas es fundamentalmente impredecible en base a determinantes previos. El determinismo implicado en muchas discusiones ideológicas no tiene base científica.
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
RT @sapinker: Here’s the study itself: https://t.co/2VHkOGiIIp
Here’s the study itself: https://t.co/2VHkOGiIIp
This isn’t just our finding. 160 teams also worked on this problem and found similar results; https://t.co/piVgoYIhnu
@BalazsAczel @dingding_peng .@msalganik et al. 2020 @FFCWS ? https://t.co/u4RlqqySgX
This is such an interesting research question that really makes you think about how researchers predict life outcomes. 😏😀 #LVCMarFam
Life is still larger than machines, at least for a few more decades. Will it ever be possible to predict people's life, and perhaps more importantly their emotions? If so are we going to keep calling it "life"?
RT @kph3k: My view of things is pragmatic: it is very difficult to identify replicable correlates of behavior, whether measures are envir o…
RT @kph3k: My view of things is pragmatic: it is very difficult to identify replicable correlates of behavior, whether measures are envir o…
RT @kph3k: My view of things is pragmatic: it is very difficult to identify replicable correlates of behavior, whether measures are envir o…
My view of things is pragmatic: it is very difficult to identify replicable correlates of behavior, whether measures are envir or bio, and all associations tend to be small. See, Eg, the Fragile Families Challenge if you want to depressed about it: https:/
“For policymakers considering using predictive models in settings such as criminal justice and child-protective services, these results raise a number of concerns.” Yep, worth a look.
RT @KarimJebari: This should not be surprising, but it certainly humbling. "Measuring the predictability of life outcomes with a scientific…
This should not be surprising, but it certainly humbling. "Measuring the predictability of life outcomes with a scientific mass collaboration" https://t.co/WcoCoQ2Rwz
Explico esto para poder resumir este estudio publicado recientemente. Un equipo gigantesco de investigadores evaluó la capacidad de varios modelos para predecir las trayectorias a largo plazo de seis características conductuales. https://t.co/EbcCFI7rFT
RT @hiddenmarkov: @databoydg I have recently used this paper by Salganik et al. for this argument https://t.co/3hNh3BqtFU It seems like th…
RT @hiddenmarkov: @databoydg I have recently used this paper by Salganik et al. for this argument https://t.co/3hNh3BqtFU It seems like th…
RT @hiddenmarkov: @databoydg I have recently used this paper by Salganik et al. for this argument https://t.co/3hNh3BqtFU It seems like th…
RT @hiddenmarkov: @databoydg I have recently used this paper by Salganik et al. for this argument https://t.co/3hNh3BqtFU It seems like th…
@databoydg I have recently used this paper by Salganik et al. for this argument https://t.co/3hNh3BqtFU It seems like the largest quantitative study that very convincingly demonstrates failure to predict social outcomes (hundreds of top researchers failed
RT @random_walker: Predicting the future is hard, whether you use a magic 8-ball or machine learning. New in PNAS—160 teams competed to pre…
@lindsey_mac @ProfAliceS looks interesting, reminds me of this: https://t.co/qWmzD5BA4y
RT @msalganik: If hundreds of scientists created predictive algorithms with high-quality data, how well would the best predict life outcome…
RT @JonathanMijs: Turns out sociologists *suck* at predicting life outcomes, but so are fancy algorithms. On the upside: life is too compl…
RT @nettwerkerin: Absolute reading recommendation: Measuring the predictability of life outcomes with a scientific mass collaboration. Spoi…
@bmwiernik @talyarkoni @timothycbates @ani_ta_twee_ta @SamHarrisOrg @JonHaidt @primalpoly @Steve_Sailer @sapinker @a_centrism Yes - life outcomes hard to predict. Here's probably the best work on this: https://t.co/QyJ2EdqtYJ (notice best predictor is $).
Study shows how prediction of human trajectories is difficult, even with mass collaboration and the best modeling tools of AI https://t.co/4UxtDg1uaA
RT @joe_deasy: Abandon all hope, ye modelers who enter here: "Measuring the predictability of life outcomes with a scientific mass collabor…
Abandon all hope, ye modelers who enter here: "Measuring the predictability of life outcomes with a scientific mass collaboration" https://t.co/5Ok2ju7bys
RT @IndexLlc: Excellent new paper showing how hard it is to accurately forecast long-term effects produced by complex adaptive systems - in this case, individuals' life outcomes. #investing #riskmanagement #CFA #RIA #CFP #nonexecs #forecasting https://t.c
RT @msalganik: If hundreds of scientists created predictive algorithms with high-quality data, how well would the best predict life outcome…
Excellent new paper showing how hard it is to accurately forecast long-term effects produced by complex adaptive systems - in this case, individuals' life outcomes. #investing #riskmanagement #CFA #RIA #CFP #nonexecs #forecasting https://t.co/mmA2dMHilB
Measuring the predictability of life outcomes with a scientific mass collaboration. https://t.co/ZoGLCeRLL2
New BUSINESS Research Available. Click Here to Access Article(s): https://t.co/TZ1Tc3RQcM
RT @nwheeler443: 100s of researchers attempted to predict six life outcomes, e.g. a child's GPA. Researchers used #machinelearning methods…
RT @nwheeler443: 100s of researchers attempted to predict six life outcomes, e.g. a child's GPA. Researchers used #machinelearning methods…
100s of researchers attempted to predict six life outcomes, e.g. a child's GPA. Researchers used #machinelearning methods optimized for prediction & drew on a vast dataset painstakingly collected over 15 y. However, no one made very accurate prediction
@attractfunding Overall distribution? https://t.co/RAFZ0yLexz
So leicht sind die Menschen nicht auszurechnen.
Read this research and read it again if you are working at any point re predictions on life outcomes / events #aiethics
RT @nettwerkerin: Absolute reading recommendation: Measuring the predictability of life outcomes with a scientific mass collaboration. Spoi…
RT @nettwerkerin: Absolute reading recommendation: Measuring the predictability of life outcomes with a scientific mass collaboration. Spoi…
RT @nettwerkerin: Absolute reading recommendation: Measuring the predictability of life outcomes with a scientific mass collaboration. Spoi…
RT @nettwerkerin: Absolute reading recommendation: Measuring the predictability of life outcomes with a scientific mass collaboration. Spoi…
RT @nettwerkerin: Absolute reading recommendation: Measuring the predictability of life outcomes with a scientific mass collaboration. Spoi…
Absolute reading recommendation: Measuring the predictability of life outcomes with a scientific mass collaboration. Spoiler: Predictability is very low. One of the co-authors is @random_walker . Thanks for this mass effort! https://t.co/jHQgv46FkV
@janusboye Tak for anbefaling 👍 og mindede mig om denne: https://t.co/Kp8WVG0z1x der er så absolut grund til at være ekstremt påpasselig med at bruge algoritmer til social risiko-score #dktech #dkpol #dksocial
Ada dong yang ngerjain kayak yang di Captain America: Winter Soldier itu tapi bukan pake DNA, lebih ke data sosial. https://t.co/bUWNk0Pi12
RT @NAChristakis: Deep and disturbing paper about social science, the vagaries of life, and our sense of control: Measuring the predictabil…