In some ways, it is easy to understand how the game of checkers eventually lost its excitement for Marion Tinsley. Easily the greatest checkers player ever, Tinsley reportedly spent much of his time as a graduate student mastering the game. And he truly mastered it in a way that few have mastered other games — the record of such greats as Emmanuel Lasker or Bobby Fischer in chess or of Wu Qingyuan in go do not even come close to Tinsley’s record in checkers: In a career that spanned some 45 years, he lost only seven games, the last two of which are mentioned below. It was a given that Tinsley would win any match he participated in. Later in life, he remarked that he had become bored playing checkers with human beings because there was no longer any challenge. In matches, he would play to win just enough games to assure his win in the match and then lazily play the remainder to a draw.
But in 1990, the computer program Chinook was developed and became a competitor that might actually present a challenge for Tinsley. The history of the 1992 tournament is one that embarrassed the bodies that govern international checkers: Chinook had won the right to play Tinsley for the World Checkers Championship, but those bodies retroactively determined that computers were ineligible for the title of World Champion. Tinsley resigned his title so he could play Chinook, leaving those bodies in a position where their champion could never truly be acknowledged as the world’s best checkers player of the time as long as Tinsley was still around.
Tinsley won that match against Chinook, drawing 33 games, winning four, and losing two — the last two of the seven he would lose over his career. He said that playing Chinook made him feel like a young man again.
This week, many are watching the television game show Jeopardy! as two of its greatest human champions compete against the computer program Watson. In some ways, Jeopardy! seems simpler even than checkers because a very big database can hold a lot of factual information. But the real challenge for Watson is not in the retrieval of information, but in its ability to parse human language in a way that it understands what is being asked. It is something humans do with ease but which computers continue to have enormous difficulty with — as anyone who has tried a natural-language search engine well understands. To put things in perspective, consider that as great as the world’s CPU power currently is, a report in Science last week notes that it is roughly equivalent to the processing power of a single human brain.
Developments in artificial intelligence are now affecting every area of human intellectual effort, and the law is no exception. Lawyers deal with enormous amounts of information that can be difficult to organize and understand so that it can be applied most effectively. Much of litigation is frequently dominated by armies of attorneys on each side of a dispute poring over many thousands of documents, seeking to identify those facts that are most relevant and that will prove most persuasive to a jury. They seek ever more effective ways of determining how to present those facts in a way that will cause the jury to return the result they want. Every beginning attorney is exposed to the stories of their highly paid friends spending Thanksgiving or Christmas locked in a warehouse of documents, endlessly reviewing them just to identify those facts. When billions of dollars are at stake, the effort is worth it.
Increasingly, attorneys are making use of software designed to streamline the task by tagging information in a way that makes it easier to identify the strengths and weaknesses of a case, as well as the role that individual witnesses can play in developing the presentation of a case. At the moment, the judgment of how to present the case is still made by a human attorney. But if we are getting to the stage where Watson can challenge the best Jeopardy! players by actually understanding the information and drawing conclusions, we are beginning to approach a time when human attorneys may be matching their wits against opponents who use artificial intelligence — to select facts, to identify connections between facts, to evaluate the persuasiveness of facts, and to determine the best ways in which to present those facts to jurors who have modes of evaluation that can increasingly be modeled.
Indeed, the use of mock juries in evaluating cases has become commonplace in preparing for large trials. Such mock juries are used to aid attorneys in understanding what evidence juries accept at face value or view with skepticism, to determine how the way individual witnesses portray themselves on the stand affects jurors’ impressions of their credibility, to evaluate the way jurors respond to certain types of language and choice of words. And perhaps above all, to gain insight into the types of emotions and hidden motives jurors apply in reaching their verdicts.
All of these aspects of trials — and more — are the subjects of active research by artificial-intelligence scientists. Programs exist and are being further developed to aid in the analysis of case law, having computers read the voluminous relevant cases to distill rules of law, identify exceptions, and to find lines of reasoning that can be exploited. Models are being progressively refined to mimic adversarial attorney interactions by predicting strategic responses and to rank different strategic approaches so that attorneys may have better insight in constructing and adapting their overall tactics. The role not only of juries in moving a trial from opening to verdict but also of judges and the impact of their evidentiary rulings are also being studied with the techniques of artificial intelligence.
At the moment, such research provides only rudimentary tools that attorneys necessarily use, but the research will continue and the techniques of artificial intelligence will have progressively more impact. In a world where companies willingly pay the most talented attorneys $1000 an hour — because the tactical legal skill they bring to bear is important enough to make a difference in the outcome of a case — there is no question that they will also willingly pay to have the best advantage the techniques of artificial intelligence can provide.
About Patrick Boucher