Thanks for this interesting question. In my opinion, the human ethical principles are difficult to be represented in a machine readable format. It might be possible to represent one single case ( not universal). Therefore, I think the optimal way to start with, is to use ontologies combined with uncertainty values, weights and tuning parameters. The ontology of such single case will be huge due to the several factors that play a role to decide if somebody is guilty or not. Thus, the problem of reasoning to inference such knowledge base is an other problem to think about it.

While one can:

1)model legal rules through deontic logic formalism

2) use data mining and fuzzy membership functions to deal with determining current decisions throughout the legal system

There still remains the  challenge that laws may have exceptions (which may be based on very arcane argumentation). The reasoning and entailment procedures I have seen do not have the required sophistication that such tool would need to handle such exceptions (not yet anyway).

In my opinion this can be done and today we have the technology developed to support. Take the example of Watson (developed by IBM) system which can suggest cancer treatment based on patience history.

Coming to legal system, the feasibility of developing it is high for following reasons:

1.       Legal structure is primarily rule based

2.       Historic data is available (in some case more than 100+ years)

To begin with, the system can provide suggestion to the judge who can make the final decision as in the case for cancer treatment, where the doctor refers Watson’s recommendations but uses his judgement to make the final decision.

Thank you all for your valuable comments.

Cheers