Artificial Intelligence in Mediation: The Future Has Not Been Written

While everyone knows that AI technology can easily vanquish the greatest human chess champions, a constructive discussion into whether AI will dominate the mediation profession may be inspired by consideration of a simulation (albeit one of purely satirical provenance).

Assume that in early 2023, the GabADR program introduced artificial intelligence to commercial mediation, as its creators promised “a seamless integration of AI technology and human interfaces positioned to disrupt the traditional Alternative Dispute Resolution model.” How might the roll-out have gone?

Simulated transcript begins …

Plaintiff’s counsel: So, our damages exceed $5 million, but we’re willing to open with a demand of $4.5 million.

GabADR: the optimum opening demand would be $2,477,777.

Plaintiff’s counsel: Very generous of you. Mind dishing how you get there?

GabADR: Algorythmically, counsel. More specifically, this case optimally will settle at between $2.2 million and $2.3 million. Therefore, an offer ten percent above that amount should attract an offer also within ten percent, thus facilitating a classic “gravitational pull.” Also, a century of sales data shows that consumers find prices ending in sevens nearly irresistable.

Plaintiff’s counsel: You might possibly be right about us settling at $2.2-$2.3 million, but not if I open with 2.5. I’ve known this defense counsel for years. She’ll take that opening as a signal from me that I’m prepared to go below $2 million – which I ain’t. So take the $4.5 million demand to them please.

… 15 minutes later

Defense counsel: So, $4.5. (Turning to client: “I win our bet”). OK, so how strong do you think our defenses are?

GabADR: You have a .00002% chance of survival. No whoops, that question somehow triggered my “Spock mode.”

… recalibrating …

Ah yes, you have a 48.7% chance of prevailing on at least one of your defenses.

Defense counsel: I feel significantly more confident than that. As you know, this is a bench trial matter and I’ve been before this judge at least a dozen times, and I’ve always backed up my claims and never over-stated my case. Does your algorithm include a factor for my credibility with the judge?

GabADR: No, that’s unquantifiable. I did, however, factor in that this judge has denied dispositive motions in 58% of cases bearing a greater than 90% correlation with the factual-legal matrix in this case.

Defense counsel: Sounds about right, but I won't be relying on dispositive motions -- too easily overturned on appeal. Anyway, I’m authorized to extend an opening offer of $95,000. You can tell defense counsel what I said about my credibility in this court.

… about two hours later

GabADR: I’m back. And I think this is the juncture to propose a bracket.

Plaintiff’s counsel: Do you have a virus?! It’s 10:45 in the morning.

GabADR: The margin is sufficiently narrow to move to brackets.

Plaintiff’s counsel: Well, in my experience brackets in mid-morning lead to an impasse by mid-afternoon. What do your calcs show about the timing of commencing exchanges of brackets?

GabADR: I have no such data. My creators find it to be unobtainable, given that virtually all mediations are strictly confidential, thus precluding data mining. Nevertheless, predictive analytics strongly dictates moving to brackets.

Plaintiff’s counsel: (Shrugs). Sure, whatever. You AIs are going to run the World I’m told. So put in a kind word for me to Skynet and convey to defendant a bracket of …

… four hours later (2:56 pm)

GabADR: So I’ve brought everyone back into general session as we are clearly at an unbridgeable impasse.

Plaintiff’s counsel: I said it! Brackets before lunch; mediation implodes before the Stock Exchange closes!

Defense counsel (sotto voce): I feel I’d like to travel back in time so I can stop myself from signing the mediation agreement ...

Plaintiff’s counsel: You know, Gab, to be blunt it seems like it’s a real problem that much – probably most – of the data you’d need to “optimally” mediate is unobtainable given the non-public nature in which mediations proceed: you know, what kinds of offers establish credibility, when incrementalism works and doesn’t work, timing as to when brackets and other “advanced” techniques are best deployed, the negotiating approaches of counsel, I mean, practically everything.

Defense counsel: Yeah, with your data deficit, how could you ever break a stalemate?

… 3 minutes later

GabADR: Are you saying you would like to play me in chess?

(Transcript ends)