Today I want to take you into the realm of Artificial Intelligence and look at its potential to create an impact in the fight against terrorism.
The aim of AI (artificial intelligence) is to get at what is happening when one’s mind silently and invisibly chooses, from a myriad of alternatives, which one makes most sense in a very complex situation. In many real-life situations, deductive reasoning is inappropriate, not because it would give wrong answers, but because there are too many correct but irrelevant statements which can be made; there are just too many things to take into account simultaneously for reasoning alone to be sufficient.
— Douglas Hofstadter, Godel, Escher, Bach: An Eternal Golden Braid
Today, artificial intelligence (AI) is a concept that is too big to ignore and too exciting to deny. We should not be frightened by it—either its “hype” or its implied shortfalls—but rather should use it as an aid in the development of practical tools. The concept has a wider application than computer systems alone and should be viewed as a vehicle that helps us to think about the way in which we reason, how we draw inferences from given facts, and what rules we use to tie those facts together.
In a recent trade journal it was predicted by market researchers Frost and Sullivan that the size of the artificial intelligence market for the U.S. military will more than double in the next five years. I say this not to suggest that by throwing money into an AI program we can produce a do-anything expert system but to encourage the use of such programs where they best serve our needs. In the area of law enforcement, Richard Larson and his colleagues at the Massachusetts Institute of Technology have been advocating the use of expert systems for a number of years. In this article I propose to explore some of the problems we face in understanding and dealing with terrorism and then to consider the use of artificial intelligence techniques as practitioner tools, commenting on their suitability.
Expert systems require a means of simulating reasoning—of inferencing—most commonly by utilizing a set of “if/then” rules. They also require a knowledge base, which may in turn necessitate the elucidation of human experts regarding how to approach a particular problem. You may recall the hackneyed example of the system envisaged to aid in medical diagnosis. A patient is seated in front of a small computer terminal. On its screen a series of questions appear. The patient responds to these questions, which are diagnostic in nature and based on the best knowledge of experts—the doctors. The system eliminates the possible medical conditions based on the known symptoms for those conditions. The result is a diagnosis with an expert opinion of the likely ailment. This is an example of a simple expert system.
With these sorts of ideas at the back of our minds, let us ask ourselves if expert systems could have some application to the various ways that we deal with the problem of terrorism.
We know that acts of terrorism can take place in diverse locations of the world and be perpetrated by numerous actors. We also know that the nature of the act, like the problem of crime itself, may change over time. Many commentators have argued that the places where crimes are committed and the times at which they are committed are unpredictable; therefore, any positive policing activity directed toward the prevention of crime is virtually impossible. I do not subscribe to this view and would argue that crimes in general have a relatively high degree of predictability. As all advocates of situational crime prevention techniques know from experience, distinctive patterns and trends can be utilized to take both preemptive and detective action.
In the case of those crimes we might call terroristic, I would argue that a similar situation exists. Analysis indicates that factors governing the commission of acts of terrorism correlate largely to sociopolitical factors pertaining in a particular country or to the interaction of events that may occur in more than one country. The acts themselves can be characterized as particular types, whether they be kidnapping, extortion, hijacking, or bombing. What I am saying is that there are definable connections among places, people, and events. Where such connections exist, statements—then rules—can be made about them.
It is the democracies of the world that are most vulnerable to acts of terrorism. One might argue that a democracy poses the greatest danger to itself by its response to terrorism. If, for example, it reacts with what are seen as harsh or oppressive security measures, it may alienate the public it is trying to protect. I have recently heard of two such instances that may be worthy of mention. When France suffered a recent series of bombings in Paris, the government put large groups of uniformed officers, heavily armed, on patrol in the shopping areas of the city. Considerable comment was made by Parisians and visitors alike about this reaction. Similarly, the sight of uniformed police officers carrying machine guns at London’s Heathrow Airport is also the subject of considerable comment. Both of these responses are what we might call “out of character.” We know why these decisions were made, but if there were another way to deter terrorists or protect the public I am sure we would choose it.
Conventional intelligence gathering in the area of law enforcement has been questioned by some observers, who suggest that it does nothing more than meet statutory requirements for data (i.e., it is a statistical or administrative function of the state). We might also question whether we have traditionally gathered the kind of information we need to prevent or detect the crimes known as acts of terrorism or whether we collect data in a less methodological way. I would ask anyone working in the law enforcement area to consider their own method of data collection in the post-incident phase of a crime. Data files often contain pages of free text that explain the activities and thoughts of the investigator in this post-incident phase; however, for the most part the data will refer to unsolved crimes. What is the currency of such data? How should we proceed to undertake its analysis? How long does it remain in some useful and retrievable form?
Questions such as these may seem fundamental, but I would suspect that when investigations into acts of terrorism are carried out, the methods of data capturing, recording, and storage are, in some way, similar. In order to achieve the best practice to produce the best intelligence, we must ensure that all our detectives are equipped to make the sort of inferences we would wish. Do they have the requisite skills, and what training do we give to help them make these inferences?
Moving on from crime intelligence, we might consider the area of threat assessment. We could look at this area as a mixture of profiles and facts. Assessments can be based on facts such as direct threats and on profiles inferred from a knowledge of world events. Different levels of analysis are obviously involved. It might be an interesting academic exercise to review the assessments made by both governmental and commercial intelligence-gathering agencies and measure their success against actual world events. Certainly, applying the expert system to this problem would combine a multidisciplinary approach to the factors affecting terrorism with the pure expertise of the “generic” expert.
An opportunity exists for the creation of systems to “advise” practitioners both in the post-incident stage of a terrorist event and in the preemptive stages. In the latter case, screening devices could be used to advise security agencies in a covert manner. For example, vast amounts of data could be utilized in the screening of passengers boarding aircraft to prevent the entry of hijacking teams and in the post-incident stage to provide advisers to negotiators.
The problems of air carriers are classic. Leaving aside the separate danger of proxy bombing, the major difficulty identified by carriers is that of ensuring the security of passengers who board aircraft passing through high-risk countries. Air carriers may be faced with screening up to 400 to 500 passengers in the short time before they board a jumbo jet. The objective is to ensure that no bombs or firearms are taken onto the aircraft, thus lessening the possibility of an armed hijacking taking place. We know that this process is expensive in human and financial terms and, from time to time, ineffective. Another way of undertaking such screening would be to identify from the 400 or so passengers those persons who may be traveling as an undisclosed group-perhaps a hijacking team. Using this method might make it possible to reduce the number of physical checks by eliminating the indiscriminate checking of all the passengers in favor of a more thorough investigation of those highlighted by the screening procedure.
Another form of expert system could be constructed to assist in the post-event phases of terrorist acts such as the critical area of negotiating. Vast experience has been accumulated in siege and hostage situations by relatively few persons. The lessons learned from these experiences form the basis of an expert system which could provide a practical tool to law enforcement personnel.
Understanding the phenomenon of terrorism as distinct from crime in general has taken considerable time and effort. Academics especially have worked on creating data bases that would help them answer questions about how groups behave and the frequency of bombings or particular incidents, but little has been done to create systems that can scan vast amounts of data before providing an answer.
There are a number of opportunities to utilize AI systems in small but important areas of the terrorism “industry”. I have already indicated several options, such as a system to advise and prompt negotiators. Numerous other possibilities exist, from providing training simulators to examining the awareness of corporate executives, who are often the targets of acts of terrorism.
I hope that in the course of this article I have been able to give you an insight into artificial intelligence techniques and expert systems and a taste of their potential within the area of law enforcement. In fact, if we have learned one lesson from our pioneering work in the United Kingdom, which goes back several years, it is that we are still learning and still discovering many potential applications. Expertise in law enforcement investigation is not as definable as the expertise of, say, the hard scientist, the geologist, the physicist, or the chemist. A lot of our expertise lies with the practitioners who, by virtue of doing their jobs over long years, have accumulated vast experience. Whether you talk about expertise or best practice, there is a considerable amount of wisdom, knowledge, and proven method in our approaches. Instead of these being lost to law enforcement when experienced officers leave, information can be gathered into systems which permit such knowledge to live and, above all, be available to the rookie cop.
Our job is about people, emergencies, tension, and adrenaline, and in the field we need fast decisions, not the more leisurely answers to problems given by pundits and professional commentators. It is perhaps an unfortunate fact that law enforcement agencies are expected to respond quickly. Expert systems can give us the cumulative wisdom and knowledge of the best of our people. It would be very shortsighted if we, as practitioners, did not recognize the benefits that these types of systems can offer, for, at the end of the day, it is our reputation and professionalism that is at stake.
So, does all this sound a little dated? It should do. I authored the article nearly 20 years ago. Did it have any profound effect? Not really. Both the British and United States governments chose not to support the development of aircraft passenger screening tools – as did the airlines themselves. And, are we not still trying to get our intelligence act together? Let us hope that if we reprint the article in 20 years things will have moved on a little.
Comments
Post a Comment