Am at the BILETA conference, liveblogging the sessions I’m attending. It’s being held this year at the Universidade do Minho, Escola de Direito, and the conference title is ‘International perspectives on emerging challenges in law, technology and education’. After the welcomes, Burkhard Schafer, from SCRIPT in Edinburgh U, on ‘Creativity in AI and Law’. Some kind words from Burkhard about our shared BILETA past… The questions he wanted to ask — what can AI teach us about creativity in law and legal education? What type of legal AI should we build?
Burkhard started with AlphaGO — it brought AI into the limelight, he said; and with it the slew of questions raised by AI’s presence in society, eg risk of human employment replaced by AI. Creativity is the dividing line — he quoted a blurb from California Western Law School — ‘the creative problem solving area of concentration is designed to help you acquire the needed skills that demand broader and deeper understanding of people, their problems, and the consequences of confronting those problems only in narrow, legalistic ways’. Interesting opposition of legalism vs creative problem solving. What does creativity in law actually mean, he asked? For academics? For lawyers? Can research in AI creativity actually help us to define what creativity in law can be.
Law and AI began, according to Burkhard, with Abba ben Joseph bar Hama (or Rava, according to the Talmud), around 300CE. Interesting discussion by Burkhard about the lesser jurist/judge-as-human vs ultimate judge, God. Recurrent themes in AI since then include –
- legal AI as perfect rule following
- wilfully non-creative
- natural language processing/dialogue
- knowledge acquisition bottleneck
Burkhard notes how the scroll is a form of blockchain — at last, someone else is taking the concept of blockchain seriously in terms of its communicative and process-driven power. I’ve discussed this elsewhere on this blog. He moved on to Voltaire, setting the old paradigm of machine rationality vs artistic creativity, which is embedded in case law — above all, it showed the a judge is a human being, not the type of unfeeling robot some would expect the judge to be’ – Keppel v BaRoss Builders, Inc.
Why do we want the human touch? Burkhard compared Dworkinian fear of the human element in Hercules with Hartian discretion as an intrinsic element of the process of judging. Traditional AI is philosophically Dworkinian. He took a case study of algorithmic music, eg Nikolaus Simrock 1797, on composing forty five trillion different waltzes without ‘understanding anything about music or composition’. B. quoted Neil Cuxbury on Random Justice as an example of this in jurisprudential thinking. Randomness though can be chosen as a form of case solution — it becomes a tool to overcome shortcomings in information. But legal AI doesn’t normally involve throwing dice to reduce complexity. Could one make the argument that randomness is sometimes good? B. points out how evolution involves the element of the unplanned and the unplannable — it is random in its direction. He compared this to the sources of error in commodities trading, and the patterns of trading in the global financial crisis. He then quoted Pennsylvania v Local Union, 1974, arguing that by the subtle tone of their objection they demonstrate either that they want black judges to robots who are totally isolated from their racial heritage and uncorked about it, or, more probably, that the impartiality of a black judge can be assured only if he disavows, or does not discuss, the legitimacy of blacks’ aspirations to full first class citizenship in their own native land. The critical issue is, what conduct by black judges will assure their impartiality? Should they be robots? Should they demand their heritage by asking for less than first class citizenship for other blacks?
For B., this overestimates robots. Because we need to look at how the robots are being coded, with human coding prejudice encoded in them. And it also underestimates the creativity of AI researchers in embodying the values of their society, social contexts and values. Should we build legal Ads with political etc preferences and biasses? So as not to overestimate their impartiality? Because we consider this in a pluralistic society a good in itself? ‘In China, Go is not just a game, it is a mirror of life’. Is the inverse true? (This is particularly true of simulations and games — they mirror choices and created pathways, created [in the best sims and games] by the learners themselves).
AI in its way can be an idiot savant — but what’s the difference between making a mistake and being a genius. Creativity, rather, can be seen as conceptual transfer, where we see new connections, and ‘mosaic’ in patent law. He cited Schafer and Taks on CoReO, and he pointed out how pace and time, in slowing down participants in a legal process can be useful, as is creativity and slowness. He cited Kwiatkowski, Vartanian, Marindale on creativity and speed of mental processing, and applied this to legal education, arguing for slowness in the curriculum. How do we teach patience, eg? In playing Go with a machine the player actually became more creative, had to think out of the box. No human or machine, nor human against machine, but human together with machine — resilience, from being thrown by the machine, to working with the machine to be better humans, better in our professional activities.
Lots to think about here, eg the blockchain point B. made. Actually (and this is my view written during questions), this is a fascinating point. Compare a blockchain contract with, eg Accursius’ Great Gloss. The critical thing about the blockchained contract is that it remains as it was when it was signed, and can be built upon because of that transparent reliability. But the Great Gloss, while a form of fixed commentary on the textura, actually was never fixed, as glosses were never fixed, because always in fluidity, in motion, as new arguments surface, older ones submerge. Could blockchain therefore be a record of that form of dynamic process? And can blockchain then act as a robot in legal education, offering advice, assistance, as I predicted could happen in the final section of Transforming Legal Education?
Fascinating keynote, one of the best I’ve heard at BILETA. Worth coming all this way just to hear this alone.