Predictability, a predicament?

Abstract - In the context of AI in general and Machine Learning in particular, predictability is usually considered a blessing. After all – that is the goal: build the model that has the highest predictive performance. The rise of ‘big data’ has in fact vastly improved our ability to predict human behavior thanks to the introduction of much more fine grained and informative features. However, in practice things are more complicated. For many applications, the relevant outcome is observed for very different reasons. In such mixed scenarios, the model will automatically gravitate to the one that is easiest to predict at the expense of the others. This even holds if the more predictable scenario is by far less common or relevant. We present a number of applications across different domains where the availability of highly informative features can have significantly negative impacts on the usefulness of predictive modeling and potentially create second order biases in the predictions. Neither model transparency nor first order data de-biases are ultimately able to mitigate those concerns. The moral imperative of those effects is that as creators of machine learning solutions it is our responsibility to pay attention to the often subtle symptoms and to let our human intuition be the gatekeeper when deciding whether models are ready to be released 'into the wild'.

Short bio - Claudia Perlich started her career at the IBM T.J. Watson Research Center, concentrating on research and application of Machine Learning for complex real-world domains and applications. From 2010 to 2017 she acted as the Chief Scientist at Dstillery where she designed, developed, analyzed, and optimized machine learning that drives digital advertising to prospective customers of brands. Her latest role is Head of Strategic Data Science at TwoSigma where she is creating quantitative strategies for both private and public investments. Claudia continues to be an active public speaker, has over 50 scientific publications, as well as numerous patents in the area of machine learning. She has won many data mining competitions and best paper awards at Knowledge Discovery and Data Mining (KDD) conference, where she served as the General Chair in 2014. Claudia is the past winner of the Advertising Research Foundation’s (ARF) Grand Innovation Award and has been selected for Crain’s New York’s 40 Under 40 list, Wired Magazine’s Smart List, and Fast Company’s 100 Most Creative People. She acts as an advisor to a number of philanthropic organizations including AI for Good, Datakind, Data and Society and others. She received her PhD in Information Systems from the NYU Stern School of Business where she continues to teach as an adjunct professor.

https://twosigmaventures.com/our-team/profile/claudia-perlich/