The Mind of Lord Denning

The next competitive edge has arrived.

Knowledge of the future is not in technology. It is in the minds of experts such as legal or medical practitioners. It is in books, writings and judgments. Knowledge on strategic thinking, for example, may be garnered from strenuously studying the 2,500-year-old Sun Tzu's Art of War. However, even better than memorising 5,000 characters in the book is to explore deeply the mind of Sun Tzu.¹

By the same token, I argue for the development of expert systems embedding legal reasoning through a study of how the minds of judges work. This article provides a background to my research work, Do Judges Think? Essentially, what this article aims to show is how the expert system relates to judicial thinking. To illustrate this further, a simulated case study (which I have named 'The Experiment') based on a structured, model-based approach for investigating the judicial mind of Lord Denning is adopted.

Mapping the Thinking of Judges
This article is a spin-off from a project entitled, Do Judges Think? It began in May 2001 at Chicago-Kent College of Law with an invitation from the International Symposium of Judicial Decision Support System ('JDSS').² The thrust of my proposal is for JDSS to flow from judicial thinking. For a surprising 30 years, judicial decision had been a potentially viable avenue - and market - for the application of artificial intelligence ('AI') tools.

Ideas evolved as I spoke on the topic. Under the aegis of the IT Committee of the Law Society of Singapore, I explored the notion of mapping the thinking of judges. Pursuing Do Judges Think? led me to engage in lively debates at the NUS Law Faculty, which were enriching on the whole.

Co-relation Between Expert System and Judicial Thinking
Ongoing, pervasive changes in technology have impacted legal practice in Singapore. Singapore now has the most efficient justice delivery system. Our courts in Singapore are technologically so transformed that the next frontier has to be the application of AI. The challenge is for computer engineers to develop the AI tools to create an artificial judge: one capable of complex, legal reasoning.

Jurisprudentially, the adoption of the AI approach to delivering judicial decisions ought to be consistent with the normative model of judicial outcomes (see Figure 1). In terms of an equation, this could be represented as:

Judges = Judicial Thinking
Judicial Thinking = Expert System
Expert System = Artificial Judge

Figure 1

Under the normative model, human judges are bound to think by certain rules. This is what whets the appetite of the expert system builders. They relish the challenge of shelling out a rule-based expert system.

If you believe that all judges must be homo sapiens, then beware. For if all a judge does is think within a tightly prescribed, logical, serial framework of 'If this, then that', then one day he or she may no longer be required. With today's technology, it is not impossible that an artificial judge may be able to render such decisions - and perhaps even far more efficiently.

In practice, rule-bound modes of judicial thinking may not always be true for all judges. Judges may not, despite what the jurists may propound, feel bound by precedents. Half a century ago, the House of Lords decided that they be not absolutely bound by their own decisions. How wise!

The Experiment
With The Experiment, I chose to begin with data that was on hand. I exposed my audience in Singapore vicariously to an in-depth interview led by law Professor (and Queen's Counsel) Smith with the late Lord Denning.³ The purpose was for participants to feel how the mind of Lord Denning works in the process of legal reasoning. The subject matter was the High Trees case⁴ and the doctrine of consideration.

At the outset, it must be said that Lord Denning is not a typical judge. Lord Denning has long been an icon among lawyers within the common law tradition. In the words of Professor Smith, Lord Denning had made 'extensive contribution' to developing the law of contract. Indeed, he had brought to bear upon the law a certain warmth, freshness, vigour, vitality and, most importantly, humanity. At the interview, he showed himself to be most charming and captivating. Interestingly, someone remarked, 'Oh for the law students, they just love Lord Denning. He stood for them justice - not what the law is but it ought to be'. Few judges are equally enduring and endearing.

The Experiment was grounded upon Wahlgren's automated legal reasoning (see Figure 2). The parsimonious model captures judicial reasoning as thinking: 'case' for 'situation'; 'reasoning' (in judgment) as 'decision'; and 'rule application, evaluation, formulation and learning' as the complex of interactions to be discussed. My theory of judicial thinking recognises the complexity of human brain functions ('complex neuronal processes'). Indeed, judicial thinking may usefully be mapped with the latest fMRI (functional magnetic resonance imaging) technology. In essence, judicial thinking is a dynamic, interactive process of mapping a given case onto rules. Rules that may be discovered from the analysis of statutes and cases.

Figure 2

Inside the grey matter of his brain, the judge weighs reiteratively a given case (say, the High Trees case) against discoverable law (doctrine of consideration). Through highly interactive processes as indicated in Figure 2 (see the two cursive arrows in the diagram) 'emergent thinking' arises. Based on such thinking the judge reasons his decision. There are two dimensions to thinking - the conscious and the unconscious. The actual writing of the judgment constitutes the more conscious dimension of judicial thinking.

The Mind of Lord Denning
On the whole, The Experiment successfully mapped Lord Denning's judicial thinking upon the parsimonious model (see Figure 3). The normative model of judicial thinking did not hold in this case. With Lord Denning, a just outcome was paramount. For him, justice was to prevail over certainty of law. Lord Denning could easily have kept to the classical, traditional, old common law rule in the High Trees case. He stated just as much in the opening line of his written judgment:

If I were to consider this matter without regard to recent developments in the law, there is no doubt that had the plaintiffs claimed it, they would had been entitled to recover ...

The outcome would be the same as in the normative model if someone had engineered an artificial judge modelled on the long line of past cases. The rule was as obvious as it was programmable - 'consideration was required for contract'. That rule may happily be engineered in the expert system as legal knowledge in the style of a straightforward dichotomy:

Is consideration present?
If yes, there is a contract.
If no, there is no contract.

Instead of reinforcing upon certainty of legal outcomes, Lord Denning argued persuasively for justice to prevail. In the High Trees case, Lord Denning had equity prevailing over common law.

The documentation by video was most useful in observing the facial expressions and gestures of Lord Denning. His manner of speaking, tone of voice and even silences yielded a sense of his thinking. Like the revving of search engines, these give clues to how intensely data was being processed.

Theory of Back Propagation
Even more interestingly, one IT-savvy legal practitioner suggested Lord Denning's thinking process to be back propagation. Technically, this is the name for the most popular learning algorithm that underpins neural networks. Here are some examples of Lord Denning's train of thinking as implied by the words and phrases that he used during the interview:

• '... dug up all these old dictum ...';
• '... call in aid ... Law Reform Committee ...'; and
• '... old case in House of Lords overlooked for years ...'.

His mind worked by searching back to justify his good (in the neural sense of the word) output or what he felt as the right judicial decision. This is exactly how neural network ('NN') works. The artificial brain goes backwards from good output by putting synaptic weights to neurons. Should this not be expected? After all, neural network was modelled upon the same principles as our biological neurons. With technology advancing so rapidly, the back propagation power of electronic neurons within an artificial judge may reason out a just decision. Technically, an artificially intelligent judge is so 'trained' (a terminology of NN) that its pattern of synaptic weights learns to become identical to Lord Denning's. For in terms of NN, intelligence is but a distributed set of synaptic weights!

The Court as a Thinking Organisation
According to Dr Lawrence Farwell, every thought of ours is imprinted somewhere deep in our brains.⁵ And there is actually a patented technology for retrieving such storage! In my working paper, I had argued for the court to be perceived as a thinking organisation. If so, expert system may be utilised to embed past legal decisions as knowledge. Perhaps, it is timely for legal practitioners to reflect upon roles for the artificial judge in our legal system.

Embarking upon an empirical programme of research requires a team of enthusiasts. Building an artificial judge in any given legal domain has to be a collaborative enterprise comprising the software engineer, legal expert and judge. An expert in the domain area must provide the framework. Then, a software engineer models it. Following which, we need to run tests by parallel experiments to see if the judicial decision of an artificial judge is identical to that of a human judge. In certain domains of the law, this may one day be realisable.

Foo Check Teck
Nanyang Technological University
E-mail: mctfoo@ntu.edu.sg

Endnotes