- Cet évènement est passé
Machine Learning and Causality: The Emperor’s New Clothes, or a New Research Pipeline? – Matthew Vowels
janvier 31 @ 10:30 - 12:00
The replicability crisis has drawn attention to numerous weaknesses in psychology and social science research practice. In this work we focus on two issues that cannot be addressed with replication alone, and which deserve much more attention: Functional and structural misspecification. We demonstrate the severity of these problems via simulation, and provide recommendations for a new research and analysis pipeline. We argue that psychologists should expand their analytical arsenal to incorporate recent developments in machine learning and causality, and show that by doing so they no longer have to rely on adages such as ‘all models are wrong, but some are useful’, or ‘correlation is not causation’, in order to justify the limitations of current practice. It is only through an awareness and understanding of varied and modern statistical methods for predictive and causal modeling that researchers will have the tools with which to answer the cause-effect questions which are fundamental to human understanding.