If the title of this blog scared you, it was meant to.
In “The Future of the Professions,” authors Richard and Daniel Susskind, who are leading experts in law and tech, look at the impact of the current technological revolution on professional services.
Their starting point is that professional service providers need to ask themselves what their clients really want. For example, a patient visiting a neurosurgeon most certainly does not want neurosurgery – they want health. And many current technological advancements in the medical field are focused on preventative medicine and health promotion through, for example, providing earlier or more effective diagnoses, rather than curing ailments.
But the legal profession so far has been more reactive than “proactive” when it comes to embracing technology. Whilst consumers of professional services generally prefer a fence at the top of the cliff (i.e., problem-avoidance) to an ambulance at the bottom (i.e., problem-solving), lawyers have been preoccupied to date with embracing technology for greater efficiency, by (in the words of the Susskinds) “equipping the ambulance better than rivals or ensuring its arrival at the scene of the problem sooner than competition.”
A few years ago, I had the pleasure of hosting a fascinating session on AI and Arbitration in Singapore by Professor Dr Maxi Scherer, and I recall her humorous quip that “tech-savvy arbitrators are as rare as vegan butchers.” This comment may still ring true.
At the recent ICC Commission on Arbitration and ADR meeting in Paris, Richard Susskind invited those in the audience to consider what their clients really want in arbitration and ADR, and to come up with solutions to address what clients really want, rather than more efficient ways of doing what we have been doing so far. This will require a complete paradigm shift in the way we think as disputes lawyers. Are we ready for this challenge?
And what do our clients really want? Based on feedback collected from business users by the ICC Commission’s task force on “ADR and arbitration” (some of which is referred to in the task force’s Report on “Facilitating Settlement in International Arbitration”), business users want – amongst other things – more dispute prevention tools; more dexterity across dispute resolution options; and arbitrators who can encourage or facilitate settlement.
The dexterity to combine different ADR services allows dispute resolution tools to be tailored to the dispute and parties at hand, and may give parties the best chance of a negotiated settlement, alongside an opportunity to maintain their business relationships.
AI is developing at an exponential pace. A number of studies have demonstrated the potential for AI to predict judicial outcomes with a high degree of accuracy, based on previous decisions, as outlined by Professor Dr Maxi Scherer in her most recent article on this topic entitled “Artificial Intelligence in Arbitral Decision-Making: The New Enlightenment?” (published in ICCA Congress Series 21 (Edinburgh 2022)). However, as identified by Professor Scherer in the same article, there are a number of key obstacles to deploying AI for arbitral decision-making:
- Firstly, a large set of data is required to establish an AI model, and international commercial arbitration awards are generally confidential.
- Secondly, the underlying data used to train the model may have been “infected” with human biases, such that such biases may be validated and even exaggerated, as they are held to be “true” for future outcome predictions.
- Thirdly, AI programs are not designed to provide reasoned legal decisions: they are based on statistics and probabilities. Whilst the outcomes they predict may look like “intelligent outcomes”, the focus is on predicting the outcome, rather than the reasons influencing why previous decisions have led to similar outcomes.
Reasons are critical in legal decision-making where the stakes are high (unlike small-value disputes, where AI has been deployed for decision-making with some success, for example in consumer disputes administered by eBay and Alibaba). As pointed out by Professor Scherer, reasons add legitimacy to decisions, allowing the disputing parties to understand why the decision was made. Reasons also enable the disputing parties (as well as third parties aware of the decision) to adapt their behaviour based on the decision, and they enable other decision-makers to follow or depart from the rationale for the decision. Professor Scherer rightly concludes that “[t]he need for reasoned decisions is therefore likely to be an important barrier to AI-based legal decision-making.”
However, much of the focus in the discourse so far has been on replacing arbitrators with AI robots. But this focus falls into the trap identified by Richard Susskind of lawyers looking to deploy AI to do what we already do – just more efficiently – by replacing existing decision-makers with AI programs.
How can we deploy AI to focus on what users really want in the realm of dispute prevention and dispute settlement?
One very useful – but currently underused – dispute resolution tool, as referred to in the ICC Commission Guide on Effective Conflict Management (available here), is non-binding neutral evaluation, which can be used by one party, or all parties to a dispute. In this scenario, a third-party neutral is engaged to provide an evaluation of the parties’ differences in order to help promote settlement of a dispute. The evaluation will not be based on all the evidence available at the final hearing of the dispute, but that is precisely the point: the neutral will provide an evaluation based on the evidence and arguments presented to him or her, which will necessarily be more limited (and therefore cost-effective) than those presented at a final hearing.
One possible option is to deploy AI to provide “neutral evaluations” of likely outcomes, based on previous decisions. The use of AI could enable neutral evaluations to be made at high speed and minimal cost, and this may encourage more take-up of neutral evaluation as a dispute resolution tool to aid settlement. The lack of reasoning would not render such evaluations useless, as a predictive outcome would be valuable to the parties, even without reasons.
The challenge would remain that there does not currently exist a large data pool of commercial international arbitration awards, but this paradigm is changing. With databases such as Jus Mundi partnering with institutions to add anonymised commercial arbitration awards to their database, the data pool of commercial arbitration awards is growing, and rapidly. Further, as pointed out by Professor Scherer, institutions could make awards available for the purpose of building AI models, even without publishing them.
The problem of hidden human bias in commercial awards would remain, but the examples most frequently cited of potential hidden bias in arbitral awards tend to be from investment arbitration rather than commercial arbitration. It remains to be explored whether human bias will have a significant impact on the outcomes generated by an AI predictive model based on commercial arbitration awards, and if so, how that impact could be mitigated.
The idea I have set out above is just one potential suggestion as to how AI may be deployed to assist disputants in preventing or resolving legal disputes. There are no doubt many more ideas out there. This blog serves as a clarion call to encourage all of us in international dispute resolution to collectively and creatively focus our minds on how AI may be deployed to help solve our clients’ problems and give them what they actually want – not what we think they want, and what we have become accustomed to providing to them in the past.
If there is just one point you take away from this blog, then let it be this: artificial intelligence is not a spectator sport. It is time to for all of us to jump on the train, or risk being left behind – or even worse, becoming extinct.
*This post represents the views of the author and not any organisation she is affiiated with.
________________________
To make sure you do not miss out on regular updates from the Kluwer Arbitration Blog, please subscribe here. To submit a proposal for a blog post, please consult our Editorial Guidelines.
I couldn’t agree more – a well-known quote by a former US Chief Justice rubbishes the notion that most people want back-robed judges, well-dressed lawyers, and fine panelled courtrooms as the setting to resolve their disputes. ‘People with problems, like people with pains, want relief, and they want it as quickly and inexpensively as possible.’ As you suggest, AI provides a credible solution that in many cases, cost-conscious, business-focused parties in dispute will readily move towards if it means they can quickly draw a line under a problem and look ahead instead of retracing historic events at great cost. Neutral evaluation using AI at an early stage to aid settlement, perhaps even as issues are identified could prevent disputes from ever escalating and save parties significant sums and time. As for access to precedent, while this will in any event improve in time, in theory, there is no reason why AI should not have access to just as much admissible precedent as any counsel in a full-blown arbitration or court hearing. And as regards the potential risk of hidden human bias in awards & judgments, this remains an issue for tribunals and judges as well – but again in time (although maybe not so much time), some may prefer to take their chances with AI…cost-effective and unreasoned rough/commercial justice.