0521280
Dangerousness is a Dangerous Concept
-1-
0521280
Introduction This paper will present the argument that dangerousness is indeed a dangerous concept with an important qualification: when the concept is used wrongly. It has to be remembered dangerousness is episteme. Knowledge in and of itself is not dangerous. Just as E=MC2 can be used to power hospitals it can also be used to create bombs of unimaginable destruction; dangerousness can serve our society well or poorly. The paper is split into two parts. First there will be a methodological analysis of dangerousness. The paper will show there are differing levels of predicting dangerousness, from completely subjective evaluations to more objective formalism of actuarial equations. But whatever method used, all have serious problems. It will then show how predictive models are stuck in a Newtonian view of science rooted in the 19th century and has not taken into account what the ‘new sciences’ reveal about our predictive ability which offers a much less optimistic account of even the ‘hard’ sciences ability to predict. Secondly a more substantive analysis will be undertaken of dangerousness in practice. It will be shown how the concept is used in two main areas in the US: capital statutes and the civil commitment for sexual offenders. The particular jurisdiction is chosen for the simple fact that predictions of dangerousness can deprive people of their lives, analysis here is most important. It will be shown that dangerousness is actually dangerous by linking it to the first part of the paper - its inability to predict. The law is preoccupied by the ‘majesty’ of scientific formalism that psychologists purportedly provide and is worryingly changing the very nature of our (democratic) legal systems,
-2-
0521280 without the evidence warranting it. As Michael Foucault identified1, a paradigm shift is occurring in our legal systems where non judicial actors (psychologists etc) take on judicial power, they have the power to deprive of liberty and even life, the burden of proof is relaxed and due process in criminal trials is circumvented. Most importantly the law judges a set of facts that have no basis in reality and accepts deterministic and reductionlist arguments when jurisprudence has long been based on culpability, freewill and a set of facts arising.
A history of dangerousness It is sometimes forgotten that the concept of dangerousness existed prior to solid clinical methods of prediction. It can be traced far back as 1911 in the US 2 and 1838 in France.3 Only the more modern approaches will be analysed in this paper. Prior to 1966 there was relatively little attention paid to how well clinicians assessed risk. In 1966 the U.S Supreme Court in Baxstrom4 held 967 offenders detained longer than their maximum sentence, due to dangerousness predictions, unconstitutional. Most offenders were released to local hospitals and eventually onto the street. After 4.5 years only two men 1
For example he stated “a general process has led judges to judge something other than crimes; They have been lead in their sentences to do something other than judge ; and the power of judging has been transferred , in part , to other authorities than the judges of the offence. The whole judicial process has taken on extra-judicial elements and personnel” Nola, Rescuing Reason: A Critique of Anti-Rationalist Views of Science and Knowledge, (Springer:2003) p443 2
Massachusetts Briggs Act cited in Predicting and Controlling Dangerousness http://faculty.ncwc.edu/toconnor/294/294lect04.htm and Günter Blau and Hans-Dieter Schwind, Festschrift für Günter Blau zum 70. Geburtstag am 18. Dezember 1985, (1985: Walter de Gruyter) p568. There was even discussion of it stretching far back as 1870s see Oliver Wendell Holmes, ‘The Path of Law’ 10 Harvard Law Review 457 (1897) at 470-471 where he discusses dangerousness as an important jurisprudential concept. 3
Castel ‘From dangerousness to risk’. In Burchell, G., Gordon, C & Miller, P. (Eds.), The Foucault effect. studies in governmentality (pp. 281–298). London: Harvester Wheatseaf (1991) p296 at note 17 4
Baxstrom v Herald 383 U.S 107 (1966)
-3-
0521280 had been re-incarcerated for committing a violent crime.5 Throughout the 1970’s more studies demonstrated that clinicians had relatively little competence in predicting violent behaviour.6 Monahan reviewed the first generation studies and concluded “the upper bound level of accuracy that even the best risk assessment technology could achieve was of the order of 0.33”7. To put it into perspective 0.5 is the accuracy one can achieve for predicting the outcome of a coin toss. The general problem for the first generation was reliance on clinical methods (‘clinical judgement’). These could not be objectively verified and they tended to over predict. In one study members of a maximum security hospital were asked “if 100 men were randomly chosen from this maximum security hospital and released, how many would commit an offence causing bodily harm to another person within one year?” 8 All made estimates too high; medically trained staff made higher estimates than non medical staff. Another study shows how psychiatrists are not even able to demonstrate prediction
5
Steadman, H. J. & Cocozza, J. J. (1974) Careers of the Criminally Insane: Excessive Social Control of Deviance. Lexington, MA: Lexington Books. 6
Thornberry, T. P. & Jacoby, J. E. (1979) The Criminally Insane: A Community Follow-up of Mentally III Offenders. Chicago, IL: University of Chicago Press. Cocozza, J. J. & Steadman, H. J. (1976) The failure of psychiatric predictions of dangerousness: Clear and convincing evidence. Rutgers Law Review, 29, 10841101 both cited in Dolan and Doyle, Violence risk prediction The British Journal of Psychiatry (2000) 177: 303-311 Megargee, E.I. (1970). The prediction of violence with psychological tests. 2 Current Topics in Clinicaland Community Psychology 98 (C.D. Spelberger ed.) cited in Trowbridge, ‘Age and Recidivism: How Accurate are Our Predictions’, Washington Criminal Defence, Nov 2004, Vol 18. No.4 7
Monahan, J. (1984) The prediction of violent behaviour: Toward a second generation of theory and policy. American Journal of Psychiatry, 141, 10-15 8
Quinsey, V.L. (1981). The long-term management of the mentally-disordered offender. Mental Disorder and Criminal Responsibility 137 (S.J. Hucker, et al eds) cited in Trowbridge Op Cit
-4-
0521280 expertise superior to that of a lay person.9 Buchannan10 and Castel both agree the reason for the higher prediction is probably the consequences of the two types of mistake. As Castel puts it: “[w]hen in doubt it is better to act, since, even if unfounded intervention is an error, it is one that will certainly never be known to be such; whereas if one abstains from intervening and the threatened act should still materlaize [Sic], the mistake is obvious and the psychiatrist is exposed to blame”.11
Other problems have been highlighted including the fact clinicians are seldom required to describe in detail the processes by which they estimate risk12. Monahan urged second generation thinking to develop actuarial techniques to resolve the problems of over-prediction and lack of objectivity.13 Actuarial tables based on ‘static’ variables such as type of crime, age and sex of victim, school records, onset of criminal behaviour were all integrated. They objectively score and provide probabilistic estimates of risk based on established empirical relationships between their items and the outcome of interest.14 For example research establishes lack of empathy is a trait of psychopaths, so that is integrated into equations. These eventually resulted in methods
9
Quinsey, V.L. & Ambtman, R. (1979) Variables affecting psychiatrists’ and teachers’ assessments of the dangerousness of mentally ill offenders. 47 Journal of Consulting and Clinical Psychology 353 10
Buchannan ‘Risk and Dangerousness’ Psychological Medicine, 1999, 29, 465-473, at 468 11
Castel Op Cit P283 12
Buchannan Op Cit p467
13
Monhan, J. (1984) Op Cit. 14
Seto, ‘Is More Better? Combining Actuarial Risk Scales to Predict Recidivism Among Adult Sex Offenders’ Psychological Assessment 2005 Vol 17 No2 156-167 at 156
-5-
0521280 such as the PCL-R15 which determined psychopathy by characterising various interpersonal traits associated with psychopaths. A score above 30 is the prototypical psychopathic range. The PCL-R was integrated into the Variable Risk Appraisal Guide (VRAG) and its accuracy was rated at 0.74.16 Other methods achieved similar results and a recent study17 combining all the major methods (VRAG, SORAG, RRASOR and Static99) has shown that there is no advantage of combining all to predict offending. In short 0.75 and the upper band of 0.80 accuracy is likely all we can squeeze out of these methods.
Problems with the methods Two general problems have already been set out: over prediction and lack of objectivity, at least for clinical methods. Actuarial approaches were seen as resolving these but these have significant problems of their own. Buchannan has identified some problems. Firstly using variables in actuarial equations is mathematical but determining them is not. For example a lack of remorse for past misdeeds is a prognostic sign and may amount to an informal mathematical model, but the decision as to whether or not the patient is remorseful can hardly be mathematical.18 Unlike with the ‘hard’ sciences we cannot take simple measurements of ‘time’ with a stop-watch, to quantify its variable in the speed =distance/time equation. Psychiatric variables are inherently subjective to the 15
Hare, R.D. (1991). The Hare Psychopathy Checklist-Revised, Multihealth Systems, Toronto. 16
Janus, E.S. and Mechl, P.E. (1997). Assessing legal standard for predictions of dangerousness in sex offender commitment proceedings, Vol 1, 3 Psychology, Public Policy & Law, pp 33-64. cited in Trowbridge. 17
Seto Op Cit n13 18
Buchannan Op Cit p467
-6-
0521280 observer19. These can be exacerbated by lack of cultural relativity. As Corbett and Westwood have argued,20 in the context of Dangerous and Severe Personality Disorder, the social backgrounds of those experts in creating/applying diagnostic criteria are most likely middle class. Thus they may impart social judgements under the guise. What may be perfectly normal behaviour among a culture or class may be deemed a variable for risk in formulating an equation or even assigning a variable to an already formed one. Unfortunately a recent study into the area of cultural bias has proven this thesis correct21. The study shows probation officers saw the presence of a white family as a positive influence and an asian one a negative influence. Using dangerousness predictive methods opens up the door for institutional racism (conscious or subconscious). Political and social ideologies influence even ‘hard’ sciences like Biology. That is why Soviet scientists saw evolution as a matter of intra and inter species co-operation and western scientists saw it purely as Darwinian competition; the conclusions reflecting their respective economic modes of production. If a hard science is influenced by politics something inherently subjective, like assessing an offender, will most likely do the same and to a greater extent. As will be shown such early initial discrepancies of measurements in a deterministic system is liable to lead to ‘chaos’ further down the line. Completely ruining any predictive power these methods have.
19
While strictly speaking Hezenberg’s Uncertainty Principle also attaches such conditions to the natural sciences, the subjectivity is no where near as prevalent. We know this from our everyday experiences, we can measure the time of a 100m race to an accuracy of 1/10000th of a second. 20
Corbett and Westwood ‘Dangerous and severe personality disorder’: A psychiatric manifestation of the risk society Critical Public Health, June 2005; 15(2): 121–133 p125 21
Hudson and Bramhall, “Construstion and ‘asianness’ in risk assessments by probation officers” (2005) 45 British Journal of Criminology 721
-7-
0521280 Buchannan also cites other problems with actuarial approaches. As stated earlier empirical research establishes which variables to integrate into equations. However, the studies themselves are inconsistent. For example schizophrenia studies on violence being more likely if patient is describing psychotic symptoms are conflicting. Several studies have suggested it is, while others have concluded it is not22. Rare events cannot be quantified into variables either, due to their transient nature. Capgras delusions are wisely reported predictors of violence but not widely rated. The resulting lack of statistical power means no conclusions can be drawn. The same is true of delusional mood. These episodes are so brief they are unable to be captured on a scale.23 Actuarial methods cannot establish which piece of research is relevant to the patient either. As Buchannan argues, “all statements on statistical probability rest on the assumption that the circumstances in which we are required to predict resembles circumstances which we have prior knowledge”24 and “the grounds for this assumption of resemblance cannot be statistical 25” In other words despite the purported objectivity actuarial methods ostensibly give, at their heart lies clinical methods of evaluation with all their weaknesses. In the end assigning variables is a matter of clinical judgment and not objective measurement.
22
Buchannan, Op Cit p468 23
ibid, p470 24
ibid 25
ibid
-8-
0521280 As early as 1976 Megargee warned personality and situational factors must be taken into account26 and many others all said the same27. Today it is accepted that dangerousness is a complex interaction between psychological, biological and environmental factors.28 In accepting this a nail has been driven into the predictive ability of dangerousness – it is no longer a closed system since it is not focused on the individual. While the improvements made from adding environmental factors is obvious 29 (0.33 to 0.75; increase of 0.42) adding environmental factors constrains any further improvement. We know this because of what the ‘new sciences’ (as they are collectively known) tell us about our predictive ability – it is heavily constrained.
The new sciences Little has been written about chaos theory and quantum theory in relation to dangerousness. But it cannot be understated how pertinent such theories are to our understanding of predicting dangerous behaviour. As one psychologist who notably attempted to incorporate the new sciences said in his lecture: “Prediction is prediction, after all, regardless of whether one is predicting the behaviour of subatomic particles, or that of sentient, multi-cellurlar organism.”30 The closest analogy one can bring of the 26
Megargee, E.I. (1976). "The prediction of dangerous behavior". Criminal Justice & Behavior, 3, 3-21. Cited in Cohen, D.Notes on the Clinical Assessment of Dangerousness in Offender Populations Psychiatry Online [www.priory.com/psych/assessin.htm] 27
Hepworth (1985), Monhan (1988), Steadman & Ribner (1982) all cited in Cohen [ibid]
28
ibid 29
Important note: Although of course not all improvements were made down to the incorporation of environmental factors into predictive models. Some came about through better empirical source data. 30
Hart, European Association of Psychology and Law, Lisbon, Portugal (June 5-8, 2001), Address: Complexity, Uncertainty, and the Reconceptualization of Violence Risk Assessment, p3
-9-
0521280 importance of chaos theory in prediction for human behaviour is in the field if economics. Economists have long used actuarial methods to predict how the market will behave. It is concerned with predicting how economic actors (viz people) behave according to certain market conditions (viz environment) and vice-versa. Chaos theory has had a considerable impact on this field31. A field primarily concerned with the actions of people and predicting their behaviour. The major problem is in accepting environmental factors into predictive models we turn a closed system into an open one. Classical probability theory was intended to model the behaviour of closed systems. All the actuarial scales are rooted in this model32. Society is an open system with “permeable walls so that matter, energy and information or entropy may cross them in either direction.”33 In open systems we must take into account Chaos theory. The theory basically states that in a complex, non-linear system (such as human-society) a small change in the initial input could produce a massive change in the output. The oft cited ‘butterfly effect’ is the prototypical example of this. A butterfly flapping its wings (small effect) in London could set off a hurricane in Massachusetts (big effect). Thus complex, non linear systems are sensitive to initial conditions. Lorenz, the pioneer of Chaos theory, modelled a computer simulation. He found that if he started his simulation with values that were only slightly different from the original, the difference being that one set is accurate to six decimal places and the 31
Murray Rothbard, Making Economic Sense, Chapter 5 ‘Chaos Theory’ available online as an eBook: http://www.mises.org/econsense/ch5.asp 32
See Hart Op Cit p6 33
Çambel, A. B. (1993). Applied chaos theory: A paradigm for complexity. San Diego:Academic Press. P42 cited in Hart p6.
- 10 -
0521280 second set down three places, then the weather produced by the computer soon veered wildly from the original34 The difference here is miniscule - a mere three decimal places versus six (say 1.001 and 1.000001). The problem hinted to before about assigning initial values in the actuarial equations would be manifestly larger. Not a matter of three decimal places but the very fact if a variable even applies to an offender. Chaos theory tells meteorologists that they cannot predict the weather accurately a week in advance. It tells economists they cannot predict long term market behaviour; it tells Biologists they cannot predict long term migratory behaviour - all due to the sensitivity of the initial conditions. What then does it say to psychologists? Surely the complex interaction between environmental and biological factors exhibits the same uncertainty. Simply stated, we cannot predict long term behaviour of all patients, only those who we can accurately assign variables to. However, due to the inherent non-scientific method of assigning variables we cannot even be sure any variables are assigned correctly. Even more fundamentally the actual equations and tables as Buchmannan pointed out, have conflicting empirical data or may be missing some important information. Chaos theory shows how actuarial methods cannot be relied upon if their initial values are subjective. Unfortunately some psychologists have fundamentally misunderstood how the new sciences apply to dangerousness. Hart is guilty of this, he summarises 400 hundred years of advances in mathematics and physics in an attempt to prove his thesis that nothing is predictable. He argues due to the advent of quantum theory 35* we cannot 34
http://www.marxist.com/science/chaostheory.html
35
Quantum theory is about the very small: subatomic particles. Classic probability theory (Newtonian mechanics) held that the universe was entirely predictable. If given enough information the mathematics of Newtown’s equations could predict the behaviour of all particles in the universe; effectively predicting the future. However, quantum mechanics holds that we cannot predict anything with a 100% certainty and only assign probabilities to it. For example, Newtown would say particle Z exists at a certain point in
- 11 -
0521280 predict anything with certainty. This is overstated in his analysis. He correctly argues that physics cannot show which way a coin will land, tossed once 36. The equations would be unimaginably complex and we cannot obtain the data in the first place. However, despite physics being unable to explain which way a coin will land tossed once, tossed a hundred, a thousand or even a million times it (rather mathematics) can give us the probability of heads arising 50% and tails 50%. Physics cannot give us the causal explanation of each subatomic particle involved in the event of a coin tossing, but probability theory can show us it is a 50/50 chance. The same is true of human behaviour. We might not understand why but we know it to be the case that those with a high rating on the VRAG are likely to offend. Hart misunderstands the two types of probability which Buchmanan citing37 David Hume in his article correctly identified as one founded on chance and the other which arose from causes38. Unfortunately most of Hart’s lecture is criticising the probability arising from causes. This is only relevant to explaining the variables chosen in the first place i.e. the diagnosis and empirical data. Even with this though, medical space/time. Where as quantum theory would say particle Z has a 99% chance of existing at a certain point in space/time. Hart attempts to argue this fact of quantum mechanics means we cannot predict anything. However we know this to be incorrect from our everyday experiences. Otherwise life as we experience it would not exist. Our world would be full of ‘quantum tunnelling’ where people can walk though walls and other microscopic quantum effects which would manifest in macroscopic objects. What Hart misunderstands is the importance of Planck’s constant (ћ). If ћ were a big value then indeed we could not predict anything because quantum effects would apply to bigger macroscopic objects. But because ћ is tiny (1.05 x 10-27) quantum mechanics only ruins predictive ability at the tiny microscopic subatomic level. It is simply incorrect to argue it does at the macroscopic level. And since human brains are made up of billions of cells, it qualifies as a macroscopic object. See generally Greene, The Elegant Universe (Vintage Books: 2003) p85-86 for an excellent, easy to understand, explanation. 36
Hart p9 37
Buchanan p469 38
Hume, (1739). A Treatise of Human Nature. Volume One. Published as A Treatise on Human Nature. Volume One (ed. T.Green and T. Grose), 1874. Longmans: London.
- 12 -
0521280 science has long since strayed away from the absolute causal certainty that physicists seek. A doctor knows mesothelioma is usually caused by exposure to asbestos, even though he is not sure of the exact aetiology. He may not know the exact cause but there is a high degree of probability it is caused be exposure to the asbestos. That is the nature of medical science, exact causes, down the subatomic level cannot be given. Unfortunately this is most likely why Hart has been seen by his peers as being ‘unscientific’39 and his relevant points not being widely accepted despite being tremendously important. To summarise this section we have shown how dangerousness is unreliable and whatever methods used all have their problems. It would indeed be dangerous to use these methods to justify execution and detention, however, as will be seen in part two this is unfortunately the case.
Application of Dangerousness This section will concentrate on application of dangerousness in the US under civil commitment for ‘sexual predators’ and capital punishment statutes. In so doing it will be seen there is bias toward psychologists expert testimony in predicting dangerousness which the evidence as presented above, does not justify. There is also a misunderstanding about how base-rates affect the number of false positives.
Capital Statutes
39
Hart p16
- 13 -
0521280 The US Supreme Court in Furman40 struck down capital punishment statutes as violating the 8th Amendment of the US Constitution on ‘cruel and unusual punishment’. The court held, inter alia, the current capital statutes were arbitrary since they had no guidelines for juries or judges. As a result some states responded to the decision by amending their capital statutes to require the jury to determine the defendants’ future dangerousness. This is where predicting dangerousness is at its most dangerous – when life is deprived because of it. 21 states require dangerousness to be predicted; Texas and Oregon require the jury to predict future danger in their decision of handing out capital punishment.41 The case of Barefoot42 is indicative of the problem. Under the Texas statute43 two special questions submitted to jury including whether “there is a probability that the defendant would commit criminal acts of violence that would constitute a continuing threat to society”. At trial the State called two psychiatrists who, in response to hypothetical questions, testified the petitioner would probably commit “further acts of violence and represent a continuing threat to society”44 Neither had examined the defendant. On appeal the defence argued that psychiatrists: 1) individually and as a group are incompetent to predict dangerousness to an acceptable degree of reliability. 2) Psychiatrists should be required to examine defendant personally before testifying. As such they petitioned the court to overrule the conviction. All arguments failed. The judgement is a vivid example of exactly why dangerousness is dangerous: its probative 40
Furman v. Georgia, 408 U.S. 238 (1972) 41
Sorensen, ‘Criminology: An Actuarial Risk Assessment of Violence Posed By Capital Murder Defendants’, 90 Journal of. Criminal. Law. & Criminology 1251, 1252 (2000) 42
Barefoot v. Estelle 463 U.S. 880 (1983) 43
Tex. Code Crim. Proc. Ann., Article 37.071 44
p884 of the judgement.
- 14 -
0521280 value is minimal and prejudicial effect maximal. This can lead to deprivation of life because judges and juries are impressed by quasi-‘science’ of prediction. Against 1) The majority45 cited previous jurisprudence under Jurek46 where seven Justices rejected the claim that it was impossible to predict future dangerousness. What was essential was the “jury have before it all possible relevant information”47. Against 2) the court cited48 a previous case, Edgar49 where the court was of the opinion that expert testimony, even if hypothetical, is commonly admitted as evidence where it might help the fact finder do its assigned job. Also the court argued that the ordinary rules of evidence governing the use of expert testimony apply to dangerousness, even in capital cases. The reasoning is dangerous because it overstates the probative value of psychiatric testimony; that is, its prejudicial effect is greater than its evidentiary value. In 1983 dangerousness predictions were wrong 2/3 times50. The Court reasoned that the ordinary rules of evidence, the adversarial nature of the courts and the availability of cross examination would stem any prejudicial effect. Justice Blackmun in his powerful dissent puts it best: The Court holds that psychiatric testimony about a defendant's future dangerousness is admissible, despite the fact that such testimony is wrong 45
Ibid , p896 46
Jurek v Texas 428 US 262 (1976) 47
Barefoot p897 48
p903 49
Spring Co v. Edgar 99 US 645 50
According to the APA brief of the case, quoted by Justice Blackmun at p916
- 15 -
0521280 two times out of three. The Court reaches this result - even in a capital case because, it is said, the testimony is subject to cross-examination and impeachment. In the present state of psychiatric knowledge, this is too much for me. One may accept this in a routine lawsuit for money damages, but when a person's life is at stake - no matter how heinous his offense - a requirement of greater reliability should prevail. In a capital case, the specious testimony of a psychiatrist, colored in the eyes of an impressionable jury by the inevitable untouchability of a medical specialist's words, equates with death itself. 51 Juries are almost solely influenced by psychiatric testimony when predicting dangerousness. Justice Blackmun cited many studies proving this to be the case. “The major danger of scientific evidence is its potential to mislead the jury; an aura of scientific infallibility may shroud the evidence and thus lead the jury to accept it without critical scrutiny”52 Surprisingly the American Psychiatric Association participating as amicus curiae53 urged the court “[t]he unreliability of psychiatric predictions of long-term future dangerousness is by now an established fact within the profession.” 54 And flatly stated they are wrong 2/3 times.55 Yet the majority of the Justices somehow deem this sufficient to be ‘beyond reasonable doubt of future dangerous behaviour’ as required by the Texas statute. An accuracy level of 33.3% (2/3) would not even be sufficient to establish dangerousness on a civil standard of proof. Worryingly the purported ‘experts’ in the case mislead the court by stating they could predict Barefoot’s future behaviour “within a reasonable psychiatric certainty” or to a “one hundred percent and absolute”
51
p916 of the judgement 52
Giannelli, The Admissibility of Novel Scientific Evidence: Frye v. United States, a Half-Century Later, 80 Colum. L. Rev. 1197, 1237 (1980) cited in Barefoot p926 53
Literally: ‘friend of the court’. In the US groups can often advice the court if the matter concerns them. 54
P920 of the judgement 55
ibid
- 16 -
0521280 certainty.56 As shown above, even with today’s technology we can only achieve 75% accuracy. If judges tend to accept psychiatrists’ recommendations about dangerousness with little regard for cross-examination or other testimony, 57 then juries with less expertise are bound to accept them. A study into jury predictions of dangerousness confirms this.58 Juries use three broad factors to predict dangerousness: prior criminal history, the instant offence and expert opinion59. The study felt expert opinion was most important in sentencing defendants to death.60 “[T]he data shows that the decision to give life versus death in Texas squarely rests on … future dangerousness.”61 It has to be remembered in predicting dangerousness the jury is predicting dangerous behaviour in prison not in society. More importantly the study found that those sentenced to death (so deemed dangerous by the jury) but later having their sentences commuted so they were in the general prison population were not more violently predatory or a disproportionate threat to inmates and staff.62 In fact 90% of death row releasees held trusty status63 in Texas’ prisons. It cannot be underscored enough how dangerous this is; peoples lives are being deprived because 56
p922 of the judgement 57
Cocozza & Steadman, Prediction in Psychiatry: An Example of Misplaced Confidence in Experts, 25 Soc. Probs. 265, 272-273 (1978) at 271 58
Marquart, Ekland-Olson, Sorenson, ‘Gazing Into The Crystal Ball: Can Jurors Accurately Predict Dangerousness In Capital Cases?’ Law Society Review Vol 23 No 3 1989 449-468 59
ibid p454 60
ibid p459 61
ibid p464 62
ibid p459-461 63
A term for reward for good behaviour. For example unarmed guards during exercise etc.
- 17 -
0521280 of it. This despite the fact the empirical base does not give capital sentencing the utilitarian goal it seeks. Even more indicative of the failure of our adversarial court system is the experts chosen to testify and the effect they have. Our litigation system simply cannot cope in its present form with predictions of dangerousness. Psychiatrics are not experts on the level that, say coroners are about causes of death or engineers about structural faults. As one commentator has put it “[s]imply because one is termed a ‘mental health professional’ does not mean that he or she understands the cognitive and behavioural processes that create dangerous behaviour64”. Since the court labels them as ‘expert’ witnesses they immediately gain the scientific aura of more ‘hard’ science like doctors or engineers which their backgrounds do not justify. As one judge stated the problem with expert testimony is, “the opinion is introduced by one whose title and education (not to mention designation as “expert”) gives him significant credibility in the eyes of the jury as one whose opinion comes with the imprimatur of scientific fact.65” One purported expert mentioned in hundreds of cases in Texas is Dr. James P. Grigson also known as ‘Dr. Death.’66 His testimony on dangerousness has been used to help convict one third of all Texas death row inmates.67 His predictions have to be seen to 64
Mark David Albertson, ‘Can Violence Be Predicted? Future Dangerousness: The Testimony of Experts in Capital Cases’, 3-WTR Crim. Just. 18, 19 (1989) p46 cited in La Fontaine, ‘A Dangerous Preoccupation With Future Danger: Why Expert Predictions Of Future Dangerousness In Capital Cases Are Unconstitutional’ Boston College Law Review 2002 Vol 44 No.1 65
Flores v. Johnson, 210 F.3d 456, 464 n.10 (5th Cir. 2000) 66
See Bennett v. Texas, 766 S.W.2d 227, 231 (1989) explaining why he got the nickname Dr. Death – because he successfully testified on many capital cases for the prosecution successfully 67
Marquart et all (1989) Op Cit. p457 and Texas Defender Service, A State of Denial: Texas Justice and the Death Penalty 26, available at http://www.texasdefender.org/study/chapter3.html
- 18 -
0521280 be believed, all are remarkably similar.68 He offers dichotomous predictions including “100%”69 certainty even “a one thousand percent chance” 70 of dangerousness. One judge on a Texas appellate court went far as to say: [W]hen Dr. Grigson speaks to a lay jury . . . about a person who he characterizes as a “severe” sociopath . . . the defendant should stop what he is doing and commence writing out his last will and testament—because he will in all probability soon be ordered by the trial judge to suffer a premature death. . . . When Dr. Grigson is shown to have testified in a capital murder case in which the defendant challenges his conviction and sentence of death, the captions of those cases in which relief was granted read almost like the following: “THIS IS ANOTHER DR. GRIGSON CASE.”[Sic] The number of these cases is so great that I will simply refer the reader to either Westlaw or Lexis and not cite them71 Dr. Grigson has been proven wrong in many cases.72 In 1995 he was expelled from the APA and from the Texas Society of Psychiatric Physicians. Both determined his binary absolute predictions violated ethics.73 Remarkably he still continues to be appointed as expert witness by Texas courts to predict future dangerousness on the capital statute. Other problems with the litigation process lead to dangerousness being dangerous. Marx’s famous quote of how formal equality leads to substantive inequality rings true here. As the Court noted in Barefoot cross-examination was sufficient to question the 68
See annex 69
Marquart Op Cit p458 70
Texas Defender Service Op Cit n65 at 30 71
Bennett v. Texas, 766 S.W.2d 227, 231 (1989) 72
The most notable being Randall Dale Adams who Dr. Grigson described as an incurable sociopath with no regard for life. The famous documentary The Thin Blue Line helped exonerate Adams and not once in his 12 years of confinement or since his release has he committed a crime cf Texas Defender Service 32, 33. Yet Dr. Grigson still asserts Adams will kill again. Paul C. Giannelli, “Junk Science”: The Criminal Cases, 84 J. Crim. L. & Criminology 105, 116 (1993) cited in La Fontaine Op Cit Also see: Marquart et all (1989) Op Cit p458 describing a defendant deemed extremely dangerous who later had his sentence commuted and never committed a violent act again. 73
Texas Defender Service Op Cit., at 29–30
- 19 -
0521280 probative value of adduced expert testimony. However, that is contrary to the empirical evidence. Jurors are more likely to rely on a psychiatrist who is 100% certain of his prediction than a psychiatrist who simply states such predictions are unreliable.74 The psychiatrist stating that predictions are unreliable cannot therefore adduce the opposite, “this defendant will not offend again”. They appear less certain75. The counter balance cross-examination provides is minimal then. Justice Blackmun again puts it best: Given a choice between an expert who says that he can predict with certainty that the defendant, whether confined in prison or free in society, will kill again, and an expert who says merely that no such prediction can be made, members of the jury charged by law with making the prediction surely will be tempted to opt for the expert who claims he can help them in performing their duty, and who predicts dire consequences if the defendant is not put to death Even more appalling is the dramatic appeals to jurors civic duty ruthless prosecution attorneys indulge in. One attorney stated, if you (the jury) acquit the defendant “upon your heads will lie the next man that's dead due to [the defendant's] hands.” 76 This is similar to problems of over-prediction in clinical methods – people do not want to be ‘responsible’ for further violence. Astonishingly the Court was aware of all these arguments in the Barefoot case. Justice Blackmun’s dissent featured literally dozens of references to studies and the APA’s brief also strongly advised the court psychiatrics cannot predict dangerousness and contained even more references. Yet the majority held that the litigation process would do justice with no citations to empirical evidence. That conclusion is contrary to the data that the court was fully aware of. 74
Mantell, A Modest Proposal to Dress the Emperor: Psychiatric and Psychological Opinion in the Courts, 4 Widener J. Pub. L. 53, 62–63 (1994) 65-66 cited in La Fontaine 75
APA’s Brief at 6, (No. 82–6080) in Barefoot 76
Fortenberry v. Texas, 4719 (1977) cited in Marquart Op Cit. p460
- 20 -
0521280 The actuarial and clinical methods set out in part one of the paper look as scientific as general relativity in comparison to the Texas case study. Even with more scientific objective methods the courts have a fundamental misunderstanding of simple probability theory: the base rate.
Problems with the base rate When there is a low base rate for violence, predictive methods that are highly valid will generate a large number of false positives. That is, incorrectly identify those as dangerous who are not dangerous. An arithmetic example is in order77. In a prison population of 1000, 15% (150) will commit another violent act if released, 85% (850) will not - a 15% base rate. Assume there is an actuarial procedure that is 90% accurate. This procedure will accurately label 135 violent persons (0.90x150). However it will falsely label 85 as violent (0.10x850). 220 are labelled violent in total, only 61% (135) will re-offend when released. 39% would not re-offend but would still require incarceration. However, as part one has shown the best we can achieve is in the magnitude of 75% accuracy. Using the same base rate (15%) out of 1000 people, the actuarial method would correctly label 11378 (0.75x150), however it would incorrectly label 22279 (0.25x887). 388 in total are labelled ‘dangerous’ yet only 113 will commit offences. The 77
Derrived from Trowbridge, ‘Age and Recidivism: How Accurate are Our Predictions’, Washington Criminal Defence, Nov 2004, Vol 18. No.4 78
Strictly speaking 112.5, but you cannot have half a person and you commonly round up in statistics.
79
221.75, rounded up
- 21 -
0521280 original 75% accuracy figure becomes 29.1%. It is not justified under even the weaker civil standard of proof, let alone the criminal standard to justify detaining 222 people undeservedly. Elder sex offender populations have very low base rates (5%)80. Yet Civil Commitment of sex offenders in the US requires dangerousness to be taken into account. Civil Commitment is where the offender has completed their sentence and are moved into a new location if deemed ‘dangerous’. The Washington Statute requires “evidence beyond reasonable doubt” that a sex offender is ‘likely’ to engage in future sexual violence.81 In Young the Washington Supreme Court interpreted this to mean “the likelihood of re-offence is extremely high.”82 The court had quoted Dr. Vernon Quinsey an eminent figure stating that the predictive accuracy of sexual recidivism “can realistically be expected to be in the 80% range83”. Court felt those falling within the procedure with 80% accuracy the “likelihood of re-offence is extremely high.”
84
The
Court had a fundamental misunderstanding, thinking 80% accuracy would result in 80% being correctly selected. They did not take the base rate into account. The Civil Commitment Program in California has a mean age of 50.9 years.85 The high mean age
80
Hanson, R.K. (2001). Age and sexual recidivism: a comparison of rapists and child molesters, Ministry of the Solicitor General of Canada, www.sgc.gc.ca/epub/corr/eage200101/eage200101.htm 4 percent of those over 60 recidivated over an average follow-up period of seven years cited in Trowbridge 81
Washington Rev.Code 71.09.020(1) 82
re Young (Wash 1993) 857 p. 2d989 cited in Trowbridge Op Cit 83
ibid 84
ibid 85
Imren, D.; Hughes, M.; Hennessy, M.; Mitter, T.; & Russell, K. (March 17, 2004). “Treatment of the Sexually Violent Predator: Where are we?” presented at the annual conference at Asilomar Campgrounds in Pacific Grove, California, of the Forensic Mental Health Association of California. Cited in Trowbridge.
- 22 -
0521280 suggests many false positives since older sex offender populations have a 5% base rate. Hundreds, perhaps thousands are detained on very weak grounds. New Jersey’s method of Civil Commitment is worse. Far from taking actuarial methods into account virtually anything is admissible to predict dangerousness; including, Freudian psychoanalysis of foot fetishes to bumping into prison female guards by accident86. These two situations actually resulted in confinement.87
Conclusion Dangerousness is dangerous because currently it is used wrongly, at least by the US courts. Part one examined the fundamental problems with predictions. In reality these problems rarely even surface because predictions of dangerousness in practice, far from using quasi-scientific methods, use completely subjective clinical evaluations. ‘Dr. Death’ and others like him distort how science can help us predict dangerousness and their testimony has extreme prejudicial effect so their probative value is minimal. In capital punishment states this results in the worrying and somewhat sad case of people dying. We
86
New York Times Report by Mansnerus on November 17, 2003 ‘Questions Rise Over Imprisoning Sex Offenders Past Their Terms’ available online http://query.nytimes.com/gst/fullpage.html?res=9C07E7D71138F934A25752C1A9659C8B63&sec=he alth&pagewanted=print Username/password access required. On detail with the author. 87
Ibid, the author notes it is impossible to determine what weight the judge gave to the evidence but psychiatrists testified having a foot fetish as a poor prognostic sign and bumping into a female guard signified. Quoted from article “Mr. Deavers's public defender, Joan Van Pelt, said the incident had been an accident and not sexually motivated, and introduced the results of a polygraph test that backed him up. But the state's psychiatrist was skeptical. ''That doesn't mean he didn't have those thoughts,'' said the psychiatrist, Dr. Charles Gnassi. ''Brushing past a woman, a man -- it's difficult, I would think, not to have some type of sexual thinking.''
- 23 -
0521280 can use dangerousness to help our legal system but its extreme prejudicial effect has to be stemmed. The current due process provisions, at least in the US cannot perform this function adequately. Admissibly of evidence rules need updating, judges and defence lawyers need educating and perhaps juries need to be abolished in capital cases. 88 In short the shift of power from the judiciary to extra-judicial elements which Foucault was so wary of needs to be stemmed – simply by education and updating court procedure. Dangerousness is propagating throughout all of laws empire - the empire needs to strike back.
88
Of course in the US trial by ones peers is enshrined in the constitution so would be hard to reform
- 24 -
0521280
Bibliography Note: All US Cases are available from Findlaw (caselaw.lp.findlaw.com) Buchannan ‘Risk and Dangerousness’ Psychological Medicine, 1999, 29, 465-473 Burchell, G., Gordon, C & Miller, P. (Eds.), The Foucault Effect. Studies in Governmentality ( Harvester Wheatseaf : 1991) Cocozza & Steadman, ‘Prediction in Psychiatry: An Example of Misplaced Confidence in Experts’, 25 Soc. Probs. 265, 272-273 (1978) at 271 Corbett and Westwood ‘Dangerous and severe personality disorder’: A psychiatric manifestation of the risk society Critical Public Health, June 2005; 15(2): 121–133 Dolan and Doyle, ‘Violence risk prediction’, The British Journal of Psychiatry (2000) 177: 303-311 Greene, The Elegant Universe (Vintage Books: 2003) p85-86 for an excellent, easy to understand, explanation. Günter Blau and Hans-Dieter Schwind, Festschrift für Günter Blau zum 70. Geburtstag am 18. Dezember 1985, (1985: Walter de Gruyter) Hare, R.D. ‘The Hare Psychopathy Checklist-Revised’, (Multihealth Systems:1991) Hudson and Bramhall, ‘Construstion and ‘asianness’ in risk assessments by probation officers’ (2005) 45 British Journal of Criminology 721 Hume, (1739). A Treatise of Human Nature. Volume One. Published as A Treatise on Human Nature. Volume One (ed. T.Green and T. Grose), 1874. Longmans: Londo La Fontaine, ‘A Dangerous Preoccupation With Future Danger: Why Expert Predictions Of Future Dangerousness In Capital Cases Are Unconstitutional’ Boston College Law Review 2002 Vol 44 No.1 Marquart, Ekland-Olson, Sorenson, ‘Gazing Into The Crystal Ball: Can Jurors Accurately Predict Dangerousness In Capital Cases?’ Law Society Review Vol 23 No 3 1989 449-468 Megargee, E.I. (1976). "The prediction of dangerous behavior". Criminal Justice & Behavior, 3, 3-21. Cited in Cohen, D.Notes on the Clinical Assessment of Dangerousness in Offender Populations Psychiatry Online [www.priory.com/psych/assessin.htm]
- 25 -
0521280 Monahan, J. ‘The prediction of violent behaviour: Toward a second generation of theory and policy’. American Journal of Psychiatry, 141, 10-15 Murray Rothbard, Making Economic Sense, available online as an eBook: http://www.mises.org/econsense/ch5.asp New York Times Report by Mansnerus on November 17, 2003 ‘Questions Rise Over Imprisoning Sex Offenders Past Their Terms’ available online http://query.nytimes.com/gst/fullpage.html?res=9C07E7D71138F934A25752C1A9659C 8B63&sec=health&pagewanted=print Username/password access required. On detail with the author. Nola, Rescuing Reason: A Critique of Anti-Rationalist Views of Science and Knowledge, (Springer:2003) Oliver Wendell Holmes, ‘The Path of Law’ 10 Harvard Law Review 457 (1897) Quinsey, V.L. & Ambtman,R. ‘Variables affecting psychiatrists’ and teachers’ assessments of the dangerousness of mentally ill offenders’. 47 Journal of Consulting and Clinical Psychology Seto, ‘Is More Better? Combining Actuarial Risk Scales to Predict Recidivism Among Adult Sex Offenders’ Psychological Assessment 2005 Vol 17 No2 156-167 Sorensen, ‘Criminology: An Actuarial Risk Assessment of Violence Posed By Capital Murder Defendants’, 90 Journal of. Criminal. Law. & Criminology 1251, 1252 (2000) Steadman, H. J. & Cocozza, J. J. ‘Careers of the Criminally Insane: Excessive Social Control of Deviance’. Lexington, MA: Lexington Books Trowbridge, ‘Age and Recidivism: How Accurate are Our Predictions’, Washington Criminal Defence, Nov 2004, Vol 18. No.4
- 26 -
0521280
- 27 -