NCFM Member Brent Oster, Do you believe the American system of justice is corrupt?


Any person who has been through the U.S. legal system, whether it was in family, criminal, civil or other courts, has probably realized how inaccessible and frustrating the process can be.

The problems that the legal system currently faces are fundamentally information science problems, along with the limitations of human beings, as there is a loss in communication and translation at each step with no error correction.

This is referred to as Developmental Stage Theory – The information used in processes becomes increasingly distorted as it becomes further removed from original theoretical sources.…/vol_4_no_1_ross_using…

This is like the game of telephone, where a message is passed from person to person, and after a few people, the output drifts far from the input, with each person making errors that are not corrected. Even still, each person on the chain thinks they heard and stated the information correctly.

Lawyers do their best to translate a client’s needs and points into legal documents. Then, in a hearing process, one side files a motion, then the other side files an objection, and the first side gets a final response — this triplet of communication is then (often partially) read by a judge. At a hearing, each side gets to speak briefly, the judge makes a decision and an order is signed (or not). In a legal proceeding, there can be dozens of hearings and one or more trials to decide major issues.

A problem is that these hearings are compartmentalized and the decisions in one hearing may be made on partial or even false information, but once an order is issued, it seals off that information, including any doubt, falsifications and error, and the next hearings are based on only the result. This propagation of a chain of decisions without overall correction can result in meaningless or disastrous outcomes in the proceeding overall.

Here is a divorce proceeding as an example, with only 1/4 of the actual hearings listed, in which a couple were completely defrauded of all their assets, including their retirement and a $1.3M house by the attorneys, a DA that was a family member of the lawyers, with a realtor, and a temp judge conspiring with them. They were able to do this because they could take advantage of the lack of information passed from hearing to hearing, with different Judges making orders on partial or false information along the way, because of the lawyers introducing false information each time.

Also, most decisions by Judges are meaningless. Each time a Judge rules on a case, they cite precedents, previous judgements on similar cases, which cited judgements on previous ‘similar’ cases to justify the decision,… all the way down through the decades and centuries in some cases, especially Constitutional rights cases which go way back.

By the time you trace from the recent ruling all the way back down to the laws on which the original ruling was made, there is so much distortion and drift from the original meaning of the law to the present misinterpretation of it – that it is not even remotely valid, for all the same reasons, Developmental Stage Theory – the information becomes increasingly distorted as it becomes further removed from original theoretical sources.

All of these problems stem from the information processing limitations of human beings – our error susceptibility when paraphrasing and translating information, in making judgment decisions on partial information, and a human’s inability to observe and assimilate a large number of steps in a court proceeding or citations as a continuous process, nor recognize distortion and drift in the process.

Now, what if we took this whole system and converted it to artificial intelligence and eliminated all the steps that only humans need: clients, lawyers, filings, hearings, judges and orders, all of which are just constructs to accommodate human cognitive and information processing limitations. Do we even need an adversarial system with two opposing lawyers and a judge?

Let’s go to the concrete example of a divorce proceeding, perhaps one of the most highly contested, potentially emotional and expensive proceeding (relative to the returns). The goal at the end of a divorce proceeding is to produce a property and asset division and a support schedule on a spreadsheet called a propertizer — and to decide custody and visiting of children and other administrative matters.

Let’s say we develop a divorce legal AI that could interview each individual, get all their documents from them, pull the records for their real estate and other large assets and then create an unbiased and legally correct division of assets — and a support schedule based on actual accounting principles, given their relative incomes and needs. The clients could also input their work schedules and note when they would like to see their children and have this information passed back and forth via the AI mediator, decided in an intelligent, rational, and mutually agreed matter,

Say we could do the same with the civil courts, starting with mediation of small-claims lawsuits, providing a service for filing these claims, for the other party to answer and for an AI solution to evaluate the claim — and an AI mediation process to seek a resolution that involves both parties so they buy into the final resolution. These two services could start small, allowing people with small claims or simple divorces to use a much simpler AI mediation system, and then it could be upgraded and expanded to handle more complex cases. Just these two AI legal systems alone would result in a much simpler and easier process for many people to get these legal services.

Because humans are fallible, with limited attention spans and scope, we are not well suited to judge matters that may span years and involve dozens of parties and attorneys and hearings. Automating key areas first, like summarizing the content and filings of the hearings with an AI process to evaluate their content so their effect on the proceeding can be balanced and fair is a good start. Flagging out-of-bounds cases for review automatically is another great application of AI that can augment Judges.

Someday, with completely automated AI mediation and claims resolution, people may not have to go into courts, or employ lawyers, and they can work with a user-friendly and intuitive mediation system to solve problems, resolve claims, and get fair outcomes that satisfy both parties. Less and less people will come to the courts, judges and courtrooms will be scaled back, and one day, the last human judge will close the last court room on the last session, and AI will preside supreme over human legal affairs.

Brent Oster

CEO ORBAI – Artificial Intelligence and Robotics
Former Deep Learning Solution Architect at NVIDIA2008–2018
MSc in Computer Science, University of California, Santa Barbara Graduated 2008

NCFM Member Brent Oster, Do you believe the American system of justice is corrupt?