Looking for law in all the wrong places: legal theory and legal neural networks

In this paper the author argues that neural networks have been poorly served by their applications to date in the legal domain. He argues that law and legal theory have been equally ignored in these applications. This paper thus raises a number of specific legal theoretical concerns. These include the di culties that the inherent nature of neural networks present, the problems of using statistical, sub-symbolic techniques in a (purportedly) symbolic field, the dangers in choosing the training set, and the legal philosophy apparent in current neural network implementations. The author suggests how we might create neural networks in the future which, perhaps, will not give rise to such concerns