Law & Disorder: Mathematical Models in a Messy World

Abstract: It's a common misconception that our laws are a neat set of rules, and that all that stands in the way of their perfect application are the humans who carry them out. Not so. Laws are ambiguous, undefined, underdefined, contradictory, and constantly changing. Applying a law is not a matter of looking up the right rule and following it. Judges must interpret, balance, and make subjective decisions about what laws mean even before getting to subjective decisions about evidence, argument, and proof. The law is messy and human. In our work at Judicata building a map of the law, we've found these attributes unavoidable.

Messiness is not limited to law. It infects all sorts of domains. Products built on data are are attempts at imposing structure on world. Much of the time, however, we run into situations that break our products. We find outliers. We discover unexpected scenarios. We realize the data available to us doesn't tell the whole story. Our attempts at structure falls short. We are building products for a messy world, one that is inevitably more intricate than we handle.

So what can we do? At Judicata, we've learned to embrace the mess. In this talk, I'll explain our iterative, high precision approach that allows us to make steady progress while keeping our error rate down. We tightly integrate domain expertise and engineering development. I'll describe how our domain experts are involved in system design, and how our engineers build systems that allow domain experts to direct improvements and to verify functionality. Building high quality data products is a team effort, and even then it isn't easy. By expecting surprises from the beginning, we are able to make progress even as the oddities and corner cases pile up.

Bio: Ben Pedrick is the CTO at Judicata, which is using technology to improve the quality and lower the cost of legal representation. Judicata is building a map of the law applying NLP to legal documents. That map enables new products that evaluate lawyers' arguments, automate paralegal work, and surface information critical to court cases. Prior to working at Judicata, Ben graduated from Stanford and worked at few other start ups in San Francisco.

Open Data Science Conference