Artificial intelligence in dispute resolution
Where we are and where we’re going
By Rick Rogow
One concept that seizes popular imagination is the idea of the digital judge. There is something appealing about the concept that one day our frustrating, chaotic human disputes will be resolved by the cool, all-knowing rationality of a fair and impartial electronic decision maker. Perhaps.
Since moving my mediation practice online in March 2020, I’ve learned a great deal about online dispute resolution (ODR). It wasn’t a big leap from my experience with video conferencing to being able to conduct a full-on mediation process through video and audio virtual conferencing with multiple parties. It would have been my loss if I had ignored the gains in technology.
Technology, Dispute Resolution and the Fourth Party
Online Dispute Resolution takes the technology another step. By Online Dispute Resolution, I refer to augmenting the use of communication and information technology with “machine learning” to help people resolve disputes. In current ODR processes, this technology is most often used as administrative support, handling tasks like case filing, reporting on statistics with some analysis and facilitating communications. It’s a tool. But obvious to those in the ODR field, computer processing is unimaginably powerful right now. So that available technology is not just a tool, it becomes a fourth party to the negotiations, capable of being a collaborator as well as so much more.
What does the term artificial intelligence (AI) refer to? How does it relate to machine learning? How are these developments used in Alternative Dispute Resolution (ADR)? And, where is it going?
Getting Used to the Machines
The expansion of technology into realms of human behavior always scares me at first, but then it intrigues me. I analyze things pretty much the way I always have, but my keen analytic skills have cognitive biases and attribution errors. But, “AI” doesn’t go about formulating its outcomes in the same way as our human intelligence. AI learns in a very particular way. It learns by looking at data, and it derives rules from that data in order to gain insights and/or to make predictions.
An important characteristic of machine learning and AI is that this data must be structured, pre-structured, into a format that the AI can make sense of; only then will the AI’s output be coherent or useful. AI algorithms are still not very good at making sense of unstructured data.
In 1999, eBay commissioned the Amherst Center for Information Technology and Dispute Resolution at the University of Massachusetts to conduct a pilot project to mediate disputes between buyers and sellers. The pilot project handled two hundred disputes in a two-week period, by far the largest number of disputes ever handled online. This capability led eBay to include dispute resolution as an option for buyers and sellers in the event that an online transaction was unsatisfying. By 2010, the number of disputes handled by eBay had reached an extraordinary 60 million.
AI and Empowerment
To better understand the DNA of AI, it helped me to think of the cognitive tutor interface, significantly advanced at CMU in 2007. As those who might be learning a foreign language interfacing with their computer, these systems provide context-sensitive hints and instructions to guide students towards reasonable next steps. A continuously-evolving decision tree. These AI systems are helping musicians co-create new music they admittedly wouldn’t have discovered on their own. This machine-learning is currently a resource in disaster claims with insurance companies. But what of high dollar-value, contentious conflict with complex histories; or claims wherein one’s legal rights, or human rights, are likely to be impacted?
Is taking an individual through a logic tree to clarify their resolution options and legal rights risky business? At what point does the consumer, the citizen, the client become vulnerable to the programmer, to the database, to the machine and lose their agency?
AI and Equity Into the Data Sets
The need certainly is there for equitable access to justice. National judiciaries and international tribunals are seeing AI and ODR with strong potential to bolster and provide equitable access to their judiciary. Countries such as Canada, England, Brazil, European Union, India are currently developing beta systems hoping to significantly increase understanding and access, to provide more consistent, even-handed justice and at less cost. Those are certainly admirable goals.
Collaboration in Design
“Automation bias,” is another term in machine learning worthy of highlighting. Many are skeptical and concerned, including me, about ethical vulnerabilities and inherent limitations of these data sets. How are legal rights and perhaps basic human rights impacted without the ‘informed” user’s knowledge? What protocols are in place to monitor and audit the information and data structures? There is a fairly significant difference between machine-learning systems which collaborate with a user learning a language or a musician using the data bases in the creative process and AI interfacing with claimants on high cost-value injury and dispute resolution options.
I see AI and machine learning as an extraordinary resource just as Lexis-Nexis was in all its availabilities. I’ve been empowered by that technology and access to information. In an odd and unconscious way, it has become an indispensable collaborator in my decision-making processes.
One of my favorite quotes is, “It’s not the genie that’s dangerous, it’s the unskilled wisher.” (Cassie Kozyrkov, Chief Decision Scientist, Google.) As legal practitioners, we intuitively understand how critical it is to know what questions will be asked so that the hoped-for outcome isn’t an unintended consequence. This seems to be a key take-away.
As professionals doing our best to neutralize bias in conflict resolution, our challenge is to assist our clients and the courts in preserving the quality of our processes and of our court-related protocols. This challenge should draw directly on our strengths and experiences: to ask the right questions and keep on asking.
Attribution: I received considerable help over the past 18 months in understanding the concepts of Online Dispute Resolution and the developing use of AI in ADR by my good friend and colleague Colin Rule. Colin, among his significant contributions to our field, is currently President and CEO of mediate.com, and Co-Chair of the Advisory Board of the National Center for Technology and Dispute Resolution at UMass-Amherst; received the Frank Sander Award from the American Bar Association in 2019 and was Director of Online Dispute Resolution for eBay and PayPal 2003 - 2011.