Recently my colleagues Leah Wing, Chris Draper, and Scott Cooper, and I began discussing the possibility of governing the development and adoption of artificial intelligence applications that would be consistent with the ethical use of technology in dispute engagement of all kinds. Those discussions resulted in the publication of Governing Artificial Intelligence1 earlier this year. What prompted us to think about this?
There is an adage in interpersonal communication theory that says “one cannot not communicate.” We suggest it is clear that, today, one cannot not talk about AI. This is certainly true for mediators, arbitrators, and lawyers who are confronted with an array of AI apps and platforms that do everything from intake and scheduling, to recommending resolutions, to handling agreements and billing. Some are built specifically for mediation, arbitration, or the law, but many are generally available apps that assist with some of the functions we routinely perform as dispute engagers (managing communication, dealing with data and information, and managing group dynamics). One thing is for certain: in the coming years there will be a large number of AI apps targeted directly to professionals in one form of dispute engagement or another.
It is reasonable to ask how closely the variety of AI apps that will be available to us track directly with the ethics and standards of practice for dispute engagement. It’s also reasonable to ask how dispute engagement professionals can generate a voice that is heard by the developers of AI, many of whom, with some notable exceptions, are not in any of the dispute engagement fields.
There have been extreme reactions to the development of AI: Either the fabric of society is going to be torn apart by AI (the Terminator fear), or we are headed for an Edenic existence where AI smooths out all of our problems (the “LaLa Land” hope). Certainly, we have heard both extremes (and some in the middle) in the discussions surrounding the use of AI for practices like mediation. It will take the “human” touch out of mediation and it will put us all out of work (the Terminator), or it will integrate into our processes and allow us to do more, quickly, efficiently, and with better outcomes for our parties (LaLa land).
The fears and hopes for AI echo the fears and hopes that have been expressed for major tech developments in the past. Printing was going to destroy memory, the industrial revolution was going to destroy creativity, etc. We have gotten used to each of the technological advances and we have integrated them into our societies. But the uncertainty about AI remains: As Geoffrey Hinton, 2024 Nobelist suggested, development of AI that thinks like humans “will be comparable to the Industrial Revolution. Instead of exceeding people in physical strength, it’s going to exceed people in intellectual ability. We have no experience of what it’s like to have things smarter than us.”2
The truth about our hopes and fears is that reality lies somewhere between, at least in the short term. Although considering our recent history with technology we can be sure that there are “bad” uses of AI that will proliferate – as a clue, remember that the creation of the Internet brought us immediate growth in online gambling and pornography.
The key question my colleagues and I have wrestled with centers on governance: essentially the questions are: can AI be controlled? and can it be responsibly integrated into our practices?
If by “control” one means enacting hard constraints around the development of AI systems, and in effect blocking all of the negative and destructive uses of AI, the answer is a very hard no.
If by “control” one means providing clear ethical guidance to developers, providers, and users of AI, the answer is at least a qualified yes.
The question of how to “control” or govern technology is not at all new. As Bob Bergman often points out, the discussion of controls on technology can be seen to start in 1942 with Isaac Asimov’s ‘First Law of Robotics’ – “A robot may not injure a human being or, through inaction, allow a human being to come to harm”.3 That notion of preventing harm is at the heart of Chris Draper’s definition of safety, and is an ethical imperative for those of us involved in dispute engagement: safety is reducing the potential for negative outcomes to an acceptable level. If you can make it safer for the parties, you are ethically required to do so.
So, what must we do to govern AI and make its ethical use in dispute engagement the norm?
First, give developers a good sense of what is ethical and marketable to us. Developers make choices about how to structure AI applications. What we can do is give them an idea of what ultimate users of their apps want, and how to develop the apps with our ethical standards in mind.
Second, give mediators and other dispute engagement practitioners a good set of ethics and standards that address technology generally and AI in particular that will let us choose AI platforms and products wisely. The NCTDR/ICODR standards that Leah Wing and I were active in creating are an example – they are one expression of what is important to those involved in dispute engagement.4
What are the domains in which the guidance for developers and users can be developed? There are three domains in which the voices of dispute engagement professionals can and should be heard.
Regulation/Law: Regulation and law exist in the domain of what Scott Cooper calls de jure control: the complex of controls that are created within legal and national borders and which carry the weight of the courts, and have consequences for those who violate those controls. De jure controls can be effective within their scope, but they are by definition limited to the venues in which they are created. The EU, arguably, has embarked on the most aggressive and useful set of regulations of AI, but their application is questionable even within the EU, and are voluntary outside the EU. Law and regulation can address the safety issue, the harm that can be caused by the ill use of AI. In Italy, a recently passed law makes it a felony to create AI material that harms an individual. Of course, defining what harm is will enrich a cadre of lawyers, and may eventually create a set of precedents that define harm in that context. But, for the most part, law and regulation are captive to interpretation and are behind the curve of the rapid development of technology. And the political process that develops laws and regulations is not targeted to address the ethics and standards of practice for our field. Even so, dispute engagement organizations may lend their voices to the political discourse that leads to regulation.
International Standards: On the other hand, there are a number of de facto standards that address AI. The ISO alone has issued at least 12 such standards.5 The strength of international standards is that they cross borders and are most often “adopted” by a number of the 175 member countries, or by members of other standards organizations. Of course, the efficacy of international standards is constrained by the fact that they are, in the end, voluntary, take a long time to develop, and, again, are not generally specifically targeted to the ethics and standards of practice of our field. But, as with the development of the ODR standard, ISO 32122, dispute engagement professionals can have a voice.
Professional Standards: Another set of de facto standards are those developed by professional organizations. For example, the ISO 32122 standard contains almost verbatim the standards of practice developed for online dispute resolution by the National Center for Technology and Dispute Resolution (NCTDR) and the International Council for Online Dispute Resolution (ICODR), and the same set of NCTDR/ICODR standards were incorporated into the ABA’s Section of Dispute Resolution ODR guidance. The power of professional standards is that they are developed by practitioners who have a very focused understanding of what is important, even mandatory, in the performance of their professional work. Another strength is that they also cross borders: a mediator in the UK will not have a set of practice standards that radically vary from those in the US or the EU. Professional standards are, of course, voluntary, but failing to adhere to them can have direct repercussions for “rogue” practitioners. And, importantly, dispute engagement professionals are the sole voice in determining the standards, as would be professionals in any other field. They are targeted, and easily communicated to app developers. The NCTDR/ICODR standards are unusual in that they not only cross borders, they cross disciplines.
What we should aspire to, in all three of the domains, is a set of regulations and standards that give guidance to those who wish to use AI in an ethical manner. There will always be bad actors, and no matter what is done in the way of law or standards, bad actors will act badly. The “wild west” environment in which AI has to date been developed can be argued to encourage bad actors by making powerful AI capabilities available to basically anyone.
For example, in Louisiana, a 13-year-old young woman was expelled for starting a fight on the bus going home from school. The person with whom this child fought was another 13-year-old in the act of sharing AI generated deep fake nude photos of her with his friends on the bus. It was only when the girl’s family sued the school board for wrongful actions that the fake nude and the sharing of the image was brought to light. At that point, the young man was expelled. How the case will play out in terms of penalties will be determined by the courts as the issue moves along. Such is the state of law and regulation. What is interesting about this episode is not that bad acts occurred, but the fact that a 13-year-old had open access to tools that allowed a realistic damaging image.
The reality is that none of the three domains work alone to deliver governance for AI. Taken all together, mixing de jure and de facto guidance, there is hope, and therein lies the qualified yes that I offered at the beginning.
It is our argument that control (governance) of AI will be in large part a function of the marketplace. All AI developers want to make money – what the users (consumers) of AI platforms will buy will help determine who “wins” and what successful AI platforms look like. What that should say to those of us involved in dispute engagement is that we should make sure we have a loud voice in the discussion about what regulations make sense, what standards we demand for our AI apps, what ethical AI looks like – and what we will buy.
As we seek governance or “control,” if the users in the equation – us – are not heard, we will get what we get – caveat emptor.
Baltimore Injury Lawyer Blog by John BrattSometimes you find great information in unlikely places. Today I got a spam email trying to sell me some editing software specifically for lawyers....
By John BrattIt is a difficult task comparing the cost of using Small Claims Court versus Mediation to resolve a dispute since there are myriad factors to consider. Much of the cost...
By Lance SoskinThis video is presented as part of Mediate.com's 25th Anniversary Conference. Link to downloadable/embeddable version here. As dispute resolution offices around the world are switching to Zoom for meetings, this...
By Clare Fowler
