Skip to main content
SearchLoginLogin or Signup

Geeks and Newbies: Investigating the Criminal Expertise of Online Sex Offenders

Published onApr 14, 2022
Geeks and Newbies: Investigating the Criminal Expertise of Online Sex Offenders
·

Abstract

While online sex offenders use a wide range of strategies to try to avoid police detection, attempts to avoid detection of child sexual exploitation materials (CSEM) and online sexual solicitation of children have received very little attention. This study aims to understand online sex offenders’ behaviors by modeling the factors associated with their use of technological data protection and anonymity preservation strategies. The data is based on a sample of 199 men involved in crimes related to the use of child pornography or sexual solicitation of minors online. The analytical strategy based on the use of an artificial neural network (ANN), a machine-learning system, identified two trends. First, those who displayed problematic substance use and sexual thoughts and fantasies as well as behaviors reported to be preoccupying did not use specific strategies to avoid police detection. Second, two combinations of factors predict use of police anti-detection strategy, suggesting that the criminal expertise of online sex offenders is manifested in two different patterns: those building on existing knowledge, and those learning skills through previous judicial experience.

Keywords

Online sex offenders, Criminal Expertise, Artificial Neural Network, Individual Characteristics, Cognitive Distortions

Introduction

Expertise refers to superior skills in a specific area (Bourke, Ward and Rose 2012). From a general perspective, expertise is the result of behavioral and cognitive processes that allow individuals to distinguish themselves from others, particularly neophytes, based on their knowledge, skills, and experience (Nee and Ward 2015). Contrary to hypotheses that suggest that committing crimes does not require any particular skill (Hirschi, 1986), empirical research has consistently shown that some criminals display higher levels of expertise than others (i.e., Meenaghan et al. 2020; Nee 2015; Nee and Ward 2015; Ó Ciardha 2015; Topalli 2005; Ward 1999) and that the expertise of sex offenders, like the expertise of other offenders, can be represented as a continuum (Nee and Meenaghan 2006; Nee et al. 2015; Topalli, Jacques and Wright 2015). Research in this area has focused particularly on off-line sexual offenses (e.g., child sexual abuse, rape, sexual homicide) and has suggested that offenders’ ability to groom victims and attempt to avoid police detection is a fundamental component of the notion of expertise (e.g., Chopin, Paquette and Beauregard 2021; Ó Ciardha 2015; Ward 1999). The behavioral expression of this capacity is displayed in strategies that make it possible to hide traces of criminal acts and preserve anonymity (Beauregard & Bouchard, 2010). It has been shown that the expertise of sexual offenders is acquired through learning from both criminal and legal experiences (Chopin, Beauregard and Bitzer 2020; Lussier, Bouchard and Beauregard 2011; Lussier and Mathesius 2012). Criminal expertise, framed by the rational choice perspective (Cornish and Clarke 1986; Cornish and Clarke 1987), is specifically grounded in the restrictive deterrence approach (Gibbs 1975; Jacobs 1996b; Jacobs and Miller 1998). This approach describes how individuals deal with the threat of criminal sanctions by attempting to avoid detection (Jacobs and Cherbonneau 2014). Restrictive deterrence is operationalized through four dimensions, one of which concerns the use of strategies to limit the probability of being detected (Jacobs 2010).

Research on the expertise of online sex offenders is limited and the analyses that exist have focused primarily on the technological tools they used (Authors in press a; Fortin and Corriveau 2015; Fortin, Paquette and Dupont 2018; Steel et al. 2020), neglecting consideration of cognitive factors and other non-Internet related behaviors. Various hypotheses have been put forward to explain the variance in online sex offenders expertise without providing empirical answers (Balfe et al. 2015). Looking at how technology is used is central to understanding online sexual crime, particularly in identifying the specific ways in which this kind of crime differs from offline sexual offenses. However, to completely explain such offenses it is important to consider the behavior and cognitive processes of the individuals who commit them. This study examines avoiding detection behaviors used by online sex offenders whose victims are minors. Specifically, this research aims to understand how behavioral and cognitive factors affect the acquisition and use of technological expertise by online sex offenders targeting child victims.

The ability to avoid police detection: rational choice and criminal expertise

The use of techniques to avoid police detection is traditionally studied from a rational choice perspective (Cornish and Clarke 1986; Cornish and Clarke 2008) centered on the notion of criminal expertise (e.g., Cherbonneau and Copes 2005; Copes and Cherbonneau 2006; Nee 2015).

Perpetrators follow a crime commission process (see Beauregard, Leclerc and Lussier 2012; Cale et al. 2021), defined as the set of decisions and actions undertaken to successfully complete a crime (Douglas et al. 2006). The rational choice perspective provides a theoretical framework for decisions made during this process, suggesting that individuals conduct an analysis of the costs and benefits associated with their behaviors and commit a crime only if the gains are believed to outweigh the risks (e.g., arrest, fine) (Cornish and Clarke 1986; Cornish and Clarke 1987). Following this perspective, a successful crime is one that 1) achieved the intended goal and 2) avoided police detection (Chopin, Paquette and Beauregard 2021; Copes and Cherbonneau 2006; Gallupe, Bouchard and Caulkins 2011). Studies of sexual offenders’ behaviors have shown that they involve a heterogeneous combination of planning and avoidance strategies indicative of a wide range of criminal expertise (Reale, Beauregard and Chopin 2021a; Reale, Beauregard and Chopin 2021b; Reale, Beauregard and Chopin 2021c; Reale, Beauregard and Chopin 2021d). While studies have identified a continuum of risk related to the profiles of online sex offenders (Beech et al. 2008; Briggs, Simon and Simonsen 2011; D’Ovidio et al. 2009; Sheehan and Sullivan 2010), there seems to be a continuum of technical expertise as well. Studies of offline sexual offenses have failed to find a connection between offender expertise and crime resolution (Chopin et al. 2019; Reale, Beauregard and Chopin 2021c). Findings indicated a lack of correlation between the criminal expertise sophistication and the outcome of the investigative process (Chopin, Paquette and Beauregard 2021). The strategies used were not effective in avoiding police detection, which depended instead on the quality of police work or even on luck (see Rossmo 2009). The restrictive deterrence literature offers an additional specific framework by connecting the criminal threats against individuals with their behavioral adaptations to limit the probability of being identified by law enforcement agencies (Gibbs 1975; Jacobs 2010). Previous studies found that the risk of being arrested impacted the behavior of individuals involved in offline (Jacobs 1996a; Jacobs 1996b; Jacobs and Cherbonneau 2018; Jacobs and Cherbonneau 2014) and online crimes (Maimon 2020; Maimon, Howell and Burruss 2021; Wilson et al. 2015). However, to our knowledge, no empirical study has analyzed the variations of these avoidance behaviors for online sex crimes.

Strategies Used by Cybersex Offenders

Cybersex offenders whose victims are minors solicit children and adolescents for sexual purposes on the Internet, in social networks, or using digital applications and also participate in the online market for child sexual exploitation materials (CSEM), both producing and exchanging images. Although in many ways they are similar to individuals who sexually abuse children offline, they also demonstrate unique psychological characteristics and behaviors (Babchishin, Hanson and VanZuylen 2015).

Cybersex offenders use a wide range of strategies to avoid police detection (see Steel et al. 2020 for a comprehensive review of the literature). These strategies are aimed at protecting both digital data, including illegal content (e.g., CSEM), as well as identity and their use reveals expertise associated with a more developed perception of risk (Authors in press a; Balfe et al. 2015; Fortin, Paquette and Dupont 2018; Fortin and Roy 2006). Storing data in the cloud and using digital deletion software and encryption can protect digital data and eliminate evidence that a criminal offense has been committed. However, given the relatively small number and conflicting results of studies that have examined the use of these techniques by individuals involved in online offending, it is difficult to estimate how prevalent they are: studies conducted in a wide variety of contexts and at different times have found that between 3% and 80% of online sex offenders use digital data protection techniques (Authors in press a; Seto, Reeves and Jung 2010; Steel et al. 2020; Wolak, Finkelhor and Mitchell 2011).

The literature also reports identity protection techniques that involve masking or hijacking Internet Protocol (IP) addresses through VPNs, proxies, or the Dark Web (e.g., via the Tor network) (Authors in press a). The prevalence of use of identity protection techniques is also difficult to estimate. Dalins, Wilson and Carman (2018) found that 1.80% of Dark Web sites were dedicated to content related to child sexual exploitation, while Authors (in press) found that 4.35% of online sex offenders in their sample had used virtual identity protection strategies.

Factors Associated with the Use of Digital Strategies to Avoid Police Detection

Various factors associated with variability in risk perception and ability to use protective strategies have been reported for online sex offenders (see Balfe et al., 2015 for a comprehensive literature review). Several studies have proposed offender age as an explanatory factor in differing levels of use of these strategies and it has been suggested that younger offenders may be less aware of the risks associated with committing crimes online than older offenders and consequently take fewer precautions to avoid police detection (Balfe et al. 2015; Kierkegaard 2011; Quayle and Taylor 2011; Zhang 2010). A similar link has been proposed for socio-economic status and education, suggesting that the higher the socio-economic status and education level, the less online sex offenders are aware of the risks and, consequently, the fewer precautions they take to protect themselves (Wolak et al. 2010). Socio-demographic factors could act as intermediaries in offenders’ criminal experience (Balfe et al., 2015) and thus, as with offline sexual offenders, influence the extent to which protective strategies are used (Davies 1992; Davies, Wittebrood and Jackson 1997). Other individual factors, including cognitive, sexual, and emotional characteristics, may also explain why some online sex offenders use strategies to avoid police detection (Balfe et al. 2015). For example, individuals with emotional regulation difficulties, particularly those experiencing intense sexual arousal, may pay less attention to how their offending is carried out, leading them to use fewer protective strategies (Elliott and Beech 2009; Prichard, Watters and Spiranovic 2011). Balfe et al. (2015) suggest that substance use or mental disorder may be associated with the absence of strategies. In contrast, belonging to a pedophile network whose values and attitudes are favorable to the sexualization of children or expertise with computers may influence perception of risk among online sex offenders, encouraging them to use protection strategies (Eneman 2009; Holt, Blevins and Burkert 2010).

Purpose of the Study

A review of the literature reveals that online sex offenders expertise in strategies to avoid police detection has received little attention. Studies in this area tend to focus on hypotheses about methods of online offending and descriptive statistics rather than looking at factors that explain the continuum of expertise among online sexual offenders. In the face of this empirical void, the present study contributes to understanding the behavior of online sex offenders by modeling the factors associated with use of data protection and anonymity preservation strategies. The lack of studies examining which factors could impact the use of avoiding detection strategies by online sex offenders renders this study exploratory by nature. Such an approach has been only examined for offline sex offending and it is difficult to formulate hypotheses. Consequently, the current research aims to identifying factors associated with the use of detection avoidance strategies among online sex offenders targeting minor victims.

Methodology

Data

The data used in this research was collected as part of the Presel project.

The goal of this research program, conducted in collaboration with the Sûreté du Québec, was to better understand online sexual crimes against children and adolescents. In the project, police case files of men convicted for online sexual charges were analyzed, including information taken from official sources (e.g., criminal records), police and expert reports (e.g., forensic computer analysis), public Court judgments, and videotaped police interviews with the suspect. The data was collected after the end of judicial procedures and included information on the characteristics of 199 men charged with and found guilty of crimes related to the use of child pornography or to the sexual solicitation of minors online in cases that occurred between 2001 and 2020 in Québec. As such, investigators were conducting interviews as they usually do without being aware of this study. Although all offenders have been identified by the police, the current study focuses on the factors associated with the factors associated with the acquisition and the use of detection avoidance strategies used by online sex offenders whose victims are minors, and not their impact on the investigative process.

A coding manual was developed by the second author, who also collected and compiled the data. To ensure adequate conceptual and content validity (Nunnally and Bernstein 1994), conceptualization and operationalization of the variables were reviewed both by the third author and by an external expert in the field of forensic psychology. Variables were selected and defined based on a wide range of scientific literature on the subject and examples were added to guide the coding process. Research assistants were then trained to use the coding manual. The information in the database relates to characteristics of individuals involved in online sexual crimes against children and adolescents (i.e., sociodemographic factors, criminal history, childhood and adolescent victimization, components of the crime – sexual, relational, emotional, antisocial and cognitive), as well as characteristics of the crime and of victims. As police interviews were part of the data and took 4 to 5 hours on average (up to 10 hours in some cases), a subsample (n = 9 cases) was coded by two research assistants for inter-rater reliability. Overall agreement on coding reached 84%.

Individuals in the sample had been arrested for various offenses: 33.67% (n=67) had been involved in child pornography offenses, 16.58% (n=33) in online sexual solicitation of minors (referred to as “luring” in the Canadian Criminal Code), 35.17% (n=70) in both online and offline sex offenses, 7.03% (n=14) in online and offline sexual solicitation offenses, and 9.05% (n=18) in child pornography and online and offline sexual solicitation offenses.

Measures

The variables used were selected based on previous studies of individuals involved in online (Balfe et al. 2015; Kierkegaard 2011; Quayle and Taylor 2011; Wolak, Finkelhor and Mitchell 2011), or offline offending (e.g., Beauregard and Bouchard 2010). Variables to measure an individual’s ability to use strategies to protect identity and avoid police detection were added.

Dependent variable. A dichotomous variable was used to distinguish between individuals who did not use techniques to avoid detection and those who did. These variables were measured using police and forensic reports. This variable (0 = absence; 1 = presence) is a composite of the variables: 1) encryption of digital data (n = 33; 18.59%) and, 2) use of proxies/VPN/Dark Web (n = 11; 5.53%). In our sample 16.58% (n = 33) of individuals used only data encryption techniques, while 3.51% (n = 7) of individuals used proxies/VPN/Dark Web and 2.01% (n = 4) used both techniques. In total, 22.11% (n = 44) of our sample used at least one technology. Because technologies evolve rapidly over time, we tested the impact of such evolution on the use of digital technologies by comparing contemporary use (i.e., 2011-2020) to earlier use (i.e., 2001-2010). The analyses indicated that there was no significant difference in the overall use of identity protection technologies (χ2 = 0.042; p = 0.838) during the total period being considered, nor was there a specific difference in the use of encryption (χ 2 = 0.788; p = 0.375) or proxies/VPN/Dark Web (χ2 = 0.081; p = 0.776), thus suggesting no influence of time on the use of digital strategies employed by the online sex offenders in our sample. Additionally, we tested the association between the continuous temporal distribution and the use of strategies. Findings suggest that there was no significant difference in the overall use of identity protection technologies (β = 0.000; p = 0.951) during the total period being considered, nor was there a specific difference in the use of encryption (β = 0.000; p = 0.997) or proxies/VPN/Dark Web (β = 0.000; p = 0.953).

Independent variables. Thirteen independent variables were selected to identify characteristics associated with the use of digital protection techniques. These variables were divided into different blocks: socio-demographic characteristics, crime characteristics, criminal career characteristics, and individual and psychological characteristics. These variables were measured using official sources, police and expert reports, public Court judgments, and police interviews.

Socio-demographic characteristics. Some studies have suggested that demographic characteristics such as age and social integration influence the perception of risk among online sex offenders and that older individuals are better at using protective strategies (Balfe et al. 2015; Kierkegaard 2011; Quayle and Taylor 2011; Wolak, Finkelhor and Mitchell 2011). We used two variables associated with these ideas, one continuous, and one dichotomous: 1) age at the time of the crime ( = 37.76, SD = 14.01, range = 15-81) and 2) being employed (i.e., as opposed to being unemployed, retired, or student).

Characteristics of the crime. Studies on the ability of sexual offenders to use strategies to attempt to avoid detection have highlighted the importance of the nature of the various acts involved (see Beauregard and Bouchard 2010; Chopin, Beauregard and Bitzer 2020). We hypothesized that individuals involved in child pornography would be more skilled in using technologies to protect identity than those who use chat sessions for virtual contact with adolescents. While consulting sites dedicated to child pornography is illegal, browsing and chatting on social networks and in discussion forums is not necessarily illegal, particularly if the communications do not encourage or involve sexual behavior involving a minor or are with adults. We used a dichotomous variable to capture this characteristic: the offense involves child pornography (considered both exclusively or in conjunction with other sexual offenses).

Criminal career characteristics. Studies have shown that offline sexual offenders who have a criminal history are more likely to use strategies to attempt to avoid detection than those who have never been involved with the justice system (Davies 1992; Davies, Wittebrood and Jackson 1997). Although this effect of criminal history has never been tested with online sex offenders, we hypothesized that experience with the justice system would increase perception of the risks associated with offending, leading to increased use of strategies to avoid police detection. We used two dichotomous variables – history of non-sexual online crimes (e.g., offense, robbery, sexual offense) and history of online sexual crimes (i.e., child pornography, luring) – to capture this.

Individual and psychological characteristics. Several studies have suggested that individual and psychological factors may influence perception of risk among individuals involved in online sex crimes, particularly substance abuse, mental disorders, paraphilias, and sexual arousal (Elliott and Beech 2009; Prichard, Watters and Spiranovic 2011). The following variables were used to test the effect of these characteristics on the use of digital protective strategies: problematic substance use (drug or alcohol use self-reported by the offender as problematic, related to the commission of a criminal offense, or having been the basis for referral to a therapeutic support group); tendency to social isolation (feeling isolated or experiencing weariness, emptiness, or melancholy); pedohebephilic sexual interest (clinical diagnosis of pedophilia or hebephilia or admission of specific sexual interest in children or adolescents); and sexual preoccupation (self-reports of intrusive, preoccupying, and obsessive sexual thoughts, fantasies, or behaviors). As expertise involves cognitive processes and maladaptive conditions may affect cognitive ability, we included variables related to maladaptive cognitions, hypothesizing that maintaining maladaptive beliefs may affect the ability to perceive risk (see Dror 2011 on the notion of cognitively influenced tunnel vision). We used the following variables: cognitions supportive of the sexualization of children (self-reported statements suggesting that children, like adults, can have intimate emotional and sexual relationships; that they are able to provide free and informed consent; and that they can desire and seek to actively engage in sexual activities); cognitions suggesting that sexual offenses are uncontrollable (i.e., the individual threw away the blame for the offenses); cognitions indicating that the world is dangerous (i.e., that adults are not trustworthy and can be expected to reject, manipulate, or betrayed); cognitions suggesting that the virtual world is not real (i.e., content obtained from the Internet does not reflect reality, Internet users are lying about their intentions and identity, everything found online is a lie).

Analytical Strategy

The analytical strategy involved an artificial neural network (ANN), a machine-learning system used in data mining and computer science. One of the main goals of neural network analysis is to identify complex patterns and relationships between several inputs (explanatory factors) that cannot be identified by the human brain (Bigi et al. 2005). This technique is used in a wide range of scientific disciplines (e.g., medicine, neuroscience, computer science) and shows particular promise in the social sciences, given the complexity of human behavior (Liu et al. 2011) (Liu et al., 2011). ANN has several advantages over prediction methods such as logistic regression or classification and regression trees. Logistic regression is known for its robustness but does not handle the possibility that different variables predict the same outcome for different subgroups of individuals (see Steadman et al. 2000). Classification and regression trees are better at identifying the possibility that the same outcome can be predicted for different groups but demonstrate a lack of stability and a risk of overfitting (Colombet et al. 2000; Dillard et al. 2007). Finally, ANN has the advantage of being stable with small samples (Cui et al. 2004; Kim 2008), unlike regression models which can handle such samples only in specific conditions. ANN models avoid the weaknesses of conventional methods and are particularly effective when the main objective is to predict and identify significant interactions or complex nonlinearity between variables for a given set of data (Liu et al. 2011; Tu 1996).

The multilayer ANN perceptron algorithm is the most commonly used network architecture. It uses inputs (independent variables), hidden layers (nodes), and output layers (dependent variables). These different layers are connected and the strength of the association is identified by synaptic weights – the closer they are to zero, the weaker the relationship; the further they are from zero, the stronger the relationship. Negative values suggest the opposite relationship.

To test the quality of the predictive model, the multilayer ANN perceptron algorithm includes a learning process and a testing process (Price et al. 2000). The percentage of incorrect predictions as well as the value for the area under the curve (AUC) are used to assess the quality of the model (Liu et al. 2011). It is recommended to use a sample with at least 50 statistical units (Cui et al. 2004), 10 times higher that the number of variables included in the model (Abu-Mostafa 1993; Haykin 2009), while ordinal, dichotomous and continuous variables can be used (Moayed and Shell 2011).

In this study, the input layer of the ANN model consisted of the independent variables and the output layer contained two units, representing the two categories (no technology use = 0; technology use = 1) of the dependent variable. To create the neural network model, 71.35% of the cases (n = 142) were randomly sampled while 28.65% of the cases (n = 57) were used to test the model.

A multicollinearity test was performed for the independent variables and the results show that the static variance inflation factor (VIF) did not exceed the threshold of 1.348 and the tolerance was greater than 0.742 (see Appendix 1).

Ethics

This study received ethical approval from the Arts and Sciences Research Ethics Board of the University of Montreal.

Results

Univariate analyses

Table 1 provides a description of the variables used in this research. The results indicate that 22.11% of the individuals in our sample used specific technologies to try to avoid police detection. In 18.59% of the cases, online sex offenders encrypted their digital data and in 5.53% of the cases used a proxy, a VPN, or the Dark Web. The average age of the individuals in the sample was 37.76 years at the time of collection and more than half (55.28%) were employed. Three quarters (76.38%) had been involved in a child pornography-related offense. Less than half (40.70%) had a history of non-sexual crimes online while 14.07% had a history of sexual crimes online. The analysis of individual characteristics shows that 20.60% of the individuals had a substance abuse problem, 35.68% had a tendency toward social isolation, and 42.21% had a pedohebephilic sexual interest. In 27.14% of cases, individuals reported a sexual preoccupation. Finally, analysis of their cognitions shows that 43.22% of the online sex offenders had expressed the belief that children are sexual, 53.77% that sexual offenses are uncontrollable and 14.07% that the world is dangerous. Finally, more than a third (36.18%) had indicated they believed that the virtual world did not correspond to reality.

Table 1. Descriptive analyses (N=199)

Variables

N

%

Use of technological strategies to protect against police detection

44

22.11

Digital data encryption

37

18.59

Use of proxies/VPN/Dark Web

11

5.53

Socio-demographic characteristics



Age (continuous)

37.76a - 14.01b [rank=15-81]

Employed

110

55.28

Characteristics of the offence



Involves child pornography

152

76.38

Criminal Career Features



History of online non-sexual crime

81

40.70

History of online sex crimes

28

14.07

Individual and psychological characteristics



Problematic substance use

41

20.60

Tendency to social isolation

71

35.68

Pedohebephilic sexual interest

84

42.21

Sexual preoccupation

54

27.14

Cognitions:

Sexualization of children


43.22

Sexual offences are uncontrollable

107

53.77

Offline world is dangerous

28

14.07

Virtual universe is not real

72

36.18

Notes.

a Average

b Standard deviation

Multi-layer perceptron ANN prediction of the use of technologies

Table 3 presents the results of the ANN model predicting the use of technologies to protect against police detection. In order to assess the goodness-of-fit of the model, we used the percentage of correct classification of the training and test samples as well as the AUC of the overall model. The results indicate that the model has a very good level of prediction, with an AUC of 0.839 (see Appendix 2). The model correctly classifies 88.30% of the cases used for learning and 87% of the cases used for testing. This model is based on five combinations of variables called nodes (hidden layer), three that predict the absence of use of technology and two that predict use of technology. In other words, three combinations of variables predict the lack of use of technologies while two combinations of variables predict their use by individuals involved in online sexual offending against minors. Nodes 2 and 4 represent the combinations of factors that predict use of technology while nodes 1, 3, and 5 represent the combinations of factors that predict that individuals will not use technology to protect their identity. Figure 1 shows the two nodes (2 and 4) that predict individuals’ use of technology. The importance of each variable in the different combinations is measured by the synaptic weight coefficients (i.e., the closer to zero, the lower is the relationship; the farther from zero, the stronger is the relationship. Negative coefficients indicate negative relationships). Node 2 suggests that individuals who are younger (-0.02), employed (0.60), have a pedohebephilic sexual interest (1.38), and report beliefs favorable to the idea that sexual offenses are uncontrollable (0.14) and that the virtual world is different from reality (1.19) are more likely to use technologies to avoid police detection. Node 4 suggests that individuals who are older (1.09), employed (0.50), have been involved in a child pornography offense (0.23), have a history of online sex crimes (0.79) or other crimes (0.30), have a pedohebephilic sexual interest (0.36), and report that the world is dangerous (0.46) are more likely to use technologies to avoid police detection.

Table 2. ANN of factors associated with technology use (N = 199)







Output layer

Input layer



Node 1

Node 2

Node 3

Node 4

Node 5

No use of technology

Use of technologies

Input layer

(Constant)

-0.57

-1.72

0.06

-0.87

0.16




Socio-demographic characteristics









Age (continuous)

-0.13

-0.02

-0.12

1.09

-0.16




Employed

-0.43

0.60

-0.23

0.50

0.01




Characteristics of the offence









Involves child pornography

-0.02

-0.82

-0.45

0.23

0.10




Criminal Career Features









Online history of non-sexual crime

-0.08

-1.48

0.25

0.30

-0.52




Online history of sex crimed

-0.04

-0.02

-0.24

0.79

-0.09




Individual and psychological characteristics









Problematic substance use

0.52

-0.06

0.27

-0.80

0.21




Tendency to social isolation

-0.17

-0.86

0.69

-1.07

-0.39




Pedohebephilic sexual interest

-0.34

1.38

0.65

0.36

-0.31




Sexual preoccupation

0.05

-1.32

0.13

-0.07

0.20




Cognitions

Children have sexual interests

-0.26

-0.44

0.43

-0.23

0.19




Sexual offences are uncontrollable

-0.11

0.14

-0.69

-0.59

-0.41




Offline world is dangerous

0.11

-1.16

-0.45

0.46

-0.32




Virtual universe is not reality

0.20

1.19

0.66

-0.88

-0.13



Output layer

(Constant)






0.48

-0.23


Node 1






0.14

-0.69


Node 2






-1.52

1.51


Node 3






1.09

-0.34


Node 4






-1.03

0.94

 

Node 5






0.30

-0.03

Classification








% correct classification (training sample)

88.30







% correct classification (test sample)

87.00







Area under the curve (AUC)

0.84

 

 

 

 

 

 


Figure 1: ANN of factors predicting technology use by online sex offenders (N=199)

Discussion

The objective of this study was to determine why some online sex offenders are more likely to use strategies to try to avoid police detection. To do this, we looked at a sample of 199 individuals involved in online sex crimes against children and youth, including child pornography and sexual solicitation of minors.

The distribution of protective strategy use in our sample supports the idea that only a small proportion of individuals demonstrated criminal expertise (Authors, in press; Wolak et al., 2011). This distribution also confirms the existence of a continuum of criminal expertise, with a majority of individuals who are reluctant to use strategies and a minority who are committed to developing technical skill in an attempt to protect themselves from identification by police (Elliott and Beech 2009; Prichard, Watters and Spiranovic 2011). To identify more precisely the factors associated with these behavioral tendencies, we tested the role of several variables suggested by the literature, using an ANN model to identify the combinations of factors that best predict strategy use among online sex offenders.

A cross-sectional analysis identified characteristics related to the use of strategies to avoid police detection. First, problematic substance use consistently predicted non-use of strategies to avoid detection. This result supports the findings of Balfe et al. (2015) and is consistent with work on the effect of substance use on criminal behavior that shows that such use affects cognitive abilities, reducing the ability to rationally analyze costs and benefits associated with behavior (see Assaad and Exum 2002; Beauregard, Lussier and Proulx 2005; Exum 2006). Substance use alters decision-making mechanisms, resulting in altered perceptions of reality, reduced understanding of costs (i.e., police identification), and overestimation of the benefits of the crime (Assaad and Exum 2002; Peterson et al. 1990). Second, sexually aroused individuals whose sexual thoughts, fantasies, and behaviors were reported to be preoccupying, recurrent, and pervasive did not use specific strategies to avoid police detection. This finding supports the hypothesis that sexual arousal influences the behavior of online sex offenders (Prichard, Watters and Spiranovic 2011; Quayle and Taylor 2011). Previous work on the impact of sexual behavior on decision-making has established that sexual arousal can lead to an increase in risk-taking as well as a reduction in recognition of associated negative consequences (Ariely and Loewenstein 2006; Skakoon-Sparling, Cramer and Shuper 2016).

Dror (2011) hypothesized that individuals’ expertise may be influenced by the presence of certain cognitions. Our results support this hypothesis as specific cognitions were found to be predictive of the use of strategies to avoid police detection. The idea that sexual offenses are uncontrollable was associated with use of strategies while online offenders who believe that their online sexual behavior is controllable apparently do not feel the need to take action to protect themselves from police detection. The belief that the world is dangerous also predicted use of strategies. In light of these results, it appears that specific cognitions may influence the decision to protect oneself from police detection.

Another of our results appears counterintuitive: while it would seem logical to expect that those who believe that the online world is a reflection of reality would take steps to protect themselves from police detection, our results show the opposite. It should be noted, however, that the cognitions of online sex offenders were identified through their discourse during police questioning. It is therefore likely that these verbalizations were formulated in a utilitarian manner and do not reflect the real beliefs of these offenders. This hypothesis is consistent with the suggestion of many researchers that sex offenders’ verbalizations may be merely ad hoc justifications aimed at excusing and minimizing the seriousness of their offending actions or attempting to lessen the negative sanctions associated with their crimes (e.g., lighter sentences, maintenance of the marital relationship; e.g., Maruna and Mann 2006; Paquette and Fortin 2021). Different research designs, particularly those that use implicit measures of cognitions, could shed light on the role of beliefs held by online sex offenders in developing expertise in avoiding police detection. Finally, cognitions supportive of child sexualization did not predict strategy use by online sex offenders. As this cognition underlies pedophilic sexual interest (see Paquette and Fortin 2021), which is a strong predictor of strategy use in our model, it is possible that these two variables share the same variance.

Skills and learning: two patterns of factors associated with strategy use

The ANN model indicates that two combinations of factors predict use of police anti-detection strategies. These combinations suggest that the criminal expertise of online sex offenders manifests itself according to two different patterns: 1) building on existing knowledge and 2) using skills learned through previous judicial experiences.

Leveraging existing knowledge. This combination of factors suggests that the subgroup of expert online sex offenders is younger. While this result appears to conflict with study results showing that younger individuals are likely to feel there is less risk involved in committing online crimes and therefore might be less likely to attempt to avoid detection (Balfe et al. 2015; Kierkegaard 2011; Quayle and Taylor 2011), it may be that younger people are simply more comfortable with and more accomplished in using digital technology. Other predictors of this pattern include the cognition that sexual offenses are uncontrollable. Cyberoffenders who demonstrate this pattern appear to be new to crime, suggesting that some may have doubts about their ability to control their online behavior, leading them to adopt strategies to avoid police detection. The results show that pedohebephilic sexual interest is the best predictor of this pattern. Holt, Blevins and Burkert (2010) showed that members of virtual pedophile networks, who are likely to join these networks because of specific sexual interests, are more aware of the risks involved in committing online sexual offenses and thus more likely to take steps to mitigate these risks. Having a job is also predictive of using strategies to avoid police detection. Although this information remains relatively general in our study, it could be related to

Eneman’s (2009) suggestion that individuals who use computers as part of their job would be particularly sensitive to the risks involved in committing online crimes and would also have a higher level of expertise. The combination of these factors suggests that some of the individuals who used strategies to avoid detection were more aware of the risks involved and used their existing skills to try to mitigate these risks.

Using skills learned from previous court experiences. This specific pattern of factors indicates that older individuals also use strategies. Among predictive factors, the importance of criminal history is noteworthy. This factor has not previously been examined in relation to online sex offenders, despite its known role in offline sexual offenses (Davies 1992; Davies, Wittebrood and Jackson 1997). Individuals who have been in contact with the justice system are more likely to understand how it works and how their offenses were detected (Chopin, Beauregard and Bitzer 2020). To the extent that mistakes lead to learning (see Metcalfe 2017), individuals who have been detained by police – evidence of a mistake – may subsequently adapt their behaviors to limit future risks. Our results show that previous criminal convictions for online sexual crimes has greater predictive power of use of avoidance techniques than engaging in other forms of crime. Such a result may suggest that online sex offenders with specific experience, having replicated past actions would be more sensitive to the specific flaws that led to their detection (i.e., digital investigation techniques used by police, see e.g., Delle Donne and Fortin 2020; Fortin and Paquette 2018). Finally, consistent with history of contact with the justice system, our results indicate that cognition suggesting that the world is dangerous predicts this pattern. This belief is characteristic of men who distrust adults, authorities, and institutions (see Bartels and Merdian 2016; Ildeniz and Ó’Ciardha 2019; Paquette and Cortoni 2020; Paquette and Fortin 2021) and, potentially, also those they consider responsible for their convictions. This belief is also associated with the tendency to break rules, including criminal codes (Ildeniz and Ó’Ciardha 2019), suggesting that these individuals may be more likely to have had previous experience with the judicial system and more likely to suspect that such contact could occur again so methods to avoid detection should be used.

Conclusion

This study is the first to examine what the digital strategies used by online sex offenders to avoid police detection can tell us about these individuals. The analysis identified two patterns for acquiring such expertise: building on existing knowledge and learning skills through previous experiences.

This study has both theoretical and practical implications. First, the results confirm that online sex offenders, like other types of offenders, have a heterogeneous criminal expertise profile that influences their perception of the costs and benefits associated with their crimes, supporting the idea that any specific elements in their decision-making are related to how the crimes are committed – online – rather than to their characteristics as a distinct group of offenders. We also found that acquiring skills in the use of strategies to avoid police detection is similar to that observed with offline sexual offenders, although it can be expected that the younger online offenders will have more technical skills acquired outside the experience of the judicial process. Second, the findings have implications for various stakeholders. Police officers, like clinical practitioners, should try to better understand how online sex criminals operate, particularly in order to prioritize investigations. The results of this study could help guide decision-making in the police context.

In conclusion, we recommend that future studies replicate the present design with a larger sample to test the validity of the model presented. The model presented in this study is exploratory and confirmatory analyses are necessary. In addition, it would be interesting to include some more variables to increase the scope of the results, particularly in relation to socio-demographic and psychological factors. It would also be useful to conduct interviews with online sex offenders to better understand their motivations for using or not using strategies to avoid police detection. Finally, a more in-depth analysis using implicit measures of cognitions would make it possible to better understand the beliefs held by individuals involved in online sex offending and their role in the development of online criminal expertise.

Although innovative, there are limitations to this research that should be noted. First, the data used in this research comes from police files. Thus, the results pertain only to individuals who were identified by the police and charged. Although the literature on offline sex crimes indicates that there is no link between the use of strategies to avoid police detection and arrest/conviction for a crime, the situation may be different for crimes committed in virtual space. It is possible that there are individuals who have never been identified by the police because they use different, more effective, strategies. Second, we considered the use of strategies as homogeneous both on and offline. However, it is possible that the factors associated with the use of digital data protection or identity protection technologies are different from those for offline use. Third, the cognition variables used in this study were taken from the discourse of individuals in police custody and we cannot rule out that some of them represent ad hoc justifications intended to reduce the negative impacts associated with their crimes rather than reflecting the actual beliefs of online sex offenders. Given police goals when interviewing suspects (e.g., the discovery of evidence), it is possible that our method was not optimal to capture how offenders perceived the online behaviors and victims. Like traditional explicit measures of cognitions (i.e., psychometric questionnaires) that are limited by the fact that offenders may hide their actual beliefs, the method employed in this study also limits the capture of the full range of offenders’ cognitions; some may simply not have been reported by the offenders. This is unfortunately an inherent limitation of the study of sexual offenders’ cognitions (see Paquette & Fortin 2021 for a discussion). It should be note however that a previous study report virtually no difference between online sexual offenders’ cognitions identified during police interviews and during interviews conducted in clinical settings (Seto et al. 2010). Moreover, our current study was exempt researcher’s confirmatory bias (i.e., tendency to over focus on our research object; see Nickerson 1998) as the goals of the police were not to identify offenders’ cognitions. Another inherent limitation of studies on cognitions relates to determining their role in the process of sexual offending. Indeed, it remains difficult for researchers to determine whether the beliefs existed prior to the offenses, suggesting that they might play a contributing role in offending, or whether, on the contrary, they were simply formulated after the crime to avoid the consequences of their crime. Further research with different design would however benefit to clarify whether these cognitions reflect actual belief or only post-hoc justifications. Fourth, it is possible that some avoidance strategies were not identified, leading to some individuals being incorrectly identified as neophytes. Finally, we used a multivariate approach with a limited sample size. While the use of ANN models is quite appropriate with small sample sizes (Cui et al. 2004; Kim 2008), studies found that the ‘factor 10’ rule-of-thumb we followed in this study could be insufficient and recommend applying a ‘factor 50’ rule of thumb (Alwosheel, van Cranenburgh and Chorus 2018). Regarding these analytical issues, we believe that the results should be understood in terms of trends (i.e., positive or negative) rather than the exact values of the statistical weight of each factor.

References

Abu-Mostafa, Yaser S. 1993. "Hints and the VC dimension." Neural Computation 5(2):278-88.

Alwosheel, Ahmad, Sander van Cranenburgh, and Caspar G Chorus. 2018. "Is your dataset big enough? Sample size requirements when using artificial neural networks for discrete choice analysis." Journal of choice modelling 28:167-82.

Ariely, Dan, and George Loewenstein. 2006. "The heat of the moment: The effect of sexual arousal on sexual decision making." Journal of Behavioral Decision Making 19(2):87-98.

Assaad, J. M, and M.L Exum. 2002. "Understanding intoxicated violence from a rational choice perspective." Pp. 65-84 in Rational choice and criminal behavior: Recent research and future challenges, edited by A. R Piquero and S. G Tibbetts. New York, NY: Routledge.

Authors. in press a.

Babchishin, Kelly M, R Karl Hanson, and Heather VanZuylen. 2015. "Online child pornography offenders are different: A meta-analysis of the characteristics of online and offline sex offenders against children." Archives of Sexual Behavior 44(1):45-66.

Balfe, Myles, Bernard Gallagher, Helen Masson, Shane Balfe, Ruairi Brugha, and Simon Hackett. 2015. "Internet child sex offenders' concerns about online security and their use of identity protection technologies: a review." Child abuse review 24(6):427-39.

Bartels, Ross M, and Hannah L Merdian. 2016. "The implicit theories of child sexual exploitation material users: An initial conceptualization." Aggression and Violent Behavior 26:16-25.

Beauregard, E, and M Bouchard. 2010. "Cleaning up your act: Forensic awareness as a detection avoidance strategy." Journal of Criminal Justice 38(6):1160-66.

Beauregard, E, B Leclerc, and P. Lussier. 2012. "Decision making in the crime commission process: Comparing rapists, child molesters, and victim-crossover sex offenders." Criminal Justice and Behavior 39(10):1275-95.

Beauregard, E, P. Lussier, and J Proulx. 2005. "The role of sexual interests and situational factors on rapists' modus operandi: implications for offender profiling." Legal and Criminological Psychology 10:265-78.

Beech, Anthony R, Ian A Elliott, Astrid Birgden, and Donald Findlater. 2008. "The Internet and child sexual offending: A criminological review." Aggression and Violent Behavior 13(3):216-28.

Bigi, R, D Gregori, L Cortigiani, A Desideri, F. A. Chiarotto, and G. M. Toffolo. 2005. "Artificial neural networks and robust Bayesian classifiers for risk stratification following uncomplicated myocardial infarction." International journal of cardiology 101(3):481-87.

Bourke, P, T. Ward, and C Rose. 2012. "Expertise and sexual offending: A preliminary empirical model." Journal of interpersonal violence 27(12):2391-414.

Briggs, Peter, Walter T Simon, and Stacy Simonsen. 2011. "An exploratory study of internet-initiated sexual offenses and the chat room sex offender: Has the internet enabled a new typology of sex offender?" Sexual Abuse 23(1):72-91.

Cale, Jesse, Thomas Holt, Benoit Leclerc, Sara Singh, and Jacqueline Drew. 2021. "Crime commission processes in child sexual abuse material production and distribution: A systematic review." Trends and Issues in Crime and Criminal Justice (617):1-22.

Cherbonneau, M, and H Copes. 2005. "‘Drive it like you Stole it’ Auto Theft and the Illusion of Normalcy." British Journal of Criminology 46(2):193-211.

Chopin, J, E Beauregard, and S Bitzer. 2020. "Factors influencing the use of forensic awareness strategies in sexual homicide." Journal of Criminal Justice 71:1-9.

Chopin, J, E Beauregard, S Bitzer, and K Reale. 2019. "Rapists’ behaviors to avoid police detection." Journal of Criminal Justice 61(2019):81-89.

Chopin, J, S. Paquette, and E Beauregard. 2021. "Is There an Expert Stranger “Rapist” " Sex Abuse Advanced Online Publication.

Colombet, Isabelle, Alan Ruelland, Gilles Chatellier, François Gueyffier, Patrice Degoulet, and Marie-Christine Jaulent. 2000. "Models to predict cardiovascular risk: comparison of CART, multilayer perceptron and logistic regression." American Medical Informatics Association Annual Symposium Proceedings Archive:156-60.

Copes, H, and M Cherbonneau. 2006. "The key to auto theft: emerging methods of auto theft from the offenders’ perspective." British Journal of Criminology 46(5):917-34.

Cornish, D. B, and R. V. Clarke. 1986. "Introduction." Pp. 1-18 in The Reasoning Criminal: Rational choice perspectives on offending, edited by D. B Cornish and R. V. Clarke. New York: Springer-Verlag.

—. 1987. "Understanding crime displacement: An application of rational choice theory." Criminology 25(4):933-48.

—. 2008. "The rational choice perspective." in Environmental Criminology and Crime Analysis, edited by R. Wortley and L. Mazerolle. Cullompton Devon: Willan Publishing.

Cui, Ying-Jin, S Davis, Chao-Kun Cheng, and Xue Bai. 2004. "A study of sample size with neural network." Pp. 3444-48 in International Conference on Machine Learning and Cybernetics. Shanghai, China IEEE.

D’Ovidio, Rob, Tyson Mitman, Imaani Jamillah El-Burki, and Wesley Shumar. 2009. "Adult-child sex advocacy websites as social learning environments: A content analysis." International Journal of Cyber Criminology 3(3):421-40.

Dalins, Janis, Campbell Wilson, and Mark Carman. 2018. "Criminal motivation on the dark web: A categorisation model for law enforcement." Digital Investigation 24:62-71.

Davies, A. 1992. "Rapists' behaviour: A three aspect model as a basis for analysis and the identification of serial crime." Forensic Science International 55(2):173-94.

Davies, A, K Wittebrood, and J. L Jackson. 1997. "Predicting the criminal antecedents of stranger rapist from his offence behaviour." Sciences and Justice 37:161-70.

Delle Donne, J, and F Fortin. 2020. "Pratiques policières et utilisation des médias sociaux." Pp. 23-36 in Cybercrimes et enjeux technologiques, edited by F Fortin. Montréal: Presses Internationales Polytechnique.

Dillard, Elizabeth, Fred A Luchette, Benjamin W Sears, John Norton, Carol R Schermer, R Lawrence Reed II, Richard L Gamelli, and Thomas J Esposito. 2007. "Clinician vs mathematical statistical models: which is better at predicting an abnormal chest radiograph finding in injured patients?" American Journal of Emergency Medicine 25(7):823-30.

Douglas, J., A. W. Burgess, A. G. Burgess, and R. K. Ressler. 2006. Crime Classification Manual : A standard system for investigating and classifying violent crimes: Jossey-Bass.

Dror, I. E. 2011. "The paradox of human expertise: why experts get it wrong." The paradoxical brain 177.

Elliott, Ian A, and Anthony R Beech. 2009. "Understanding online child pornography use: Applying sexual offense theory to internet offenders." Aggression and Violent Behavior 14(3):180-93.

Eneman, Marie. 2009. "Counter-surveillance strategies adopted by child pornographers." International Journal of Technology and Human Interaction (IJTHI) 5(4):1-17.

Exum, M Lyn. 2006. "Alcohol and aggression: An integration of findings from experimental studies." Journal of Criminal Justice 34(2):131-45.

Fortin, F, and P Corriveau. 2015. Who is Bob_34?: Investigating child cyberpornography. Vancouver: UBC Press.

Fortin, F, and S Paquette. 2018. "Online Sexual Exploitation of Children: Reactive and Proactive Policing." Pp. 237-56 in Sexual Offending: A Criminological Perspective, edited by P Lussier and S Paquette. New York, NY: Routledge.

Fortin, F, S Paquette, and B Dupont. 2018. "From online to offline sexual offending: Episodes and obstacles." Aggression and Violent Behavior 39:33-41.

Fortin, F, and J Roy. 2006. "Profils des consommateurs de pornographie juvénile arrêtés au Québec : l’explorateur, le pervers et le polymorphe." Criminologie 39(1):107-28.

Gallupe, O, M Bouchard, and J. P. Caulkins. 2011. "No change is a good change? Restrictive deterrence in illegal drug markets." Journal of Criminal Justice 39(1):81-89.

Gibbs, J. P. 1975. Crime, punishment, and deterrence: Elsevier.

Haykin, Simon S. 2009. Neural networks and learning machines. New York: Prentice Hall.

Holt, Thomas J, Kristie R Blevins, and Natasha Burkert. 2010. "Considering the pedophile subculture online." Sexual Abuse 22(1):3-24.

Ildeniz, G. , and C Ó’Ciardha. 2019. "A dangerous world implicit theory: examining overlap with other criminogenic constructs." Journal of sexual aggression Advanced Online Publication.

Jacobs, B. A. 1996a. "Crack dealers and restrictive deterrence: Identifying narcs." Criminology 34(3):409-31.

—. 1996b. "Crack dealers' apprehension avoidance techniques: A case of restrictive deterrence." Justice Quarterly 13(3):359-81.

Jacobs, B. A., and M Cherbonneau. 2018. "Perceived sanction threats and projective risk sensitivity: Auto theft, carjacking, and the channeling effect." Justice Quarterly 35(2):191-222.

Jacobs, B. A., and J. D. Miller. 1998. "Crack dealing, gender, and arrest avoidance." Social Problems 45(4):550-69.

Jacobs, Bruce A. 2010. "Deterrence and deterrability." Criminology 48(2):417-41.

Jacobs, Bruce A, and Michael Cherbonneau. 2014. "Auto theft and restrictive deterrence." Justice Quarterly 31(2):344-67.

Kierkegaard, Sylvia. 2011. "To block or not to block–European child porno law in question." Computer Law & Security Review 27(6):573-84.

Kim, Yong Soo. 2008. "Comparison of the decision tree, artificial neural network, and linear regression methods based on the number and types of independent variables and sample size." Expert Systems with Applications 34(2):1227-34.

Liu, Y. Y., M Yang, M Ramsay, X. S. Li, and J. W. Coid. 2011. "A comparison of logistic regression, classification and regression tree, and neural networks models in predicting violent re-offending." Journal of Quantitative Criminology 27(4):547-73.

Lussier, P., M Bouchard, and E Beauregard. 2011. "Patterns of criminal achievement in sexual offending: Unravelling the “successful” sex offender." Journal of Criminal Justice 39(5):433-44.

Lussier, P., and J Mathesius. 2012. "Criminal achievement, criminal career initiation, and detection avoidance: The onset of successful sex offending." Journal of Crime and Justice 35(3):376-94.

Maimon, David. 2020. "Deterrence in Cyberspace: An Interdisciplinary Review of the Empirical Literature." Pp. 449-68 in The Palgrave Handbook of International Cybercrime and Cyberdeviance, edited by T. J. Holt and Adam M Bossler: Palgrave Macmillan.

Maimon, David, C Jordan Howell, and George W Burruss. 2021. "Restrictive deterrence and the scope of hackers’ reoffending: Findings from two randomized field trials." Computers in Human Behavior 125:106943.

Maruna, Shadd, and Ruth E Mann. 2006. "A fundamental attribution error? Rethinking cognitive distortions." Legal and Criminological Psychology 11(2):155-77.

Meenaghan, Amy, Claire Nee, Jean-Louis Van Gelder, Zarah Vernham, and Marco Otte. 2020. "Expertise, Emotion and Specialization in the Development of Persistent Burglary." The British Journal of Criminology 60(3):742-61.

Metcalfe, Janet. 2017. "Learning from errors." Annual review of psychology 68:465-89.

Moayed, Farman A, and Richard L Shell. 2011. "Application of artificial neural network models in occupational safety and health utilizing ordinal variables." Annals of occupational hygiene 55(2):132-42.

Nee, C. 2015. "Understanding expertise in burglars: From pre-conscious scanning to action and beyond." Aggression and Violent Behavior 20:53-61.

Nee, C, and A Meenaghan. 2006. "Expert decision making in burglars." British Journal of Criminology 46(5):935-49.

Nee, C, and T. Ward. 2015. "Review of expertise and its general implications for correctional psychology and criminology." Aggression and Violent Behavior 20:1-9.

Nee, C, M White, K Woolford, T Pascu, L Barker, and L Wainwright. 2015. "New methods for examining expertise in burglars in natural and simulated environments: preliminary findings." Psychology, Crime & Law 21(5):507-13.

Nickerson, R. S. 1998. "Confirmation bias: A ubiquitous phenomenon in many guises." Review of General Psychology, 2: 175-220.

Nunnally, Jum C, and I. H. Bernstein. 1994. "The assessment of reliability." Psychometric theory 3:248-92.

Ó Ciardha, C. 2015. "Experts in rape: Evaluating the evidence for a novice-to-expert continuum in the offense behavior and cognition of sexual offenders." Aggression and Violent Behavior 20:26-32.

Paquette, Sarah, and Franca Cortoni. 2020. "The development and validation of the Cognitions of Internet Sexual Offending (C-ISO) Scale." Sexual Abuse 32(8):907-30.

Paquette, Sarah, and Francis Fortin. 2021. "A Peek into Their Mind? An Exploration of Links Between Offense-Supportive Statements and Behaviors among Men Who Sexually Exploit Children and Adolescents Online." International Journal of Offender Therapy and Comparative Criminology:0306624X211013523.

Peterson, J. B., J Rothfleisch, P. D. Zelazo, and R. O. Pihl. 1990. "Acute alcohol intoxication and cognitive functioning." Journal of Studies on Alcohol 51(2):114-22.

Price, Rumi Kato, Edward L Spitznagel, Thomas J Downey, Donald J Meyer, Nathan K Risk, and Omar G El-Ghazzawy. 2000. "Applying artificial neural network models to clinical decision making." Psychological assessment 12(1):40.

Prichard, Jeremy, Paul A Watters, and Caroline Spiranovic. 2011. "Internet subcultures and pathways to the use of child pornography." Computer Law & Security Review 27(6):585-600.

Quayle, Ethel, and Max Taylor. 2011. "Social networking as a nexus for engagement and exploitation of young people." Information security technical report 16(2):44-50.

Reale, K, E Beauregard, and J Chopin. 2021a. "Criminal Expertise and Sexual Violence: Comparing the Crime-Commission Process Involved in Sexual Burglary and Sexual Robbery " Criminal Justice and Behavior.

—. 2021b. "Expert Versus Novice: Criminal Expertise in Sexual Burglary and Sexual Robbery " Sexual Abuse.

—. 2021c. "Is Criminal Expertise a Feature of Unsolved Sexual Assault Involving Theft?" International Journal of Offender Therapy and Comparative Criminology.

—. 2021d. "The role of criminal expertise in serial sexual offending: a comparison to “novices”." Journal of criminal psychology Advanced Online Publication.

Rossmo, K. 2009. Criminal investigative failures. Boca Raton, FL: CRC Press.

Seto, Michael C, Lesley Reeves, and Sandy Jung. 2010. "Explanations given by child pornography offenders for their crimes." Journal of sexual aggression 16(2):169-80.

Sheehan, Valerie, and Joe Sullivan. 2010. "A qualitative analysis of child sex offenders involved in the manufacture of indecent images of children." Journal of sexual aggression 16(2):143-67.

Skakoon-Sparling, Shayna, Kenneth M Cramer, and Paul A Shuper. 2016. "The impact of sexual arousal on sexual risk-taking and decision-making in men and women." Archives of Sexual Behavior 45(1):33-42.

Steadman, Henry J, Eric Silver, John Monahan, Paul Appelbaum, Pamela Clark Robbins, Edward P Mulvey, Thomas Grisso, Loren H Roth, and Steven Banks. 2000. "A classification tree approach to the development of actuarial violence risk assessment tools." Law and human behavior 24(1):83-100.

Steel, Chad MS, Emily Newman, Suzanne O’Rourke, and Ethel Quayle. 2020. "An integrative review of historical technology and countermeasure usage trends in online child sexual exploitation material offenders." Forensic Science International: Digital Investigation 33:300971.

Topalli, V. 2005. "Criminal expertise and offender decision-making: An experimental analysis of how offenders and non-offenders differentially perceive social stimuli." The British Journal of Criminology 45(3):269-95.

Topalli, V, S Jacques, and R Wright. 2015. "“It takes skills to take a car”: Perceptual and procedural expertise in carjacking." Aggression and Violent Behavior 20:19-25.

Tu, Jack V. 1996. "Advantages and disadvantages of using artificial neural networks versus logistic regression for predicting medical outcomes." Journal of clinical epidemiology 49(11):1225-31.

Ward, T. 1999. "Competency and deficit models in the understanding and treatment of sexual offenders." Journal of sex research 36(3):298-305.

Wilson, Theodore, David Maimon, Bertrand Sobesto, and Michel Cukier. 2015. "The effect of a surveillance banner in an attacked computer system: Additional evidence for the relevance of restrictive deterrence in cyberspace." Journal of Research in crime and Delinquency 52(6):829-55.

Wolak, Janis, David Finkelhor, and Kimberly Mitchell. 2011. "Child pornography possessors: Trends in offender and case characteristics." Sexual Abuse 23(1):22-42.

Wolak, Janis, David Finkelhor, Kimberly J Mitchell, and Michele L Ybarra. 2010. "Online “predators” and their victims: Myths, realities, and implications for prevention and treatment."

Zhang, Xiaolu. 2010. "Charging children with child pornography–Using the legal system to handle the problem of “sexting”." Computer Law & Security Review 26(3):251-59.

Appendix

Appendix 1

Diagnosis of multicollinearity of independent variables

Variables

Variance inflation factor (VIF)

Tolerance

Age (continuous)

1 179

0.848

Employed

1 081

0.925

Offence involves child pornography

1 226

0.816

History of online non-sexual crime

1 180

0.847

History of online sex crimes

1 143

0.875

Problematic substance use

1 167

0.857

Tendency to social isolation

1 216

0.822

Pedohebephilic sexual interest

1 282

0.780

Sexual preoccupation

1 348

0.742

Cognitions:

Children have sexual interests

1 199

0.834

Sexual offences are uncontrollable

1 085

0.922

Offline world is dangerous

1 226

0.816

Virtual universe is not reality

1 141

0.877

Appendix 2

Area under the curve of the technology use prediction model

Comments
0
comment

No comments here

Why not start the discussion?