Automated method for predicting psychosis has 100 percent accuracy, small sample size

Computers may “understand” mental disorders better than humans do.Columbia researchers helped create the first computer-driven method of predicting with 100% accuracy the development of psychosis. This automated method relies only on words and word order to predict the disorder. By combining automated speech analysis with machine learning, the researchers correctly predicted which of five individuals from a sample of 34 at-risk youth would develop psychosis over a period of 2.5 years. The study, which was published in Nature, documents a vast improvement over other, nonautomated predictive models, which range from about 70 to 80 percent accuracy. Despite its success, the small sample size of just 34 individuals casts doubt on the validity of its results. To verify that the accuracy of the study is not a fluke, the researchers will…


Link to Full Article: Automated method for predicting psychosis has 100 percent accuracy, small sample size