Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
MEMORANDUM Research and Legal Issues R. Klrkland Gable, Ed.D., J.D. Associate Professor of Psychology California Lutheran College Thousand Oaks, California (1) It was indicated in a summary of the November 13, 1980, planning committee meeting that more than half the subjects the Secret Service deals with have mental health problems. This seems to be an unusually high percentage. Is is possible that the Service over-selects some persons because they have a mental history? This could result in an unnecessarily high percentage of false positives. Professor Zimring's proposed research design using "proxy" behaviors could help to answer this question if subpopulations are sampled from persons with and without mental histories across categories of persons receiving various Service interventions. (Incidentally, additional "proxy" behaviors might include aggressive driving behaviors or offenses, violent sports such as hunting, child abuse, sleep disturbance, and behaviors related to agitated depression.) (2) Some aggressive behaviors may be state-dependent. They may occur primarily with the use of alcohol or prohibited drugs or in the absence of psychotropic medication.* If this is so, then perhaps interviews and other assessment procedures might be conducted when the subjects are in these particular states of conciousness. Similarly, therapeutic intervention might be most effectively conducted in these altered states. Finally, certain Secret Service surveillance, restraint, or other intervention procedures might be legally conditioned upon the subject's likely states of consciousness and medication compliance. (3) Assessment might involve some consideration of the extent to which the dangerous conduct is spontaneous (impulsive) or planned (instrumental) or a combination of these modes. If the subject's aggressive target is narrow and specific, therapeutic intervention might involve stimulus narrowing and desensitization procedures. If stress interviews appear to increase the probability of subsequent *See V. I. Reus, H. Weingartner, and R. M. Post, "Clinical Implications of State-Dependent Learning," American Journal of Psychiatry 136 (1979): 927-31; and C. L. Cunningham, "Alcohol as a Cure for Extinction: State Dependency Produced by Conditioned Inhibition," Animal Learning and Behavior 7 (1979): 45-52. 171
dangerous behaviors, they might be followed by relaxation interviews conducted by clinicians to mitigate the effects of stress interviews. Some useful information might also be obtained from these interviews by assessing the extent to which a subject can remain relaxed in the presence of target stimuli that usually elicit aggressive responses. (4) Although it is tempting to compromise the legal rights of subjects in balancing potential costs and benefits, caution should be exercised in this regard. Management and intervention strategies might be designed with the doctrine of "least restrictive alternatives" in mind. However, certain assumed "alternatives" may in fact be mere illusions of choice if they do not have empirically measurable consequences. Thus, the formulation of legal and policy recommendations should be sensitive to empirical findings in the difficult task of balancing competing, legitimate public and private interests. Some Thoughts Stimulated by the Conference (1) There may be some value in more extensively using metal detection devices for screening persons in proximity to the President. This was mentioned at the conference. (2) As a time-limited research project, persons in public crowds near the president might be randomly sampled and screened for possession of firearms, mental history, Secret Service investigative status, and so on. This would give researchers some impression as to the risk of public appearances of the president. It could also be used to assess the effectiveness of Secret Service interventions as shown by a reduced number of "problematic" persons in public crowds. (3) Systematic or discretionary searches might be made of persons in proximity to the president without the usual legal standard of probable cause. Persons within a specified proximity to the president would be deemed to have given express or implied consent to a search. By analogy, implied consent is given for blood alcohol testing by the act of driving a motor vehicle in some jurisdictions. (Here we are dealing with equal or substantially greater risks presented by the illegal conduct of persons.) In addition, express consent could be obtained, if necessary, by making public announcement, both verbal and posted, prior to the president's arrival. (4) Electronic monitoring equipment is readily available and has been used to record the location of mental patients and parolees as a method to protect the public. While the use of such equipment is not necessarily recommended, its capability is worth noting. 172
Bibliography Schwitzgebel, R. L., and Schwitzgebel, R. K. Law and Psychological Practice. New York: John Wiley and Sons, 1980. 173
MEMORANDUM On Assessment Kenne th R. Hammond, Ph.D. Professor of Psychology Center for Research on Judgment and Policy University of Colorado Boulder, Colorado There has been considerable research on human judgment and decision-making over the past 25 years. (Persons familiar with the field estimate that the number of empirical papers published in refereed journals exceeds 3,000 and may be as high as 5,000.) There is, in short, an empirical research basis for the following conclusions: 1. Judgments in general are highly fallible. 2. Fallibility increases as the "softness" of the judgment task increasesâthat is, as the judgment becomes increasingly intuition-dependent. 3. Predictions of behavior based on judgment are particularly fallible. 4. Expert judgment regarding the prediction of behavior is almost certainly no better than that of the nonexpert. (Clinical judgment has been studied since the.early 1950s; there is no evidence whatever that clinical judgment of any kind is as good as actuarial prediction, and considerable evidence that it is worse.) These conclusions are as firmly based as any conclusions in the field of behavioral science. The empirical findings noted above, together with (a) theoretical, analytical treatment of the prediction problem, and (b) general descriptions of the situation faced by the Secret Service agent in the field, and his consultants, lead to the conclusions: 1. These four findings indicated above are applicable to the judgments that must be made by the agent in the field, and 2. to any mental health professional to whom he might turn for consultation. Furthermore, 3. because of the agents' duty to intervene (that is, to interview and occasionally restrain persons of interest), it will be very difficult, and perhaps impossible, to carry out research aimed at improving the predictive accuracy of the judgments that must be made by the Secret Service agents and their consultants from the mental health profession. 175
Recommendations: 1. It will not be useful to recapitulate the experience of the past 25 years by carrying out new studies with Secret Service personnel. Such studies would be expensive, time-consuming for both researchers and Secret Service personnel, and have a very low probability of teaching us anything new. In the absence of highly compelling reasons to believe that Secret Service agents or their consultants possess some special abilities heretofore wholly unobserved in other professionals, research that merely recapitulates previous work cannot be justified; therefore it should not be undertaken. Retrospective analyses will almost certainly be worthless in view of the fact of intervention. 2. The research that has been done on judgments, decision-making, and prediction of behavior suggests, but does not prove, that there is some value to be derived from decision support systems. In the case of the Secret Service, an effort to develop such a system might prove to be worthwhile, particularly if it were closely related to a management information system that was developed with research needs in mind. Such a system may decrease the work load of the agency by leading to improvements in ⢠the information system upon which the agency depends ⢠the retraceability of decisions ⢠the uniformity of the decision-making process ⢠the training of agents with regard to predictions of behavior. 176
MEMORANDUM The Escalating Incidence of Assassination Brian M. Jenkins Director, Security and Subnational Conflict Program Rand Corporation Santa Monica, California 1. Assassination: A Growing Problem Assassinations worldwide have increased. The total volume of terrorist activity increased during the 1968 to 1980 period, and assassinations also increased as a percentage of all terrorist incidents. More than 500 assassinations occur annually worldwide. Not only do we see an increase in assassinations by terrorist groups, but government-backed assassination campaigns directed against foreign or domestic foes abroad also have increased. For example, Libyan assassins are believed responsible for a number of murders last year of Libyans living abroad who failed to heed Colonel Qadaffi's warning to return. The United States has not been immune to this crime. Last year, anti-Castro Cuban emigres assassinated a Cuban diplomat in New York. In 1976, anti-Castro extremists killed a former Chilean cabinet minister in Washington at the behest of the Chilean secret police. In 1980, a gunman or gunmen, believed to be in the employ of Iran, shot to death a former Iranian official in Washington. 2. Rand's Research on Political Violence Since 1972, the Rand Corporation has been engaged in research on various aspects of political violence, terrorism and low-level conflict. This research effort has examined the origins and theory of modern terrorism, the mindset and modus operandi of various terrorist groups, the specific problem of political kidnappings, and trends in terrorism. None of these studies specifically addresses evaluating threats to the president and other persons protected by the U. S. Secret Service. However, one current research project does deal with the problem of assessing the credibility of threat messages involving nuclear-related extortion. 3. Rand's Threat Credibility Assessment Project Since 1970, there have been approximately 50 threat messages in which the author or authors threatened to detonate a nuclear 177
explosive device or disperse radioactive material in an American city. These events, although generally not publicized, create considerable disruption and could, if publicized, cause alarm and panic. Following such a threat in Los Angeles in 1975, it was deemed necessary to create a capability for rapidly assessing the credibility of nuclear threat messages. Two developments are making the traditional method of assessing the credibility of nuclear threats more difficult. Although there is confidence in U. S. safeguards and accounting systems, there is an increasing amount of nuclear material beyond American control. The spread of nuclear programs throughout the world makes it increasingly difficult to say with a high degree of confidence that the authors of such a threat do not possess nuclear material. The spread of nuclear knowledge increases the number of people who possess at least a theoretical knowledge of nuclear weapons design. Even a layman, with little technical background, can use the right words. It becomes more and more likely that the designs or equations submitted to support a nuclear threat will be correct. These developments place a growing burden on a behavioral assessment in support of the technical assessment. The development of a behavioral assessment capability is Rand's current task. When a nuclear message is received by the FBI, which has jurisdiction over violations of the Atomic Energy Act, it is immediately forwarded to a special FBI desk in Washington. The Emergency Action Coordination Team in the Department of Energy is notified. This team then activates the Threat Assessment Team (TAT). The TAT consists of separate teams linked by computer but working separately, at least until after the initial assessment is made. The teams at Los Alamos and Lawrence Livermore laboratories address the technical and operational aspects of the message. Other teams address the behavioral aspects. The combined assessments are communicated to the Department of Energy. The research and actual threat assessment experience thus far have demonstrated the feasibility and utility of a behavioral assessment capability. We are in the process of systematizing conclusions from research and lessons learned in evaluating threats into a list of credibility criteria. The behavioral assessment capability is being expanded to other areas. While the Department of Energy's system for assessing nuclear threat messages is elaborate and involves teams including physicists, psychiatrists, psychologists, propaganda analysts, and persons with other specified skills, portions of this work may provide a model for developing a similar, although necessarily less elaborate, capability in support of Secret Service activities. 178
MEMORANDUM Prediction Research John Monahan, Ph.D. Professor of Law and Professor of Psychology University of Virginia Law School Charlottesville, Virginia I would offer four observations for consideration by workshop participants prior to our March 8 meeting. (1) I strongly support the suggestion by Frank Zimring* for "proxy studies" of the validity of Secret Service predictions of violent behavior. Without the use of reasonable proxies for attempted assassination, I doubt that validation research in this area can ever be done due to the extremely low base rate of the criterion. Such a research strategy could have useful secondary benefits as well. If it were found that agents in the course of their work were relatively accurate at predicting violent behavior, but that the violence was usually directed at persons other than those the Service protects ("proxies"), implications for cooperation with local authorities might be forthcoming. For example, to the extent that the persons predicted to be violent were also believed to be mentally ill, referral to a local department of mental health for civil commitment may be advisable. While this would not further the specific goals of the Secret Service, it would, to the extent the predictions were accurate, result in a decrease of violence in society. This would be no small thing. (2) There is reason to question the generalizability of the existing research on violence prediction to the concerns of the Secret Service. While agents certainly should be aware of the factors that seem to anticipate violence in other contexts, it is unknown to what extent those factors apply to the kinds of "political" violence of interest to the Secret Service. All of the research with which I am familiar deals with "street violence." Perhaps, as Frank's memorandum implies, only certain kinds of street violence are the functional equivalents of political violence. I raise this point because I am struck by the fact that one factorâraceâthat is always implicated in predictive accounts of street violence seems not at all implicated, or even implicated in the opposite direction, in political assassination. While blacks are highly overrepresented in street violence, all of the attempted presidential assassins I am aware of were white. *Page 183. 179
(3) "Process studies" of the reliability of agent decision-making on predictive tasks may be of value. Given the difficulty of empirically studying the accuracy (validity) of agent predictions, perhaps one useful fall-back strategy would be to study the factors that lead agents to disagree on the extent to which the same persons are a threat. This was begun in the 1976 study by Hay Associates that the Service commissioned. While improving the reliability of agent judgments will not necessarily lead to increased validity, such an increase in validity is unlikely to occur if the judgments of one agent bear little resemblance to the judgments of another. In this regard, if one could establish the decision rules employed by the "best" agents in making predictions (e.g., the most experienced agents, or those agents whose judgments are most respected by their peers), then at least one could begin to raise the performance level of all agents to that of the "best" agents. Again, it should be clear that this approach does not seek to answer how "good," in terms of validity, the "best" agents actually are. (4) Some attention might be given to the interventions that are initiated when an agent predicts that a threat exists. A first question might be the integrity of the interventionsâthat is, the extent to which they are actually carried out as planned. When a family member is requested to stay in contact with a threatening individual while a protected person is visiting the area, for example, does the family member actually do it? How often do the family members "lose" the individual for certain periods during the visit? How often, for that matter, do subjects manage to elude the Secret Service agents surveilling them? One could assess, in this regard, the relative integrity of various modes of surveillance (e.g., family member versus agents). 180
MEMORANDUM On Case Management and Decision-Making Loren H. Roth, M.D., M.P.H. Director, Law and Psychiatry Program Western Psychiatric Institute and Clinic University of Pittsburgh Pittsburgh, Pennsylvania I would like to present a somewhat different perspective about the problems of the Secret Service. To put it bluntly, I see Secret Service personnel as having two problems: first (of course) to protect the president; second, to ensure that agent behavior is "reasonable" or "appears reasonable" in relationship to the Secret Service mission. Some problems of the Secret Service agents appear no different than problems confronting emergency room physicians who must evaluate potentially violent persons. The problem is not just "what predictions do we make?" but also "how should we act?" or "what decision rules should we employ with respect to a host of potential next steps to be taken or not taken?" Assassination attempts and other attempts to harm public officials are extremely rare events. Developing methods of prediction for the most relevant outcomes will be difficult (probably impossible), even utilizing a "proxy" approach. The first goal, protecting the president, requires assessment of the factors that Saleem Shah has nicely identified in his paper (assailants, targets, and situations). Were I a Secret Service agent, I might, however, perceive the problem somewhat differently; for example, how should I act when a threat is brought to my attention? For the Secret Service, the problem is thus not only that of "prediction," but also of planning interventions and adequately monitoring and assessing the threat situation. The worst fallacy for the Secret Service might be for it to assume that dangerousness is predictable, or to act as if a decision about a subject's dangerousness may be reliably or validly made at a single point in time. In evaluating potentially dangerous persons in the emergency room, we consider different management strategies and alternatives. Decision rules (to act or not to act) will differ as pertains to (1) the decision to hospitalize voluntarily or involuntarily, (2) notify 181
the police or potential victims, (3) question other persons about the patient (e.g., relatives and friends), (4) obtain other information and records, (5) send the person home under the surveillance of relatives or friends, (6) phone the patient tomorrow, (7) do outreach, and/or (8) follow the person over time. The base rate for missile attack on Pittsburgh is presently zero. Nevertheless, there is (I hope) a system in place for tracking and monitoring of enemy activity, including a number of decision points wherein certain steps will be taken when a "signal" is detected indicating that missiles are or may be coming. By analogy, clinicians know that preventing future violent behavior, even for high-risk persons, involves continuous monitoring of subjects at risk not only to assess (â¢) changes in mental state and (b) the provocative value of current situations, but also to (c) continue the "minuet" with the person at risk. We most effectively prevent future violent behavior by adjusting our interventions over time in light of ongoing information and feedback about a person's courseânot by making long-term predictions. I thus suggest that one focus of our meeting should be to discuss and critique the procedures, monitoring efforts, and additionally collected data, that the Secret Service utilizes once a potential assailant is identified. The Secret Service's problem in protecting the president is not only scientific, but also political. To state this is not to denigrate the problem or to make a cynical comment. It is to emphasize that Secret Service agents face problems that are somewhat similar to problems clinicians face in emergency rooms. Secret Service agents and clinicians must make behavioral interventions, but also must develop and implement monitoring procedures which are reasonable (and which, were tragedy to result, would be judged reasonable) in light of present day knowledge. It is these "mundane questions" that interest me as much as the more theoretical issues about the prediction of dangerousness. We need to explore questions such as: Under what circumstances might the Secret Service perform two different evaluations by two different agents, rather than having to rely on the judgment of only one agent? What type of background information about subjects is routinely collected, how is it collected, what is its reliability and validity, and how is it used to establish profiles for continuing surveillance? What happens when a "Quarterly Investigation" subject "gets lost" to follow-up? 182
MEMORANDUM Secret Service "Dangerousness" Research Franklin E. Zimring, J.D. University of Chicago Law School Chicago, Illinois This memorandum outlines three research soundings the United States Secret Service might undertake. The first two would involve agency files and field staff follow-up, while the third requires a prospective experiment comparing the stress interview with alternative techniques. These proposals are sketched out in this form to provoke reactions from other committee members and to inspire similar preliminary memoranda from my colleagues. (1) Classification of dangerous subjects: a proxy validation. When subjects are brought to the attention of the Service, Initial investigations are conducted to determine whether, and to what extent, the subject represents a threat to personal security that falls within the jurisdiction of the Service. These predictions of possible dangerousness are a major field activity of the Secret Service: surveillance and case closure decisions are based on initial classifications as well as subsequent events. Thus, if the Service screens 100 persons who threaten protected parties, an important early task is identifying the relatively small number who are "dangerous." Obviously, it would be nice to know how effective the Service is in classifying subjects at present, and what objective criteria predict jurisdictional danger. The problem is that overt attempts against persons within the protective jurisdiction of the Service are too rare to permit direct validation. An imperfect alternative, still worth trying, would be to use five year follow-up data now collected by the Service to see whether those classified as dangerous are more likely to engage in events which are plausible "proxies" for jurisdictional danger. Two plausible proxies are (1) acts of interpersonal violence, and (2) suicide, particularly suicide attempted or committed in obviously attention-getting fashion. I am unprepared to argue that these "proxy" behaviors are perfect or even adequate substitutes for political violence. However, finding that the group classified as dangerous is more inclined towards these behaviors in the five years after coming to the attention of the Service would be an encouraging indication that current Service criteria are efficient predictors of overt behavior. Finding no difference between groups would be a sobering indication of lack of linkage between Service criteria and overt behavior during the 183
follow-up period. Further, this kind of follow-up study can provide associations between objective factors, known at the point of initial investigation, and subsequent overt behavior. Certain factors obviously related to risk, such as whether the individual is at large or under; restraint, would have to be controlled in the comparison. And the findings would be far from definitive. But the above study seems like the "least worst" way to approximate a validity study of Service predictions. (2) Determinants of the volume of threats. Little is known about what determines the volume of different kinds of threats against political figures over time. Since the Service does not control its own "in box," there may be value in modest explorations of variations in different types of jurisdictional threat over time, seasonal variations, and the extent to which publicized events precipitate increases or decreases in the volume of particular types of threatening gestures. Unlike the proxy validation study suggested above, preliminary soundings on variations over time or with events on the volume of jurisdictional threats could begin as a low cost, in-house activity of the Service. Indeed, some of the basic data may have already been collected for housekeeping or analytic purposes. Exposing social and behavioral science consultants to these preliminary soundings could then lead to more sophisticated designs for collaborative research. (3) A "stress interview" experiment. Questions about the stress interview occurred early in Secret Service request for technical assistance. Here there might be room for a random assignment experiment. Given the large numbers of new cases, the question is really how to determine to what extent increasing the "temperature" of an interview increases its informational yield. Randomly assigning all or some classes of Secret Service interviews between stress and non-stress modes might help answer the question. It is possible to find put whether stress interviews increase the initial confidence of the agent. Further, the five year follow-up might tell the Service whether this particular device improves prediction of dangerousness. 184
