The Journey to Solomon

The roots of the present day Solomon Project stretch back two decades to another effort, “The Mediation Project” at Santa Barbara’s Family Therapy Institute. That project was designed by my mediation partner Laura Collins and I to bring the benefits of mediation to divorcing couples who could otherwise not afford them.

The project was very successful in attracting demand for its services, which were free or very low cost to the clients of the clinic. The desire of divorcing couples in the community for mediation was overwhelming; divorcing couples lined up to participate, mostly parents, often hoping to end their marriages amicably for the sake of the children. Laura and I mediated these divorces as efficiently as possible to keep up with demand on a one-day-a-week volunteer schedule.

Something else was happening at that moment that affected our mediation practice in Montecito and the fate of the Mediation Project as well. Awareness of mediation as a far better process than litigation for many divorcing couples was exploding, and the lines at the Mediation Project were stretching far beyond our ability to accommodate them. At the same time, growth at our Montecito practice was challenging our ability to stay on top of our cases in the four days a week we had available.

The Mediation Project ran for two years at the Family Therapy Institute, but the writing was on the wall; the interest in mediation in the community was outgrowing the Mediation Project’s ability to make a difference. As we witnessed the expanding awareness and interest, we were heartened by the recognition that more and more people wanted to work through their differences and come to agreements without doing battle in court. But two volunteers working feverishly to accommodate that new demand was not a realistic answer. We finished up our ongoing cases and closed the Mediation Project’s doors.

But the intuition lingered. There was something important missing in the legal system —in our culture— that the kind of personalized, legally and psychologically sophisticated, high quality facilitation that we offered to our clients in Montecito, could be a game changer if it could be made available at minimal cost to nearly anyone who desired it.

Fast forward to a couple of years ago, when the rise of the generative AI models began to suggest that a different kind of highly intelligent, always available, and even somewhat emotionally perceptive assistant might be made possible by AI. The idea of the Mediation Project now resurrected itself in a new form: could an application powered by a Large Language Model step in to provide at least some of the functions of a lawyer/therapist team like ours?

So, The Solomon Project was conceived – formulated unimaginatively at first as IDRA, the Intelligent Dispute Resolution Assistant. We were at that point thinking only of an assistant to expedite a divorce mediation. Envisioned as a stand-alone application, the AI assistant would allow a mediator to complete cases in relatively few sessions, allowing them to handle more cases, much more effectively, efficiently and affordably.

In addition, this vision inspired us to think that we could again make mediation services available to lower and middle-income couples who desired the safety and relative peacefulness of a “conscious unmarrying.” And we could achieve this, not by working furiously to expedite cases, but by structuring a system to take over much of the work of gathering information, conceptualizing a case, educating clients about the processes of mediation, dispensing legal information (not advice!), and generating creative settlement options. Even more importantly, as we began to understand the evolving capabilities of the AI models, the AI assistant could develop and present at least a preliminary analysis of the goals and needs of our mediation clients, along with a preliminary picture of each client’s emotional state and psychological traits.

This is where we are with Solomon Now: developing an MVP that evaluates the strengths and weaknesses of different AI models using an architecture that calls tools and augments retrieval with a growing store of embeddings, so conversations better capture users’ needs, analyze interactions, and continuously improve.