Predicting the Dangerousness of a Criminal
How Dangerousness is Predicted
There are two broad ways in which predictions can be made. The first is in a controlled environment (through clinical analysis), and through the analysis of statistics based on prior criminals and crime types (also referred to as actuarial or statistical prediction). The latter type looks at past offenders and what is generally known about their offences. It will also attempt to determine from this an offender's suitability for release.
Both methods have their benefits, and understandably, both have potentially negative consequences for the community into which the individual may be released and for the individuals themselves. For example, if an assessment is being made of a violent offender in a population whose level of violence is low, we may see an unacceptable level of false positives [Adair].
The ability to label someone as dangerous may be expressed most easily by comparing the conclusion of the clinical or statistical assessment to the actual behaviour of the individual (admittedly, this may not be known in some circumstances). The outcome is usually assessed as a false positive, false negative, true positive or true negative. A false positive incorrectly identifies someone as a future risk when in reality they are not. A false negative is when an individual is determined not to act in future dangerous ways, but they, in fact, do. A true positive is a correct assessment that the individual will be dangerous, while a true negative is a correct prediction that the person will not be dangerous. While it may sound confusing at first, it is a rather easy concept to understand, and perhaps best understood in the following way:
Conversely, the accuracy of assessing the behaviour of violent individuals among a general population of other violent individuals (for instance, in a prison population) is increased greatly [Adair], though in this situation the need for assessment tools is questionable, though not dismissed entirely.
The available assessment tools fall into two broad categories: clinical and statistical. Clinical tools use a controlled setting during which an assessment of the person in question occurs. This may also be performed without any personal interaction between the professional and the offender. That is, with the clinician having access to full accounts of the individuals behaviour, which may also include any psychological tests performed, rather than access to the individuals themselves. Of the variety of different assessment tools available, perhaps the most subjective are clinical judgements. These may be heavily influenced by the bias of the clinicians, their exposure to a particular type of case (or lack of exposure), and the degree of training they have in certain areas. This type of assessment is made based on experience with a certain clientele. For example, a psychologist may propose that "based on my experience with these offenders, I would suggest that this person will re-offend within the next 12 to 18 months". It should be apparent that there might also be some overlap in the methods, with clinicians sometimes also relying on statistical probabilities to assist them in their determinations.
Statistical tools are the quantified, impartial, and systematic use of factors from the file or history of the person [Adair]. This notion is extended to the knowledge of offender and offence types previously mentioned. A statistical prediction might operate on the following assumption: 75% of offenders released from prison who were originally incarcerated for offences of violence, who have no family or peer support network and no stable employment upon release, will go on to commit further crimes and again be placed in custody.
Another form of statistical prediction is the use of typologies. A typology is a clustering together of individuals by virtue of shared characteristics. These shared characteristics often include factors such as the motive for offences committed by the group, the behavioural precursors to offending and different probabilities for repeat offending [Blanchette]. Typologies are applied by assessing the crime scene or behaviour of the offender during the offence, and then by comparing this to the information supplied in the typology. This information is then reapplied to the case under examination and a determination of the offender's behaviour is then made. Typologies often present a problem when used in this fashion because they generalise the behaviour of one offender based on past similar offenders. The main reason that this is a problem is that different offenders will do similar things, but for different reasons. It is often argued that these typologies should be used as a starting point only, and not for providing detailed offender information.
Other more specific methods are available for predicting the future dangerousness of an individual depending on their criminal history. For example Turvey summarises the work of Quinsey and others when examining the issue of sex offender recidivism. Quinsey has noted that a suitable predictor of re-offending is inappropriate age selection. This refers to offenders who choose victims whose ages are not concordant with their own, as is the case with many child molesters. The sexual arousal of this group of offenders is measured phalometrically, where a machine known as a penile plethysmograph measured genital arousal in response to a number of sexual stimuli. This type of physiological assessment uses various standardised stimuli (such as video, audio and photograph) to determine age and gender preference, as well as interest in sexual violence relative to consensual sexual interactions [Blanchette].
Turvey notes the following findings of Quinsey in relation to predicting future dangerousness:
- Offenders who exhibit a preference for sadism or violence during rape.
- The degree of psychopathy of an offender (psychopathy is a personality disorder, not a mental illness as often portrayed by the media and other sources).
- Inappropriate sexual age preferences as phalometrically measured.
Behavioural observations may also be used in determining future dangerousness. These predictions are based on how a particular person acted in the past in a similar situation [Bartol], which may also assist in determining appropriate treatment options for particular offenders if it is known that the offending behaviour came about as the result of a particular situation. For example, it might be known that a non-preferential child sex offender committed his offence during a period of intoxication. This may indicate that some form of drug or alcohol treatment is warranted to reduce the likelihood of re-offending.
A study conducted by Shaffer, Waters and Adams determined the accuracy of predictions in a sample of a number of subjects. This study found that there was a 90% correct classification of non-violent offenders, and an 80% correct classification of violent offenders with a 10% rate of false positives (incorrectly identifying someone as dangerous when they are not) in a hospital sample. In the prison sample, correct classification occurred 80% of the time with non-violent subjects and 76% correct -identification of violent subjects with 20% false positives. The following criteria were used selectively depending on the group under examination (non-violent or violent):
White or non-white
Married or single
Stable or unstable
Receiving welfare during the developmental period
Non-intact or intact
Juvenile arrest history
Having no arrests or one or more arrests
Having no arrests or one or more arrests
History or no history
Score on standardised tests
Score on the standardised Minnesota Multiphasic Personality Inventory (a psychological test)
Level of arousal
Based on psychiatric mental status reports
History of violence
History or no history of violence during the past five years