The term artificial intelligence – AI for short – has become ubiquitous. It’s easy to be dazzled by the promise of smart technology as we see it in popular culture and science fiction. Westworld makes it all seem so real.
In reality, an enormous gulf remains between what we want this technology to be able to do and what it can do in practice. In the courtroom, there’s a real danger that jurors – as well as judges and opposing counsel – can be easily persuaded by expert witnesses and litigators to accept without question any conclusions produced by this technology.
Sometimes even the experts cannot agree on the definitions of the terminology, so it’s easy to see how jurors can be led astray. The confusion starts with the terminology. “AI” as it is commonly used actually comingles four categories of computer functions:
- Algorithms – These generally describe sets of rules or commands within a program for solving a problem, such as Google’s highly secretive formula for determining search engine page rank.
- Robotic Process Automation (RPA) – IBM defines RPA as using “bots” that automate highly repetitive tasks. Think of a national bank that might use a computer to process and pay utility bills that are received from account holders electronically in its online bill pay system.
- Machine Learning – With machine learning, technology can go a step further than RPA by being trained or taught to evaluate data so that its ability to analyze improves over time.
- Rapid Data Search – Computers that have been loaded with an enormous amount of data are programmed to search that data very rapidly when certain keyword or phrases are entered.
None of these functions qualify as artificial intelligence compared with AI in popular culture and in the mindset of a typical juror. AI in popular culture refers to computer systems that can think, reason and make independent decisions like a human. Think of a computer that can perform the above four categories, all at the same time, while also applying reasoning and judgment. So far, and for the foreseeable future, commercially available computers cannot do all four of those things at once, much less apply reasoning and judgment.
Sadly, the adage that computers are only as good as the programmers who program them is still true. One only has to look at a “chatbot” on the websites of virtually every company we interact with today. It might seem like we’re interacting with a human, but other than assisting with the most basic of topics, it is almost impossible to get a useful answer to the questions we type into the chatbot. Why? Because the AI that is claimed to exist at most companies is nothing more than a computer performing rapid data search.
Attorneys are probably most familiar with these tools as they relate to searching emails and other digital documents as part of e-discovery. These are powerful advancements that can make everyone’s work more efficient, but they are only as good as what has been programmed – typically through keywords and phrases along with “seed sets” of documents entered as examples of the types of data you are looking for in a larger population.
What does this have to do with expert witnesses, forensic accounting and the courtroom? As a practicing forensic CPA, I am beginning to see witnesses – both expert and fact – referring to AI in their testimony as if the use of AI or related technology makes their work correct, infallible or superior.
The term AI in a court setting should be a blinking red light for litigators. These models are not intelligence – and they’re far from infallible. Moreover, they have no ability to think, reason or apply judgment to a situation. When used for damage quantification, they are more akin to a series of Excel spreadsheets than they are to any form of intelligence. This issue is important whether you are presenting a witness on direct or cross-examining an expert who attempts to use these buzzwords to explain or enhance the credibility of his or her work.
While you can’t cross-examine an algorithm or a computer model, it’s important to remember that there are human decisions and biases behind all of these technologies, and testimony based on these tools is only as good as the data entered on the front end. By helping judges and juries understand the zeros and ones inside the technology, trial lawyers working with forensic experts can help shed the mystique from this type of testimony and create a clearer picture for judges and juries.
Larry Kanter is the founder of Dallas-based Kanter Financial Forensics, LLC and is an adjunct professor at SMU’s Cox Graduate School of Business, where he teaches forensic accounting and forensic data analysis. He can be reached larry@kanterforensics.com.