Description

Register for this Event

Data have always been the lifeblood of the sciences, giving rise to new theories, invalidating old. Data are costly to acquire and in most fields have long been a constraint on progress. Over the last few decades this situation has structurally changed in several fields, with information technology leading to rapidly declining costs of data acquisition.

Since more data are better than fewer, ceteris paribus on quality, growing data has been welcomed. But in the last few years, in multiple fields, there have appeared so many data that no extant theories are able to explain more than a tiny fraction of such data. In this talk Prof Axtell will illustrate this phenomenon based on quasi-comprehensive micro-data with respect to American business firms. There are a large number of theories of firms, none of which appear capable of rationalising these data. Prof Axtell will discuss the use of machine learning approaches for modeling data of this type, the value of explicable models, and how machine learning techniques can be used to create new data structures in order to better understand the data. Finally, he will mention complications that arise in the social sciences when individuals use machine learning to make sense of their environment.

REGISTRATION

This talk is live in-person and online

About the speaker:

Professor Axtell works at the intersection of the computational, social, behavioural and economic sciences. His research group combines agent-based computing with micro-data to build large-scale models having high verisimilitude with the real-world. They have worked on a variety of policy issues, from housing to fisheries, behavioural aspects of retirement and science policy.

He has been visiting Professor at the MIT, University of Oxford, the New School for Social Research (NY), Middlebury College (Vermont), and Google Research.

Professor Axtell supervises Ph.D. students in Mason's Computational Social Science Ph.D. program. He also serves on Ph.D. committees from other departments when the work involves agent-based modeling, e.g., Economics, Systems Engineering and Operations Research, Public Policy, and Civil and Infrastructure Engineering. He has been visiting Professor at the MIT, University of Oxford, the New School for Social Research (NY), Middlebury College (Vermont), and Google Research.

His research has been published in leading general interest journals ("Science," "Nature" and "Proceedings of the National Academy of Sciences"), in field journals (e.g., "American Economic Review," "Economic Journal", "Computational and Mathematical Organization Theory," "Journal of Industrial Ecology"), and reprised in newspapers (e.g., "Wall St. Journal," "Washington Post") and technology publications (e.g., "Scientific American," "Technology Review," "Wired").