Free beta access. 50 free AI tokens each week. Optional top-ups available. See what is included
Opening Word document…

Predictions of linear chains resulting high magnitude conflicts are impossible and always serve to discount ongoing violence

Masur 7 [Jonathan, Bigelow Fellow and Lecturer in Law, University of Chicago Law School, “Probability Thresholds”, 92 Iowa L. Rev. 1293]

Deference to the executive has become a judicial shibboleth and, in most cases, appropriately so n157 - the executive is peculiarly capable of [*1329] assessing and evaluating national-security threats. n158 But when courts afford this deference, inflation of the threatened harm is one of the natural consequences. As the subject matter moves further outside of the courts' competence - in other words, as the probability of an event drops and it becomes less susceptible to judicial evaluation - that effect increases, and deference becomes an ever more decisive factor in the weighing of costs and benefits. n159 Moreover, low-probability, high-magnitude harms are particularly susceptible to governmental inflation. The lowest-probability harms involve especially acute informational asymmetries between the government and the court (and the opposing party). The less likely an event is to occur - and the more dramatic the threat - the less likely it is that background information relating to the event's potential causes or antecedents will have seeped into the public domain. The explanation echoes in the nature of both low-probability events and high-magnitude harms. First, a potential harm of great magnitude n160 can assume only a few forms (remember that this must also be a harm catalyzed by speech). A war. A riot. A revolution. An act of terrorism or violence, necessarily with a substantially destructive weapon. A significant combat loss, such as the sinking of a troop transport. These categories define nearly the entire universe of relevant high-magnitude speech harms. Such harms share one essential, salient characteristic: they all concern topics that are especially - if not exclusively - within the expertise of the executive branch. n161 Only the [*1330] executive can know if the publication of a ship's schedule will lead to the sinking of a transport because only the executive knows when the transports sail or how much submarine activity they are likely to encounter. n162 Only the executive can know whether the publication of partial plans for the hydrogen bomb will actually make it more likely that some terrorist or rogue nation could acquire the bomb, since only the executive knows what other information is yet to be published or how difficult it is to obtain plutonium and tritium on the world market. n163 And only the executive can know whether opening the deportation hearing of a potential terrorist to the news media will give other terrorists valuable information about American anti-terrorism methods because only the executive knows what those methods are or how they relate to what the hearing might divulge. n164 High-magnitude harms are national-security-implicating harms, and national-security-implicating harms are the province of the executive. Second, and in concomitant fashion, low-probability harms rarely involve simple causal relationships - if A, then B, but only with small probability. More frequently they require a series of only loosely dependent probable occurrences, the interaction of disparate parties, or fortuitous conjunctions of unrelated events. The greater the number of moving parts and the more complex the relationships that a particular speech act must catalyze in order for a harm to occur, the more difficult and costly it is for a court to make an independent assessment of that harm's likelihood. The link is with Brandenburg's "imminence" requirement: once a potential causal chain has become so attenuated that an event will not necessarily occur "imminently," that chain likely involves a sufficiently large number of links that a court will have difficulty assessing the relevant probability absent some strong indication from the government. n165 It is endemic to the structure of American government that these types of decisions fall within the near-exclusive competence of the executive. n166 The question that remains is how to cope with the systemic distortions that result. [*1331] Throughout history, when the executive has exaggerated threats to justify speech restrictions, it typically has exaggerated low-probability threats, transmogrifying them from negligible dangers into constitutionally cognizable harms. n167 These are the types of dangers that will be more particularly within the executive's competence than the courts', and these are the types of dangers for which small adjustments in probability will produce the greatest impact. n168 B. Low-Probability Mathematics and the Difficulty of Weighing Costs and Benefits What the executive cannot do for itself, the courts and the populace will do for it. The cognitive-science literature offers a fairly robust description of the extraordinary difficulties that individuals have when coping with low-probability threats or even when rationally evaluating low-probability events. n169 Individuals tend to drastically overrate the statistical value of low-probability harms as a purely mathematical matter, substantially overpricing those harms when compared with harms of more easily quantifiable probabilities. It is worth pausing now to note that by the use of the word "individuals" here and in the following Sections, I do not mean only the great proverbially unwashed masses, as distinct from the noble and rational judiciary. A significant body of literature substantiates the belief that judges are highly susceptible to the same sorts of cognitive errors and biases that plague non-judicial decision-making. n170 And while there is reason to believe that experts may not succumb to all of the biases that interfere with a layperson's ability [*1332] to make rational choices, n171 judges are not experts (at assessing national security threats, for instance) - they are intelligent, educated generalists. n172 That fact is, of course, part of what makes deference to administrative agencies and executive actors appropriate in the first instance. n173 Absent any strong indication that judges will succeed in dodging the cognitive pitfalls that speckle typical rationality, this Article treats judicial and nonjudicial actors alike. n174 1. Low-Probability Value Inflation Study after study has demonstrated that individuals experience great difficulty, purely as a matter of estimation and intuition, when dealing with high-magnitude, low-probability threats. For many years, economists, lawyers, regulators, and other types of decision-makers routinely have used statistical measures of the value of a human life when performing cost-benefit analyses and making regulatory decisions. Analysts obtain these value-of-life numbers by looking at the salary premiums that workers demand for more dangerous jobs, n175 the amounts that people will pay to avoid certain every-day risks, n176 or by simply performing surveys that ask hypothetical questions regarding what an individual would be willing to pay to avoid a peril of some fixed probability. While these estimates are hardly exact, they are well accepted and widely used within the risk-management industry and usually conform to a reasonably limited range of values. The value of life is most commonly priced at around $ 7 million. n177 Importantly, though, these calculations are made almost always with reference to fairly [*1333] commonplace risks (workplace accidents and the like) that occur with significant probability. n178 Yet when the probability that a harm will occur drops significantly, toward a chance of one in hundreds of thousands or one in a million, individuals' responses to the same types of risk-assessment questions change drastically. The average person is willing to pay far more to avoid an exceedingly low-probability harm than the standard $ 7 million value of life would predict. In one survey, respondents confronted with a probabilistic threat to their lives of 1 in 100,000 were willing to pay approximately $ 220 to avoid the danger, which corresponds to a value of life of $ 22 million - already well above the typical $ 7 million benchmark. n179 But when a parallel set of respondents were asked about a threat to their lives of only 1 in a million probability, they were still willing to pay on average $ 100 to avoid it, placing their estimated value of life at $ 100 million - nearly fifteen times the usual number and five times the value for a threat of tenfold greater probability. n180 Another study displayed this effect in even starker relief. Respondents were willing to pay, on average, $ 15 per airplane flight to lower the risk of a terrorist attack on that flight from 1 in a million to 1 in 10 million. n181 This price, for a risk reduction of 9 in 10 million, corresponds to a value of life of approximately $ 16 million, already well above the typical amount. But this effect exploded when respondents were asked what they would pay to eliminate that final 1 in 10 million chance. The average person was willing to spend more than $ 16 in the face of only a 1 in 10 million probability of harm, pricing her life at a whopping $ 160 million. n182 Other studies have revealed a similar effect, with individuals even willing to pay more to avoid smaller risks than to avoid larger ones. n183 The consequences of this type of value inflation are straightforward and severe. A judge or juror performing cost-benefit analysis on a speech act that [*1334] threatens a harm of high magnitude and low probability will be inclined to substantially overvalue the "cost" side of the ledger, compared to a higher-probability harm of similar statistical value, simply because of the low probability itself. This willingness to overpay to avoid low-probability threats will result in systematic oversuppression of speech in the presence of low-probability, high-magnitude harms; too little speech will occur when compared to either the result of an unadulterated cost-benefit analysis or to a balancing of speech against only higher-probability harms. By screening out the very lowest-probability events and their drastic distortions of cost-benefit analysis (again, valuations of life increase fivefold as the probability of harm drops from 1/100,000 to 1/1,000,000), a probability threshold will act as a crude check on this anti-speech bias. 2. Low-Probability Risk Insensitivity The studies described above reveal two coordinated effects. First, as threats grew smaller, the actuarial value of life increased rapidly. n184 In the presence of small probabilities, individuals functioned as poor cost-benefit balancers. But what should not be overlooked is the fact that these inflated cost-of-life values resulted from peoples' willingness to pay approximately equivalent amounts of money to eliminate all "small" risks, even when they actually differed in probability by an order of magnitude. Low-probability value inflation is thus part and parcel of what one might describe as low-probability insensitivity - layperson inability to distinguish between various tiny probabilities, even when they differ by as much as a factor of ten. n185 Another experiment highlights this effect and is worth describing in some detail. A 1998 survey told respondents a fictional story about a chemical plant that was at risk of releasing dangerous chemicals into the atmosphere above a nearby community. n186 The survey described an insurance company engaged in writing policies for affected community members and then provided an estimate of the probability of a chemical release (some surveys listed it at 1 in 100,000, some at 1 in 1 million, and some at 1 in 10 million). n187 Each survey provided only one of the three values. The surveys then offered comparison data regarding the probability of a car accident and asked four questions: [*1335] (1) whether the plant posed a serious health and safety risk to those living in the community, (2) whether the plant could operate in a manner that was safe for the community, (3) how serious is the risk of death posed by the plant, and (4) how close to the plant respondents would be willing to live. n188 Remarkably, respondents proved almost completely insensitive to the ten-fold gradations in risk between each of the three scenarios. In their responses to the four questions, they graded the threat of a chemical release as "medium" (corresponding approximately to a risk of "3" on a scale of 1 to 5) irrespective of which of the three probabilities appeared on the survey they were given. n189 The survey respondents apparently viewed small risks as exactly that - and only that; the statistically enormous differences between them did not even register. As one commentator suggests, it is likely that for risks in these domains "the relevant probability is "low, but not zero,' and ... finer distinctions are unhelpful." n190 A number of other studies have documented the same effect even among individuals presented with two differing risk situations side-by-side. n191 In contexts ranging from automobile fatalities, n192 to serious automobile injuries, n193 to hazardous waste, n194 to pesticides used on fresh produce, n195 and even to oyster-borne illnesses, n196 survey respondents displayed an utter inability to modulate their willingness to pay for increases in safety according to how much those safety increases actually would diminish the probability of harm. In one study, nearly half of survey respondents were willing to pay [*1336] more for a reduction in the probability of death from an unspecified cause from 2 in 100,000 to 1 in 100,000 than they were for a reduction in the probability of death from a similarly unspecified cause from 20 in 100,000 to 15 in 100,000 - despite the fact that the second reduction is fivefold greater. n197 It seems that individuals simply experience great difficulty at evaluating low probabilities with any consistency or accuracy. n198 Like its partner value inflation, probability insensitivity will tend to bias cost-benefit results systematically in favor of curtailing speech to protect against prospective harms. As probabilities of harm decrease, individuals pay less and less attention to fine gradations, pushing results towards the high-probability end of the spectrum. Failure to recognize or appreciate the true size of a small probability represents the type of cognitive error that tends to compel decision-makers to accept restrictions on otherwise beneficial activities that they would not permit under an accurate cost-benefit analysis. 3. The Tyranny of Small Mistakes Compounding the errors introduced into low-probability cost-benefit analysis by the cognitive biases described above is the fact that when it comes to low-probability events - indeed, when it comes to small numbers generally - small mistakes can produce enormous consequences. Estimations of large-magnitude damages are relatively robust. Underwriters, actuaries, and other risk assessors have become adept at gauging the damage that a putative danger might cause within reasonable limits. Working values for the cost of human life, major injuries, or damage to particular types of property exist, and risk experts employ them widely. n199 But more to the point, small mistakes are dwarfed by the enormous monetary baselines employed in such risk assessments. A very conservative estimate might place the value of the damage from the September 11th attacks on the World Trade Center at approximately $ 25 billion. n200 Even if that figure is off-target by a substantial amount, such as $ 2 billion, the overall effect on any resulting computation will be less than 10%. Probability assessments, on the other hand, introduce far greater uncertainty into any calculation. As this Article describes, speech harms [*1337] depend upon the actions of one or more intervening parties, as well as any number of other compounding or confounding factors. The probabilities involved in these events - which themselves invoke questions regarding, for example, the motivations or capabilities of some terrorist or military organization - hardly conduce to easy rational estimation. n201 There is a reason why these cases are discussed so frequently in qualitative terms and so rarely with reference to quantified probabilities. n202 Moreover, small errors in low-probability estimations introduce enormous variation into cost-benefit calculations. The mathematical point is both mundane and crucial: if the government believes that an event will occur with a probability of 2 in 100,000 if certain speech is allowed to take place, but the actual probability is, in fact, 1 in 100,000, n203 then the statistical estimate of harm is off by 50%. When the uncertainties endemic to assessing speech-related harms are coupled with standard probability indifference, errors of this magnitude seem almost inevitable. It is important to note that low-probability errors of this type are not themselves systematically biased toward either overestimation or underestimation of statistical harm. The mathematics of small probabilities work only as a lens, magnifying the effect of whatever errors have already taken place. Yet, as the rest of this Section describes, those errors are likely to be mistakes of over-estimation: miscalculations tending to inflate and overemphasize the dangers attendant to speech. Enhanced by the mathematics of small numbers, these upward biases will magnify ephemeral threats into constitutionally significant events. [*1338] C. Dreadful Harms, Emotional Responses, and the Availability Heuristic In addition to the purely mathematical errors that attend efforts to cope with small numbers and low probabilities, the types of low-probability, high-magnitude harms involved in free-speech cases often invoke strong emotional responses that further skew attempts to perform rational cost-benefit analysis. Two complementary cognitive effects are manifest. First, individuals tend to overreact to harms from "dreadful" or "unknown" sources or other types of harms that invoke a particularly significant emotional response. High-magnitude harms resultant from speech are almost always of this type; terrorism, riots, and the like are not the stuff of ordinary experience, and their effects upon the public mood is far greater than the number of casualties would indicate. Second, individuals tend to overvalue harms that have special temporal salience, harms that resonate with recent experience, or threats that seem similar to ones they have heard about in the near past. The effect of this type of cognitive bias - the "availability heuristic" - is obviously context-specific; modern citizens are not likely to inflate the threat of a troop transport being sunk in the same fashion as Americans living during the Second World War might have. Yet, at least one class of potential high-magnitude speech harms, terrorism, remains of obvious salience today. Moreover, the low-probability threats that present themselves in any era - those threats that galvanize the government into attempts to curb speech - are likely to have great contemporaneous relevance, if for no other reason than for the fact that the government will attempt to make the citizenry aware of them as a part of its campaign to justify its restrictions on speech. High-magnitude, low-probability speech harms are thus, by their very nature, likely to tap into these heuristic biases. The result is a systematic overemphasis of low-probability threats, an effect that the First Amendment's probability threshold is meant crudely to counteract. 1. Emotions, Fear, and First Amendment Events It is not surprising to learn that individuals respond more vociferously to emotionally charged threats or events. Emotion is, by definition, something that individuals experience outside of their purely rational modes of cognition - but that nevertheless influences individual preferences n204 - and a phenomenon that taps into both rational and extrarational centers will undoubtedly produce some type of heightened effect. What is striking is the degree to which emotional appeals can penetrate what ought to be purely rational decision-making processes and the extent to which they can skew the outcomes of those endeavors. [*1339] Experiments have demonstrated that individuals are willing to pay substantial premiums to avoid emotionally laden dangers above and beyond what they would spend to eliminate banal risks of the same magnitude and probability. In one study, subjects were willing to pay as much as 45% more to avoid a risk that had been described to them in emotional terms than they were willing to pay to avoid the same risk - a risk of death with the same probability - when they had been offered the risk in emotion-neutral terms. n205 Another study obtained similar results by looking at responses to threats of electric shock. n206 These examples, alongside many others, n207 reveal a more particular trend: individuals will respond more strongly to "dreadful" or "catastrophic" dangers and to dangers that they cannot control than they will to typical, everyday harms. n208 The effect can become extremely pronounced in the context of improbable threats, as individuals "focus[] their attention on the outcome rather than the low probability of the harm." n209 It is the "catastrophic potential" of a threat - not its low probability or the statistical risk it actually poses - that dominates individual assessments. n210 The most immediate application of this notion is to terrorism, and indeed, many commentators have concluded that mere invocation of the threat of terrorism "evokes vivid images of disaster, thus crowding out [rational] probability judgments" and enhancing the amount that individuals would sacrifice to avoid such a harm. n211 But the principle extends to many of the types of low-probability, high-magnitude harms that might be expected to arise from speech. To some degree, the effect reaches all low-probability, high-impact harms; a "catastrophic" harm is simply one that is large, and individuals will naturally respond most strongly to large, fearsome threats of all types. But the harms associated with reckless speech are particularly susceptible to these types of emotional over-reactions, partly because they necessarily involve the machinations of faceless third parties who act upon the speech. Wars, riots, terrorism, military disasters, and the [*1340] chaos of the mob - these speech-induced threats conjure up dreadful images and inspire fear in a way that even mass disasters such as earthquakes and floods apparently do not. n212 Fear is a powerful inducement to action, n213 and speech-borne threats catalyze reactions far beyond what quotidian risks such as car accidents might produce, n214 whether or not they offer a comparably smaller statistical threat. n215 One intuitive consequence of this response is that individuals will tend to focus systematically upon the "worst case," since it is that case that will trigger the strongest emotional response. n216 In one experiment, subjects received information about a potential risk from two parties: industry and the government. n217 For some subjects, the government described the risk as high and industry described it as low; for some subjects it was the opposite. Yet, in all cases, respondents viewed "the high risk information as being more informative," regardless of the source. n218 Experiments such as this one would make it seem that our skewed perspective on the probability of dreadful events might be triggered, in part, by our pessimistic willingness to believe that they may well occur - whether or not that willingness is attributable entirely to the vivid emotional picture that such harms paint. n219 The likely consequence is a decision-making populace prepared to believe the worst whenever they are provided with competing characterizations of a level of risk. In speech cases, the worst-case scenario will of course be the one suggested by the censor. Therefore, it may be that even if the government did not hold an information monopoly over assessments of speech harms n220 - and even if the government's opinion were not entitled to [*1341] special deference n221 - the mere fact that the government was forecasting the most severe outcome would tilt the scales against other predictions. 2. The Availability Heuristic Operating alongside the overemphasis on "dreadful" or terrible harms is the tendency of individuals to overestimate the likelihood that an event will occur when an example of that event comes readily to mind. n222 This effect, known as the "availability heuristic," is the product of the very human tendency to focus on whatever evidence is most cognitively salient at any given moment. n223 It is by definition context-specific; what is news today - and therefore foremost in the minds of most individuals - may well not be news tomorrow. Thus, the availability heuristic may favor certain impressions on one day and a completely different set mere weeks or months later. Nor does the availability heuristic necessarily systematically favor the government or the speaker. A widely publicized story of police abuse - the unwarranted beating of a suspect, for instance - can trigger the availability heuristic and cause the public to misjudge the extent to which law-enforcement authorities are overstepping their lawful boundaries. One week later, news of the urban riot that erupted as a consequence of this act of police brutality could induce the same public to believe that the citizenry has run amok and that the authorities are doing an inadequate job of policing the community. n224 The only factor is how easily a decision-maker can summon an instance of the event in question to mind. It is worth noting also that the evidence and theory behind the availability heuristic would seem to conflict to some degree with the observation - offered by the same commentators who have recorded the enhanced public reaction to particularly dreadful or frightful events - that individuals respond more strongly to "unknown" dangers. n225 Yet the disagreement is not, in fact, so stark. The answer is revealed through the types of risks that these commentators classify as "unknown": nuclear accidents, genetic mutations, food irradiation, and the like. n226 The claim, then, is not the facially bizarre assertion that individuals are responding to risks that they do not know of; instead, they react most strongly to threats [*1342] that have been much discussed within the press but that are sufficiently complex or "scientific" that the average layperson cannot comprehend them. The availability heuristic would predict similar outcomes. The impact of this tendency on how the American public perceives the types of low-probability harms that speech might engender is far from one-sided. Certain types of these threats would appear foreign or anachronistic to modern eyes. Recent American history is hardly full of examples of speech acts inciting urban riots or pogroms, n227 and the United States has not had a troop transport sunk due to the publication of its sailing date in quite some time. The availability heuristic might, in many cases, lead decision-makers to underestimate the probability that those types of harms will materialize; these are among the types of catastrophic events that are hard to imagine taking place in the modern United States. In fact, the period of McCarthyism during the 1940s and 1950s, an era characterized by what are now viewed as excessive assaults on the freedom of speech, n228 arguably retains greater salience than nearly any other series of events with such an obvious speech valence. The availability heuristic thus might augur generally greater protection for speech than conventional cost-benefit models would prescribe. There is, however, at least one more powerful counter-consideration: the importance of terrorism. As of 2005, the terrorist threat remains perhaps the single most salient issue of public policy, and the events of September 11th maintain a position of prominence in the minds of people across the country. n229 The availability heuristic thus would predict that decision-makers will systematically and severely inflate their estimates of the probability of a terrorist event as one or more memorable examples come immediately to mind. Not coincidentally, the vast majority of cost-benefit speech cases that have come before the courts in recent years - and the vast majority that are expected to arrive in courtrooms in the near future - are terrorism cases. n230 [*1343] The availability heuristic may not be applicable ubiquito sly, but it is likely to have a substantial impact in the most crucial of contexts

Sign in to recut, update, or save this card.