Jim Smith is a foundational Bayesian Statistician and Decision Analyst based at Warwick University. He started as a mathematical statistician specialising in dynamic dynamic models especially their implicit geometry and algebraic description and still publishes in these areas. However most of his work over the last twenty years or so lies on the interface between machine learning, statistics and operations research particularly within the realm of complex models and high dimensional data. He is especially skilled in developing bespoke graphical representations of problems - often dynamic ones such as dynamic Bayes nets, flow graphs multiregression dynamic models and most recently chain event graphs.
He has also written widely on Bayesian causal reasoning, developed methodologies for eliciting the structure of models from domain experts and developing decision support systems that integrate in a coherent way data sources that are very different form each other but pertain to the same processes. He has worked in a wide range of domains including, most recently, food security modelling, nuclear safety, biological regulation, brain connectivity, public health, military decision making and radicalisation processes.
He is working under a number of themes within Turing. One stream of work is the statistical modelling of violent criminal populations of various kinds. Another stream of work focuses on forensic inference - its graphical representation and the combination of different evidence types. Finally he is CI on a project designed to help government use data sources wisely and in a balanced way. He has also developed an interest in natural language processing and is currently researching ways causal relationships can be extracted from natural language descriptions of text written by engineers.