RT @melindacmills: #MachineLearning models predict life trajectories only slightly better than simple benchmark model - new @PNASNews wit…
RT @melindacmills: #MachineLearning models predict life trajectories only slightly better than simple benchmark model - new @PNASNews wit…
RT @melindacmills: #MachineLearning models predict life trajectories only slightly better than simple benchmark model - new @PNASNews wit…
RT @melindacmills: #MachineLearning models predict life trajectories only slightly better than simple benchmark model - new @PNASNews wit…
#MachineLearning models predict life trajectories only slightly better than simple benchmark model - new @PNASNews with our @ridhikash07 @OxfordDemSci led by @msalganik in massive collaboration https://t.co/DZPjIJC3xK
@HayesAMark The graph reminded me of my big takeaway from the following paper: complex models and fancy tools did not perform better: https://t.co/MiQIieZDts
RT @NuffieldLibrary: Measuring the predictability of life outcomes with a scientific mass collaboration https://t.co/l6KTfjQudU new paper i…
Measuring the predictability of life outcomes with a scientific mass collaboration https://t.co/l6KTfjQudU new paper in @PNASNews with contributions from @NuffieldCollege Fellow @ridhikash07
Ummm...maybe of interest...however, unless i am mistaken, there are mainstream data measuring fewer traits that have relatively high predictability...there may be too much variability &/or insufficient specificity in these data to yield reliable result
@SolomonKurz @DrKathrynGordon @aaronjfisher Great addition here. Everything in psychology is scale dependant and thinking needs to change as you go from individual->dyad->family->community->organisation->society. Explaining results like thes
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…
prediction of life events is hard, description might be where our science needs to go -Measuring the predictability of life outcomes with a scientific mass collaboration https://t.co/Ce2bQ5Eft5
RT @Descartes_Ghost: @AurelLevasseur Il y a une étude récente qui donne un poids scientifique à ce ressenti. Il n'y a pas de prédictabilité…
RT @Descartes_Ghost: @AurelLevasseur Il y a une étude récente qui donne un poids scientifique à ce ressenti. Il n'y a pas de prédictabilité…
RT @Descartes_Ghost: @AurelLevasseur Il y a une étude récente qui donne un poids scientifique à ce ressenti. Il n'y a pas de prédictabilité…
@AurelLevasseur Il y a une étude récente qui donne un poids scientifique à ce ressenti. Il n'y a pas de prédictabilité basée sur n'importe quelle méthode statistique qui permette de faire...mieux que la moyenne. https://t.co/ivWHc0qnrZ
@paulportesi It is even worse (two birds in one stone) https://t.co/kNkiGLEr7B
RT @chris_bail: 1/n Can machine learning solve social science problems? This unprecedented study led by @msalganik and dozens of co-authors…
Having a rich data set does not necessarily mean important life outcomes can be predicted. Results of maximizing the prediction in a complete dataset do not hold up under cross-validation. Yes we know already for ages that simple models can outperform com
RT @HansKoppies: 160 teams van sociale wetenschappers probeerden 6 uitkomsten (onder meer schoolsucces, 'grit' kind, materiële ontbering) t…
RT @HansKoppies: 160 teams van sociale wetenschappers probeerden 6 uitkomsten (onder meer schoolsucces, 'grit' kind, materiële ontbering) t…
160 teams van sociale wetenschappers probeerden 6 uitkomsten (onder meer schoolsucces, 'grit' kind, materiële ontbering) te voorspellen voor het leven van een 15-jarige (n=5000). Geen van de voorspellingen klopte. https://t.co/8iwiHUfK5t
Measuring the predictability of life outcomes with a scientific mass collaboration? @pnasnews https://t.co/OYzmL7s0Ct
Measuring the predictability of life outcomes with a scientific mass collaboration #ArtificialIntelligence via https://t.co/bDTgBUIWtG https://t.co/R7W8TYGOoV
RT @random_walker: Reminder: machine learning is good at many things but not at predicting the future. Using the word "prediction" to refer…
Measuring the predictability of life outcomes with a scientific mass collaboration [Social Sciences] https://t.co/cejagh7sEV @PNASNews
My favorite takeaway:: Simple models with a few predictors were as informative as complex, machine-learning models.... Would like to more such comparisons
RT @chris_bail: 1/n Can machine learning solve social science problems? This unprecedented study led by @msalganik and dozens of co-authors…
RT @random_walker: Reminder: machine learning is good at many things but not at predicting the future. Using the word "prediction" to refer…
A new paper in @PNASNews analyzes the Fragile Families Challenge, a mass collaboration between 160 teams of researchers. In 2017, a team led by Human Dynamics head @alex_pentland won 3 of 6 categories in the Challenge. https://t.co/5VDFbIPUbh https://t.
RT @etumnamboa: Peut-on, avec du machine learning, prévoir les trajectoires de vie ? Measuring the predictability of life outcomes with a…
Peut-on, avec du machine learning, prévoir les trajectoires de vie ? Measuring the predictability of life outcomes with a scientific mass collaboration https://t.co/EnSiPGsTGi
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
@OppenheimerEvan Maybe unknown immortal orchestra? They blew me away at first, not sure whether they held up. Your question reminds me of this: https://t.co/OqgYBw5S6e
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
Limits of predicting life outcomes, with implications for policymakers e.g. in criminal justice. 100s of researchers + vast dataset collected over 15y + machine-learning, but best predictions only slightly better than those of a simple benchmark model http
If a field of study has to have "science" in its name, it might not be a science.
RT @Psychologician: "Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were…
RT @Psychologician: "Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were…
RT @Psychologician: "Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were…
RT @Psychologician: "Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were…
The predictable failure of predictions: "...despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were not very accurate and were only slightly better than those from a simple benchmark model" htt
RT @BeEngelhardt: Extremely excited to see the Fragile Families Challenge paper out. @msalganik asked if ML worked to predict life outcomes…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @Kathy_Rastle: Mass collaboration approach using machine learning finds it is pretty tough to predict life outcomes from a massive birth…
RT @Kathy_Rastle: Mass collaboration approach using machine learning finds it is pretty tough to predict life outcomes from a massive birth…
RT @Kathy_Rastle: Mass collaboration approach using machine learning finds it is pretty tough to predict life outcomes from a massive birth…
Mass collaboration approach using machine learning finds it is pretty tough to predict life outcomes from a massive birth cohort study with thousands of variables. Big questions for social science and policy applications. @Nightingale_P
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
Bookmarked this to read. Firstly, does this support or throw doubt on mixed methods. Secondly, does it just show that conclusions in such questions are due to political pressure and all we can hope for is illumination. @cecanexus
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
Measuring the #predictability of #life #outcomes with a scientific mass collaboration "Results suggest limits to predictability of life outcomes in some settings & illustrate value of mass collaborations in social sciences" https://t.co/KUyMQAg9tO ht
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
RT @paul1kirby: ‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research…
‘Hundreds of researchers found that they could not predict the life outcomes for individuals which social science research says are very predictable’. Fascinating new report. https://t.co/roRrlIdHbY https://t.co/lObV2lcQqa
RT @RebeccaSear: Measuring predictability of life outcomes with a mass collaboration:"160 teams built predictive models using data from Fra…
Measuring predictability of life outcomes with a mass collaboration:"160 teams built predictive models using data from Fragile Families & Child Wellbeing Study. Despite using machine-learning methods optimized for prediction, predictions were not accur
RT @Psychologician: "Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were…
RT @Psychologician: "Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were…
RT @Psychologician: "Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were…
1/2 Massive data fancy models, terrible predictions. I hope ecologists read/do more papers like this, makes me think that the biggest thing we have to discover is our own ignorance. https://t.co/I40zY9WDVq
RT @Psychologician: "Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were…
RT @Psychologician: "Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were…
"Despite using a rich dataset and applying machine-learning methods optimized for prediction, the best predictions were not very accurate and were only slightly better than those from a simple benchmark model." https://t.co/L7zxeEVUsw https://t.co/qv1hwIvU
RT @gerrygcarter: Predicting a child's future is hard: 160 teams of researchers tried to predict 6 outcomes for the life of a 15-yr-old (eg…
Individual outcomes are barely predictable, although aggregate ones are: https://t.co/ZByoaI5irM. Explanation: individual R^2 in social science are low because of high interpersonal variation, which can be reduced by averaging across groups. Commentary:
Predicting a child's future is hard: 160 teams of researchers tried to predict 6 outcomes for the life of a 15-yr-old (eg GPA, hardship) using advanced machine-learning and a vast high-quality dataset (5000 children). None of the predictions were accurate.
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…
Be careful of using predictive analytics to set policy. This was certainly an eye opener for me. https://t.co/xUbXDqOigC
Measuring the predictability of life outcomes with a scientific mass collaboration | PNAS https://t.co/hoVXQ0P5F9
While it's true ML is not magic, the only takeaway from this study is that the input features used (mostly based on surveys) are not good enough to predict social outcomes. Original study at https://t.co/6K0c76PXrU
RT @patilindrajeets: Measuring the predictability of life outcomes with a scientific mass collaboration https://t.co/zpLXEa6PGB "Despite u…
RT @patilindrajeets: Measuring the predictability of life outcomes with a scientific mass collaboration https://t.co/zpLXEa6PGB "Despite u…
RT @patilindrajeets: Measuring the predictability of life outcomes with a scientific mass collaboration https://t.co/zpLXEa6PGB "Despite u…
Measuring the predictability of life outcomes with a scientific mass collaboration https://t.co/zpLXEa6PGB "Despite using a rich dataset and applying ML methods optimized for prediction, the best predictions were not very accurate and only slightly better
Princeton Uni asked hundreds #researchers to predict 6 life outcomes for #children, parents, and households using nearly 13,000 data points on over 4,000 families. None of the researchers got even close to a reasonable level of accuracy #policy #science h
RT @jaytiesse: Measuring the predictability of life outcomes with a scientific mass collaboration https://t.co/1reUuR7Cqc = conclusion: mac…
Measuring the predictability of life outcomes with a scientific mass collaboration https://t.co/1reUuR7Cqc = conclusion: machine learning isn't very predictive when it comes to life outcomes (thanks @manovich for pointing me to this article).
No one knows anything! Now we know for sure!
Limits of predictive analytics in/and social sciences. https://t.co/62iprvJQll
RT @nils_gilman: Hundreds of researchers attempted to forecast life outcomes using machine-learning methods optimized for prediction, drawi…
RT @mattzook: New study trying to measure the predictability of life outcomes from massive amounts of data and machine learning. "no one…
RT @nils_gilman: Hundreds of researchers attempted to forecast life outcomes using machine-learning methods optimized for prediction, drawi…