Feb 2021

No Code automated reasoning for case law with Logiak

Machine learning techniques have some demonstrated applicability with legal tech, notably in helping lawyers to navigate the often vast document oceans they are faced with. It remains the case though that in this field the “other” kind of AI — automated reasoning — remains an intuitively appropriate partner technology.

We describe here what that can mean and how a No Code approach is directly applicable. Statute Law with its explicitly stated rules is the more amenable to automation, but there is also clear application in Case Law to draw together and “compile” expertise.

Case Law

Case Law reasoning is formidably difficult to automate because the logic with which one needs to reason is implicit in preceding case judgements whose features are never identical with the facts of future cases.

There is always the need to use quintessentially human techniques such as analogical reasoning to draw the consequences of existing judgements onto new cases.

Nevertheless, there is method in the madness and researchers have been working for some time to make methods explicit and amenable to software assistance.

One way is to accept that we will not in the near future be in the position to automate analogical reasoning and to leave that to “manual” analysis or “knowledge elicitation”, but at the same time have a clear methodology for doing such elicitation and a clear target formal representation that can be useful in automation.

Researchers use the word “factors” to denote the legally relevant aspects of cases which judges use and refer to in making judgements, and a lot of research has examined how to model these factors and how to reconstruct patterns of argument based on them, which can help understand the way in which a judgement might likely fall for a new case.

Example: Noise Induced Hearing Loss

If your work subjects you to noise, you may be in danger of losing some of your hearing unless your employer takes appropriate precautions such as providing ear protections.

If you suffer hearing loss and your employer has not taken all appropriate precautions, you might decide to take your employer to court. Whether you succeed or not depends on how the judge regards where your case sits with respect to the many cases which have already been judged and the judge will be isolating the relevant factors of your case in order to do so.

The University of Liverpool have developed a methodology for modelling the relevant factors governing judgements in this domain, the output of which is an “ADF” (Abstract Dialectical Framework). An ADF is an analysis of a case law domain which identifies the factors relevant to judgement, but which also provides rules to guide the evaluation of new cases.

Application of this methodology to the domain of Noise Induced Hearing Loss is the subject of the following co-authored paper : “Realising ANGELIC Designs Using Logiak” in JURIX 2019 proceedings, available as an e-Book: Michał Araszkiewicz, Víctor Rodríguez-Donce, Legal Knowledge and Information Systems http://ebooks.iospress.nl/ISBN/978-1-64368-049-1

Implementing ADFs

Even when the hard work of elicitation and modelling an ADF is complete, as described in the paper referred to above, there is a need for an implementation in software via which it is possible to input the case facts and automatically derive conclusions dictated by the ADF.

For case law implementations, the underlying reasoning engine has to be flexible. There is not an international standard which defines ADF reasoning, and researchers need to experiment flexibly, for example sometimes attaching weightings to factors and defining rules concerning how those weightings combine.

Such requirements are difficult and laborious to satisfy if writing code from scratch. Certainly, one would normally not want to write any kind of inference engine from scratch if one can avoid it.

Modelling case law reasoning in Logiak

With an ADF in hand, one can create an interactive case law reasoning system in Logiak with remarkably little effort.

Logiak has logical inferencing built-in, but at the same time the inferencing can be easily modified by configuration in order, for example, to achieve reasoning with weightings.

But the most significant feature for modelling case law is that Logiak imposes a clear distinction between declarative and procedural logic.

This means we are able to echo the structure of an ADF quite directly which makes implementation and maintenance simple, so the factor remains entirely intelligible as an ADF factor. Implementation of a system in Logiak maintains the basic logic at this high level. The “base factors” via which the case facts get into the system are just as easily implemented — namely as questions in a process.

Logiak is a system for creating interactive systems without programming. Once the ADF structure is mirrored into logic in the way described above, an interactive system exists: there is hardly anything more which needs to be done.

This means focus can be always on the nature of the factors themselves and how they combine. An implementation project is not the submerging and calcifying of a specific state of the ADF: Logiak permits the ADF to remain explicit and editable.

Conclusion

Logiak’s Process Logic combines formidable flexibility with simplicity. The task of creating and maintaining a Case Law system for evaluating whether certain insurance claims should be contested or settled using Logiak is very straightforward. The hard work is in the analysis and elicitation of an ADF.

In the future, we expect Logiak’s utility as a tool for the drafting case law specifications such as ADFs will emerge as significant— as with medical decision support, case law reasoning can only be enhanced with Logiak‘s use as a computer-aided specification tool.

What is particularly nice about No Code environments though is, with small caveats: once the specification is complete, the system is done.

Read about ADF implementations in more detail in the paper “Realising ANGELIC Designs Using Logiak” in JURIX 2019 proceedings, available as an e-Book: Michał Araszkiewicz, Víctor Rodríguez-Donce, Legal Knowledge and Information Systems http://ebooks.iospress.nl/ISBN/978-1-64368-049-1