Swapping Decision Trees for River Logic

My experience to date with legal knowledge engineering has consisted of using decision trees to automate legal documents in a field known as document assembly. I have never done hard coding or played with expert system shells. Indeed, there are not many of them to play with. The only ones I am aware of are those developed by Neota Logic (formerly Jnana) and RuleBurst (since acquired by Oracle).

So it was interesting to meet Dr. Pamela Gray, a legal knowledge engineer from Charles Sturt University, and her son Xenogene Gray, a computational physicist. Together they have developed a legal expert system shell called eGanges that relies not on decision tree logic but rather on something called river logic, a refined Ishikawa fishbone.

The Ishikawa fishbone was developed by Kaoru Ishikawa of Japan in the post World War II era. It originally was used in quality control as part of quality defect prevention. Visually, a fishbone diagram looks like a fish skeleton, with the end result, or effect, forming a fish’s head and the various causes of that effect forming the skeleton of the fish. The deeper into the causes you go, the finer the bones of the fish.

Another way to visualize the Ishikawa fishbone is as a river. eGanges has adapted the Ishikawa fishbone to apply to legal problems but it uses the metaphor of a river rather than a fish skeleton. Legal rules flow downstream until they achieve a particular outcome at the river mouth.

River logic works in some ways like a decision tree in reverse. You can begin at any point of the river, but if you begin at the mouth of the river (the final result) you can trace the logic back up the various tributaries, then to the creeks, and ultimately to the different sources of those creeks. As you progress upstream, you may answer the questions put to establish the points for the user’s case. At any time during a consultation with an eGanges application, the cumulative result can be viewed for your case. When you are finished, you have a complete list of the factors that led to the final current result.

Take, for example, a breach of contract case. The outcome at the river mouth is predetermined as entitlement to compensation. The rules which lead to the entitlement constitute the river system. The main tributaries of the river might consist of contractual rights and performance. Along the contractual rights tributary might be nodes such as time provision, notification term, and rights term. Alternative creeks might flow into the rights term, such as right to compensation and right to be reimbursed for payment of expenditures.

It feels easier and more intuitive to map out legal problems as rivers rather than as decision trees. A decision tree has no logic structure, just backward and forward chaining. You have to build all the logic and spell out all possible alternatives for each node. With river logic, you don’t have to be as explicitly detailed. eGanges adds intelligence at the node level. There are built-in heuristics, which is another way of saying that the machine is able to work out implicit knowledge. Each node has four possible answers: yes, no, uncertain (there is evidence of uncertainty), and no answer (there is no evidence). Uncertainty identifies areas where further evidence is required, due to such factors as the memory failure of a witness, conflicting evidence of witnesses, or lack of credibility of a witness. The software keeps track of the answers for each node.

In our previous example, if work is required to be done to effect performance of the contract (so that you can ultimately claim the compensation), then the river mouth outcome will never be positive unless the work node is also positive. If you change the work node answer, the river mouth outcome will update automatically.

There can also be a gloss, or annotation, placed on each node, similar to the judicial glosses that we are familiar with in court decisions. In addition, there is an option to insert neutral nodes. For example, in contract law, a request for further information about an offer does not terminate the offer – it does not have the same consequence as a counter-offer. Thus, it does not matter to the formation of a contract whether or not there is a request for further information. 

River logic can be used to build a case against an opponent, to ferret out the logical errors in new statutes, or to automate the interpretation of a particularly complex statute. Working with well-known Sydney solicitor Philip Argy, Dr. Gray built a Spam Act application using eGanges several years ago.

Here is a link to a sample eGanges applet:

Free copies of the software are available in return for providing Grays Institute with an application that can be posted as an applet on its planned new website, The Grays Institute website will be open for free public access in January, 2012.

Comments are closed.