Why We Should Replace Private Arbitrators With A Specialized Version Of Chat GPT
An LLM AI will make more accurate, more unbiased, faster & cheaper Arbitration Awards than fallible, biased and sometimes stupid human judges
By David Grace (Amazon Page — David Grace Website)
A friend of mine who specializes in probate/estate litigation recently participated in an arbitration where the privately-paid arbitrator was a retired judge. There were four parties in the case and four lawyers showed up for the one-day arbitration hearing.
Human Arbitrators Are VERY Expensive
The attorney’s fees for that one day of arbitration were at least $15,000. The arbitration service that provided the private judge/arbitrator charged a fee of $11,000 to adjudicate that ONE DAY of arbitration.
On top of those charges, the attorneys submitted post-arbitration briefs and no doubt there will be other post-arbitration documents prepared, exchanged, and argued over, including competing forms of the proposed final judgment.
The final arbitrator’s award will have to be filed with the Superior Court and further proceedings will be held to turn that award into an enforceable court order.
Money, money, money. Time, time, time. What an expensive, cumbersome, slow, and time-consuming process.
A Clearly Wrong Arbitration Decision Cannot Be Fixed On Appeal
Plus, PLUS, short of proof of bribery or other serious misconduct, there is no appeal from an arbitrator’s award, so if your handsomely-paid retired judge was unprovably biased, lazy, stupid, crazy or confused and, for any reason short of serious misconduct issued a judgment totally against the weight of the evidence or contrary to the law, you’re still stuck with it.
Let me repeat, there is no appeal from an arbitrator’s decision even if the ruling is obviously and completely contrary to the law. And that is always a risk.
Paid Arbitrators Have A Bias In Favor Of Lawyers Who Can Send Them More Business
And consider this: the arbitration company the judge works for wants continuing business, repeat customers.
If one side of the dispute is represented by a large firm that participates in lots of arbitrations and the other side is represented by a lawyer who rarely goes to arbitration, the arbitration company and their employee retired judge will have a material desire to keep the lawyer from the big firm happy so that his/her office will use that arbitration service and that particular judge again in the future.
I will leave to your imagination the question of how much the arbitrator’s desire for repeat business from one of the litigant firms may influence the his/her decision.
For these reasons, I would almost never voluntarily take a dispute to private arbitration.
Training An AI LLM
Now let’s assume that OpenAI created a version of Chat GPT 4 or maybe GPT 5 that was trained exclusively on the following materials:
- All California appellate decisions.
- All California statutes.
- All federal decisions relating to federal litigation within the State of California.
- Basic legal references.
Data An AI LLM Would Use To Make An Award
In a Chat GPT-decided case, the parties would complete their discovery which would consist of depositions, answers to interrogatories, subpoenaed documents, business records, and the like.
Each party would prepare what it contends is an accurate statement of facts and its written argument in support of its view of the facts.
Each party would prepare a response to the other side’s statement of facts and legal arguments.
The parties would then submit all these written and visual materials together with a joint set of questions to be decided to this specialized iteration of Chat GPT.
Chat GPT would then decide each of these submitted issues. A few minutes after submission, Chat GPT would issue its written opinion.
An AI Arbitration Would Give Both Parties A Chance To Fix Something They Originally Missed
Each of the parties would have some period of time to conduct one more round of discovery and submit further briefs, documents and depositions.
After including the new discovery, the issues would again be submitted to Chat GPT and this time its decision would be final. That written opinion would then be submitted to the Superior Court for the issuance of a civil judgment in the same way that a civil judgment is issued today based on the written decision of a human arbitrator.
There Would Not Need To Be A Material Change In The Law
The only change that would be required in the law would be the necessity for a statute providing that a decision of an AI arbitrator that was issued pursuant to a written agreement by all the litigants will be codified in a judgment of the Superior Court in the same manner and under the same rules that the decision of a human arbitrator is today confirmed by a judgment issued by the Superior Court.
The Benefits Of A GPT Arbitrator
If we used Chat GPT as the arbitrator, the rulings would not be biased in favor of attorneys who might bring more business to an arbitrator whose decision went their way.
The case would be decided by a judge who, unlike a human arbitrator, would not be tired, crazy, stupid, misunderstand the law, or confused about the applicable legal principles and could not be bribed or improperly influenced.
GPT’s decision would be more reliable, more free from bias, more in conformance with the law, and more immune to human foibles and prejudices than a decision by a human arbitrator.
It would not require a physical hearing. All testimony would be through the submission of depositions, videos, and documents collected weeks or months prior to the submission of the case.
It also would not cost $11,000/day. It might well cost only a few hundred or a few thousand dollars, AND GPT’s award would be issued in minutes rather than weeks or months.
The Role Of Attorneys Would Change
Of course, the role of the attorneys would materially change. Skilled GPT practitioners would be expert in drafting the statements of facts, law, and the questions for decision. Familiarity with structuring questions for and submissions to a Large Language Model AI would be a valuable skill.
Eventually, the LLM’s training might be augmented with expertise in body language which GPT could then apply in evaluating the witness’s demeanor during video-taped deposition testimony.
Unrealistic Fears
I know some people will be fearful of having their case decided by a “soulless machine.” Other more simple-minded people might fear that the AI will go rogue like HAL in 2001 or that it might render an erroneous decision because it was hacked by one of the parties.
All of those fears are groundless, and they ignore the very real competing risks and flaws in submitting disputes to privately-paid human judges who pose a far greater risk of bias or error.
An AI Judge Is Safer Than A Human Judge
The relevant question is not “Is it possible that the AI might make a bad decision?” but rather “How much more likely is it that a human, privately-paid judge might make a bad decision than that an AI judge might make a bad decision?”
In my opinion, the human judge is many, many times more likely to get it wrong than a specialized AI whose knowledge set includes every statute, every appellate opinion and every principle of law that could possibly be applicable to the case, an AI that cannot forget or overlook any evidentiary fact or legal principle and cannot succumb to bribery, stupidity or bias.
Nothing is perfect, but an LLM AI, especially one with expertise in reading a witness’ body language, is many, many times more likely to make the right decision in a legal dispute than some run-of-the-mill $11,000/day retired judge.
What I Think OpenAI Should Do
I think that right now OpenAI should begin providing a Chat GPT variant specialized in the law of each state to act as an AI Arbitrator so that we can get some real-world idea of how this new procedure could make the legal system work better, faster, and cheaper.
— David Grace (Amazon Page — David Grace Website)
