The conceptualization of dynamic risk factors in child sex offenders: An
agency model

Roxanne Heffernan, Tony Ward

PII: S1359-1789(15)00086-5
DOI: doi: 10.1016/j.avb.2015.07.001
Reference: AVB 929

To appear in: Aggression and Violent Behavior

Received date: 21 May 2015

Revised date: 17 June 2015
Accepted date: 24 June 2015

Please cite this article as: Heffernan, R. & Ward, T., The conceptualization of dynamic
risk factors in child sex offenders: An agency model, Aggression and Violent Behavior
(2015), doi: 10.1016/j.avb.2015.07.001

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 ACCEPTED MANUSCRIPT
1

Running Head: Dynamic risk factors and sexual offending

The Conceptualization of Dynamic Risk Factors in

PT
Child Sex Offenders: An Agency Model

RI
SC
Roxanne Heffernan

Tony Ward

NU
Victoria University of Wellington
MA

Corresponding Author:
ED

Dr. Tony Ward

School of Psychology
Victoria University of Wellington,
PT

PO Box 600.
Wellington, New Zealand.
E-mail: Tony.Ward@vuw.ac.nz
CE
AC
 ACCEPTED MANUSCRIPT
2

Abstract

PT
The current preoccupation of criminal justice practitioners and policy makers with

RI
the prediction of reoffending has resulted in a conceptualization of risk as simply

clusters of factors that correlate with recidivism. The reliance on these phenomena as

SC
explanations for the causes of sexual offending and as guides for treatment is a

NU
mistake, and in our view, the conceptualization of dynamic risk needs to be

reexamined. This article begins with a discussion of the factors that increase and
MA
decrease risk of sexual offending; the focus is then widened to include agency,

motivation, and values. These normative features are integrated with risk-related
ED

factors within the action-based Agency Model of Risk (AMR). This dynamic,

interactional model highlights the importance of the relationship between the agent
PT

and context, with both proposed to contribute to the patterns of behavior resulting in
CE

an offense. The AMR is applied to a number of dynamic risk domains for sexual

offenders, and its utility in explaining behavior and informing treatment discussed.
AC

Key Words: sex offenders; dynamic risk factors; protective factors

 ACCEPTED MANUSCRIPT
3

The assessment of dynamic risk factors is a major concern of contemporary

researchers and practitioners in the sexual offending field (Mann, Hanson, &

Thornton, 2010; Ward, 2014). In part, this is because of a demand for better risk

PT
prediction measures and risk management technologies to protect the community

from dangerous individuals. It is easier to take steps to minimize the risk of harm

RI
through precautionary efforts than to deal directly with the consequences of sexual

SC
abuse. Identifying the features of individuals who go on to offend and factors that

minimize the chances of this occurring are priorities. However, another reason for

NU
the current preoccupation with dynamic risk and protective factors is that they offer

hope of explaining offending, and as such can be used to structure assessment and
MA
treatment of individuals who commit sexual offenses. The recruitment of dynamic

risk factors to formulate cases and to inform treatment is now standard clinical
ED

practice (Ward & Beech, 2015).

PT

Running against the tide of contemporary practice, we think that the

importation of dynamic risk (and protective) factors from the domain of risk
CE

prediction into the domains of etiology and treatment is deeply problematic. It

assumes that dynamic risk factors are causes and refer to the psychological,
AC

biological, and social processes that actually result in sexual offenses. In our view,

this is a dubious assumption, one that could quickly lead the field into theoretical and

practice dead ends (Ward & Beech, 2015). A pressing concern is the lack of coherence

between theoretical explanations of behavior and the prediction and management of

risk (Ward, 2014). While there have been recent attempts to link possible causal

psychological factors with empirically supported variables in the sexual offending

literature (e.g., Mann, Thornton, & Hanson, 2010), there is still an emphasis on

prediction at the expense of explanation (Ward, 2014). Prediction is useful, but it is

 ACCEPTED MANUSCRIPT
4

incorrect to assume that psychometric validity alone is proof of a causal relationship,

or that it can adequately guide practice.

Over 10 years ago Beech and Ward (2004) explored the possible conceptual

PT
and causal relationships between etiological theories and static and dynamic risk

factors in their Etiological Model of Risk. This model combined several influential

RI
etiological theories concerning child sex offenders with empirically derived risk

SC
factors. However, despite this work, and promising theoretical developments since

that time, the relationship between risk and etiology has been relatively neglected

NU
and important questions remain unanswered. For example, it is unclear how dynamic

risk factors and psychological mechanisms interact to cause and maintain sexual
MA
offending, or whether they actually refer to causes at all- as opposed to predicting

recidivism.
ED

In this paper, we take a closer look at the nature and function of dynamic risk
PT

factors, and their ability to explain sexual offending and to directly guide treatment.

The first section of this paper concentrates on the current dominant

CE

conceptualization of risk, and explores how this account could be improved on

through a more in-depth exploration of the content of dynamic risk factors. Four
AC

broad risk domains will be examined (cognition, self-regulation, interpersonal

functioning, and sexual), and unpacked into their various components and functions,

and protective factors are briefly discussed. The second section will explore what is

missing from the contemporary view of risk, namely motivational and normative

features. In this section we also outline a new etiological model of risk, the Agency

Model of Risk (AMR), illustrating its ability to link the various facets of dynamic risk

factors with sexual offending behavior. In the third section the capacity of the AMR to

accommodate facets of dynamic risk factors is explored. Finally, in the fourth section,
 ACCEPTED MANUSCRIPT
5

we evaluate the theoretical and practical implications of the AMR, and offer

suggestions for future development of the model.

The primary aim of this paper is to encourage greater discussion concerning the

PT
motivational, psychological, and behavioral processes relating to criminal risk. We do

not pretend that the AMR is the last word on the etiology of risk and appreciate that it

RI
is simply one attempt to create a conceptual bridge between risk assessment and

SC
etiological theory. However, the roots of the AMR in agency theory and evolutionary

research, and its claim that dynamic risk factors are composite constructs, are new

NU
ideas that will hopefully advance our thinking about the nature of risk.
MA
1. Risk

Empirical and theoretical research supports the conclusion that there are at least
ED

four major domains of risk (and their various components) that a causal model of

sexual offending ought to accommodate: cognition, self-regulation, interpersonal

PT

functioning, and sexual (Hanson & Harris, 2000; Mann et al., 2010; Thornton, 2013)
CE

(see Table 1). These risk factors are much like observable symptoms of psychological

disorders. We cannot confidently say why they are present or how they function, but
AC

they are widely recognized (and empirically supported) as problems experienced by

individuals who sexually abuse children. However, as things presently stand the

relationship between dynamic risk factors and sexual offending is primarily one of

association and does not imply direct causation.

--------------------------------------------

Place table 1 about here

--------------------------------------------

In addition, as recently argued by Ward and Beech (2015), risk factors

 ACCEPTED MANUSCRIPT
6

identified in the literature contain both causal and descriptive elements, which need

to be teased apart in order to avoid conflating distinct constructs. For example,

despite both being included in the dynamic risk category of interpersonal problems,

PT
lacking emotionally close adult relationships is an observable problem whereas

emotionally identifying with children is better viewed as one of its possible causes.

RI
Dynamic risk factors are composite constructs in at least three distinct senses.

SC
They are frequently discussed in the literature and appear in psychometric models as

a type of construct whereas in fact the four general domains are heterogeneous

NU
collections of specific dynamic risk factors; thus, the general category labels are

simply place-holders for clusters of heterogeneous and concrete factors. For example,
MA
de Vries Robbé (2014) unpacks the general category of dynamic risk factors evident

in sex offenders into such elements as sexual preoccupation, deviant sexual interests,
ED

offense-supportive attitudes, emotional congruence with children, impulsiveness,

PT

poor cognitive problem solving, grievance/hostility, and lack of concern for others.

The difficulty is that later in his monograph he outlines a proposed explanatory

CE

model that uses the general term risk factors in which he explores its relationship to

violence and protective factors (de Vries Robbé, 2014). At best, this use of the term is
AC

misleading, while at worst, it conflates levels of constructs. That is, it is not clear

whether it refers to the overall set of domains of risk factors, the four domains

themselves, or to the particular dynamic risk factors that are contained within each of

the domain categories.

A further problem related to the composite nature of dynamic risk factors is

that each domain is typically broken down into further features, some of which

causally exclude each other. For example, in his recent summary of risk and

protective factors in adult male sexual offenders, Thornton (2013) listed sexual

violence and sexual interest in children as subdomains of the general dynamic risk
 ACCEPTED MANUSCRIPT
7

factor of sexual interests. The problem is that the “umbrella,” so to speak, of deviant

sexual interests consists of qualitatively different variables, which arguably refer to

distinct causal processes and their associated problems. Finally, the description of

PT
dynamic risk factors is vague and seems to include both dispositional and state

aspects. For example, the stable dynamic factor of general self-regulation includes

RI
negative emotionality (a mental state) and poor problem-solving (an enduring

SC
psychological feature). Taking into account the heterogeneous nature of dynamic risk

factors, we propose that they are composite constructs, requiring further analysis to

NU
disentangle their various, inter-related components. We will explore this issue in

greater depth below.

MA
1.1. Risk Assessment: The Pre-occupation with Prediction
ED

Risk prediction largely involves actuarial (statistical) assessment tools that

combine a number of empirically derived “risk factors” to form a total risk score for
PT

the individual. Risk factors were initially categorized as static or dynamic, based on

whether or not they could be directly or indirectly modified. Static factors include
CE

demographic features and historical events (e.g., age and past offenses), which are
AC

unable to be altered. Dynamic risk factors can in theory be changed, for example

offense-supportive attitudes might be replaced with prosocial attitudes and their

associated beliefs; changes which predict reduced recidivism. For this reason a subset

of dynamic factors have been termed criminogenic needs, and are preferred

treatment targets in sex offending and general correctional treatment programmes

(Andrews & Bonta, 2010). Thus, researchers and clinicians believe that dynamic risk

factors influence behavior because when they are successfully targeted in treatment

reoffending rates decrease (Hanson & Morton-Bourgon, 2009). It is assumed that

these factors are causal, but without an adequate explanation of their nature and
 ACCEPTED MANUSCRIPT
8

functions it is difficult to determine whether or not this is the case (Illari & Russo,

2014). Certainly dynamic risk factors are useful for prediction because they are

associated with offending in some way, but on their own they do not explain how or

PT
why offending occurs (Ward & Beech, 2015).

Dynamic risk factors have been further divided into stable and acute subtypes,

RI
based on their persistence over time (Hanson & Harris, 2000). Stable factors persist

SC
across weeks or months (e.g., antisocial attitudes), while acute factors increase the

imminent risk of offending (e.g., intense emotional states). In their etiological model

NU
of risk, Beech and Ward (2004) reconceptualize this distinction in terms of traits and

states (see below).

MA
1.2. Etiological Theory and Risk
ED

Theory creation is crucial in informing and generating empirical research, as

well as guiding practice. Yet, it has been relatively neglected in recent years as
PT

interest has shifted to psychometric and empirical research (Ward, 2014). Relatedly,

the conceptual and practical relationship between the tasks of theory creation and
CE

risk assessment appears somewhat tenuous, and arguably, overlooked (Ward, 2014).
AC

As stated above, Beech and Ward’s (2004) etiological model began to address this

issue by combining the most influential etiological theories and empirically derived

predictors into a conceptual model. This model outlined how developmental factors

lead to vulnerability (traits), which when triggered by contextual features resulted in

acute dynamic risk states. Static risk factors are conceptualized as historical markers

of existing psychological vulnerabilities, hence their usefulness in risk prediction

(Beech & Ward, 2004). Risk factors are largely a combination of both long-term

vulnerabilities (as evidenced by past offenses) and their manifestations in certain

contexts. Thus categories of risk, such as “intimacy deficits”, are more accurately
 ACCEPTED MANUSCRIPT
9

viewed as composite constructs, containing causal, descriptive, developmental, and

contextual elements. A weakness of Beech and Ward’s (2004) risk etiology model is

that it simply grouped dynamic risk factors into etiological phases and did not specify

PT
how they combine, or interact with contextual variables, to cause sexual offending.

Mann et al. (2010) further investigated the existence of causal factors when

RI
they proposed a list of psychologically meaningful risk factors for persons who

SC
sexually abuse children. According to Mann et al. (2010), in addition to being

empirically linked with recidivism, these factors must be plausible causes of

NU
offending– this is where theoretical knowledge becomes crucial. These authors

conceptualized dynamic risk factors as propensities to offend that are triggered in

MA
certain contexts, suggesting that they are vulnerabilities that may or may not cause a

sexual offense. Mann et al. (2010) also proposed that it is not necessary for
ED

psychologically meaningful factors to be completely changeable. For example, a

PT

sexual preference for children may endure despite treatment, but be managed

through “compensatory strengths or prosthetics” (Mann et al., 2010, p. 195). In their

CE

paper, they list propensities with strong empirical support, for example “offense-

supportive attitudes”, and “general self-regulation problems” (Mann et al., 2010).

AC

Mann et al. (2010) provide a noteworthy contribution by highlighting the most likely

causal factors for sex offenders. However, they acknowledge that a “deeper

conceptualization of each risk factor would provide coherent and empirically justified

accounts of how these risk factors develop and how they cause offending” (Mann et

al., 2010, p. 209). Furthermore, they only concentrated on the difference making

aspect of causality (i.e., identified sexual offending’s possible causal factors and

relationship to their effects- reoffending) and overlooked the need to also specify the

mechanisms that mediate the relationship between causal factors and their outcomes

(for the requirements of establishing causality in science see Illari & Russo, 2014).
 ACCEPTED MANUSCRIPT
10

We will now examine the four major classes of dynamic risk factors in greater

detail to support our argument that the concept of dynamic risk factors is a composite

construct, and as such, does not pick out unified types of causal mechanisms.

PT
1.3. Cognitive domain

Offense-supportive attitudes or beliefs are widely accepted as important

RI
treatment targets, and are central causes within all comprehensive theories of sexual

SC
offending (Ward & Casey, 2010; Ward, Polaschek, & Beech, 2006). These factors are

often termed “cognitive distortions” and broadly defined as norm-violating attitudes

NU
and beliefs, biased interpretations, and post-hoc excuses, minimizations,
MA
rationalizations, and justifications (Ó Ciardha & Ward, 2013). One problem with this

risk factor is its broad application to several types of problematic cognition, which
ED

differ in their functions over time. Like other risk factors, the cognitive domain is a

composite construct, consisting of descriptive and causal elements. For example,

PT

problematic cognitions are likely to be ongoing vulnerabilities (i.e., distorted content

of beliefs), leading to risky states (i.e., biased cognitive processing), and descriptions
CE

of their observable manifestations (i.e., verbal justifications).

AC

Ward and Keenan (1999) propose that cognitive distortions develop from

implicit theories (i.e., maladaptive core schema) concerning the nature of the self,

others, and social environment. Ward and Keenan (1999) identified five implicit

theories for child sex offenders. The first two implicit theories, “children as sexual

beings” and “the nature of harm”, primarily involve children as potential victims.

They represent beliefs that children are capable of choosing to have sex, and that this

is not harmful (Ward & Keenan, 1999). These theories function by facilitating

distorted interpretations of children’s actions, then justifying offenders’ subsequent

sexually abusive behavior in response to these interpersonal cues (Ward & Keenan,
 ACCEPTED MANUSCRIPT
11

1999). The third implicit theory “dangerous world” states that nobody can be trusted

and leads to the intimacy deficits often associated with child sex offenders (Ward,

Keenan, & Hudson, 2000). The final two theories concern an individual’s self-

PT
perception, specifically his lack of self-control and entitlement to sexual satisfaction

(Ward & Keenan, 1999). It is important to consider that while these theories

RI
represent broad categories of belief, they are not exhaustive, nor present in all

SC
individuals (Ward & Keenan, 1999).

1.4. Self-regulation domain

NU
The capacity for self-regulation has been discussed extensively in the sexual
MA
offending literature, and assessed with the risk factor “general self-regulation

problems”. Self-regulation may be defined as “the internal and external processes

ED

that allow an individual to engage in goal-directed actions over time and in different

contexts” (Ward & Hudson, 1998, p. 702). This definition is strikingly similar to the
PT

concept of agency, which will be explored later in this paper. Self-regulation requires

skills such as insight, problem solving, goal setting, inhibition, and the ongoing
CE

adjustment of action in response to internal and external challenges. These skills are
AC

usually learned early in life, through modeling of coping and problem solving

behaviors by others, and through social reinforcement. The key to successful self-

regulation is selecting adaptive (pro-social) goals and effective means to achieve

these. Goals, values, and individual preferences motivate action towards desired

outcomes, based on the individual’s beliefs, expectations, and ability to self-regulate.

Ward and Hudson’s (1998) Self-regulation Model is based on an offense chain,

including the triggering events, desires, goals, and plans that lead to “high risk

situations.” In this model, sexual offending can occur when goal directed actions are

disrupted in some way by external events or internal states. For example, substance
 ACCEPTED MANUSCRIPT
12

abuse may impair self-regulation by virtue of its adverse impact on attentional

control, thereby lowering inhibitions and resulting in impulsive sexually abusive

behavior. In this case, poor self-regulation is conceptualized as a state, rather than an

PT
enduring trait. On the other hand, some individuals engage in a degree of planning

(whether implicit or explicit) when committing a sexual offense, and the

RI
manipulation of physical and social environments to facilitate offending can become

SC
more sophisticated over time (Smallbone & Wortley, 2004). Thus, there appears to be

a contrast between individuals who impulsively take advantage of offending

NU
opportunities and those who spend time grooming victims and gaining their trust. Of

course these are two opposite ends of a continuum, and offenders vary in their degree
MA
of planning and the sophistication of their self-regulation strategies, depending on

their criminal expertise and the external context.

ED

It is worth noting that the self-regulation domain of risk is heavily influenced

PT

by problems within the other three domains - cognitive, social, and sexual -, as well as

being motivated by the desire for a valued outcome. These desires (values) might be
CE

translated into interpersonal goals where the individual seeks intimacy but does not

feel able to achieve this in more appropriate ways. On the other hand, the goal may be
AC

purely sexual, for example, when the individual is feeling sexually aroused, or more

specifically, is only sexually attracted to children. A third option is when an offense

related goal is a combination of both, in individuals who conflate their needs for

intimacy and sexual gratification. The interpersonal and sexual domains of risk relate

to why an individual offends, tapping into unmet needs. On the other hand, core

beliefs and self-regulation influence how an individual justifies and actually engages

in the offense process. What this reveals is the way dynamic risk factors, when broken

down into their component causal strands: (a) are composed of multiple causal
 ACCEPTED MANUSCRIPT
13

strands, and (b) that they combine to result in sexual offending (Durrant & Ward,

2015).

1.5. Interpersonal domain

PT
The risk factor “intimacy deficits” or “interpersonal problems” indicates an

incapacity for appropriate intimacy, or a lack of access to interpersonal supports and

RI
opportunities. Its key components in child sex offenders include deficits in intimacy

SC
with adults and emotional identification with children. These problems arise from

many causes, both developmental and contextual (Thornton, 2013; Ward et al.,

NU
2006). For example, an individual who suffered abuse or neglect during childhood
MA
could find it difficult to trust other adults. He may avoid adult relationships in favor

of children or not possess the interpersonal skills necessary to establish and sustain
ED

adult romantic relationships (Navathe, Ward, & Rose, 2013).

Interpersonal problems have a significant influence on individuals’ other

PT

domains of functioning. It has been argued that the quality of interpersonal

attachments throughout life influence emotional states (Beech & Mitchell, 2005),
CE

which can act as offense triggers for some individuals. Relatedly, it has been proposed
AC

that insecure attachment can lead to other risk factors such as maladaptive coping

and poor self-regulation skills (Ward et al., 2006). It is also important to consider

how attachment experiences shape core beliefs about the self, others, and the world.

If a child experiences rejection or abuse, he or she might be more likely to form

maladaptive attitudes and expectations concerning future interactions with adults. If

an individual lacks the capacity for emotional intimacy with other adults, he may

engage in impersonal or deviant sexual behavior in order to reach some level of

intimacy. As this type of behavior is not a viable long term strategy in traditional

adult relationships, children may be perceived as more accepting, and better able to
 ACCEPTED MANUSCRIPT
14

meet sexual needs. Thus, in keeping with the composite nature of dynamic risk

factors, problems with intimacy can take many forms: loneliness (mental state), poor

quality relationships with adults (situational), distrust of other people (disposition),

PT
lack of communication skills (disposition), and so on. These may exist independently

or co-occur, depending on the person, their particular history, and the current

RI
situation.

SC
1.6. Sexual domain

It is often assumed that all child sex offenders are pedophiles motivated by

NU
deviant sexual interests. However, this is not the case, and in fact many prefer adults
MA
as sexual partners (Seto, 2008). It has been estimated that 25 to 50 % of convicted

sex offenders can be diagnosed as having pedophilia (Schaefer, Mundt et al., 2010).
ED

There are many reasons why this might be the case, including biological,

psychological, and developmental factors (Seto, 2008). For those who do not meet
PT

the necessary criteria to be diagnosed with pedophilia, it may be that they are

incapable of intimacy with adults. These individuals could be using a child as a

CE

pseudo adult, seeing them as a “safer option”, easier to engage and control, while
AC

being less likely to reject them (Navathe et al., 2013). This type of offender might

have problems relating to other adults, or have distorted beliefs concerning what is

appropriate sexual behavior. Other child sex offenders may be sexually pre-occupied,

using sex as a coping mechanism and/or lacking the inhibition necessary to direct

their sexual urges towards appropriate targets.

While sexual deviance is an influence, it is not simply sexual attraction

towards children that drives sexually abusive behavior. There is significant overlap

between core beliefs about sex and children, interpersonal problems, self-regulation

skills, and harmful sexual behavior. Offenders may be sexually attracted to children
 ACCEPTED MANUSCRIPT
15

and yet hold adaptive beliefs that help them to inhibit their problematic sexual

desires (Seto, 2008). On the other hand, some individuals are primarily attracted to

adults, but in certain states (e.g., intoxication) or situations (loss of a spouse)

PT
consider children as viable alternative sexual partners. While there must be some

degree of sexual arousal present during an offense, its actual causal role can vary

RI
depending on the individual’s situation and psychological features. Sexual arousal

SC
could be motivating behavior because of its intensity or because it is experienced in

high risk situations. Other possibilities are that sexual arousal in the presence of a

NU
child is mediated by deviant sexual preferences or instead arises from other

vulnerabilities such as unmet needs for intimacy or control. Thus, deviant sexual
MA
arousal can be simply a causal contributor or a primary cause of sexual offending,

and in addition, may reflect enduring propensities or temporary states.

ED

1.7. Summary: Risk

PT

This section has illustrated the composite nature of dynamic risk factors, and

underlined the need for researchers and practitioners to refer to etiological theory
CE

when deciding what particular causal factors are operating in specific instances of
AC

sexual offending. Each dynamic risk factor is really a summary of disparate features

and processes, internal and external. The focus in mainstream research has been on

developing categories of risk, and creating tools that predict reoffending better than

chance; although it is acknowledged that there is still room for significant

improvement (Beech & Craig, 2012). There is a significant amount of unexplained

variance when using any assessment tool, highlighting the fact that something is

missing from current risk conceptualizations (Hanson & Morton-Bourgon, 2009).

1.8 Protective factors

 ACCEPTED MANUSCRIPT
16

In recent years, there has been increasing interest in the role of protective

factors in promoting desistance and predicting recidivism alongside dynamic risk

variables (Thornton, 2013). The concept of protective factors has been imported from

PT
the child maltreatment area to forensic and correctional arenas. Child maltreatment

theorists have argued that it makes sense from therapeutic and policy perspectives to

RI
understand why certain individuals are less affected by traumatic or stressful

SC
experience while others go onto develop a number of psychological and social

problems. In the child maltreatment area, Afifi and MacMillan (2011) define

protective factors as:

NU
MA
“A protective factor may influence, modify, ameliorate, or alter how a person

responds to the adversity that places them at risk for maladaptive outcomes”
ED

(p.268).
PT

However, in the criminal justice domain, there has been a noticeable shift in
CE

the meaning of the concept of protective factors, especially in terms of what is

considered to be the protected object, the source of the potential harm, and the time
AC

frame concerned. The offender undergoing risk assessment is viewed as the source of

the harm and the protected object is the community or certain groups of potential

victims such as children or adult women. According to de Vries Robbé (2014),

protective factors are:

“factors that can compensate for a person’s risk factors” (p. 25) …………..

“characteristics of an offender, or alternatively, his environment or situation,

that reduce the risk of future violent behaviour” (p. 26).

 ACCEPTED MANUSCRIPT
17

Thus, protective factors in sex offender assessment are those variables that

reduce the chances of recidivism occurring. The interesting thing to note about this

definition is that the temporal scope of protective factors has been extended and any

PT
aspect of an offender’s characteristics or environment may be a protective factor as

long as it is risk reducing. The difficulty with broadening the range of the concept of

RI
protective factors so much is that it makes it harder to distinguish between protective

SC
factors, maturational effects, and therapy induced change or desistance events. Thus

it is no longer clear that it is protection from harm that is being referred to rather

NU
than simply the change process that reduces the possibility of harm occurring in the

future (see below).

MA
ED

2. Integrating Risk-related Factors, Values, and Agency

An exclusive focus on risk-related dimensions provides a description of

PT

vulnerability but cannot explain the purpose of behavior, arguably a necessary

CE

requirement for both explanation and intervention. A benefit of the upsurge of

research on risk is that there is greater awareness of the many different types of risk-
AC

related factors; although there is an absence of causal models capable of describing

how they interact and influence behavior. As stated above, risk-related factors are

composite constructs, which do not exist independently of the context of reoffending

prediction; they are predictive not explanatory concepts. Therefore, it is timely to

provide an agency-based model depicting the proximate mechanisms- those that are

temporally and causally linked to behavior- resulting in sexual offending (see Durrant

& Ward, 2015).

From a rehabilitation perspective, the Risk-Need-Responsivity Model

(Andrews & Bonta, 2010) struggles to accommodate the heterogeneity of child sex
 ACCEPTED MANUSCRIPT
18

offenders because it prioritizes statistically derived predictors over their real-world

experiences. On the other hand, the Good Lives Model (GLM; Ward & Maruna, 2007)

is a rehabilitation framework that is based on an agency conception of individuals,

PT
including their values and strengths. For this reason, arguably the GLM provides a

more useful rehabilitation framework within which to embed the agency model

RI
outlined below.

SC
When risk-related domains of functioning (particularly self-regulation) are

labeled as deficits or risk factors, there is a possibility of undermining individuals’

NU
perceived agency and thus responsibility for their actions. This is also a problem from

a research perspective as there is compelling evidence that all organisms are actively
MA
goal seeking, and in this respect, exercise agency (Christensen, 2012; Thompson,

2007 - see below for more detail on this issue). A first step in developing a causal
ED

model of risk is to acknowledge that, as human beings, child sex offenders behave in a
PT

goal directed manner- even if we disagree with their goals and strategies for achieving

them. Therefore, it is necessary to consider individuals’ identities, core beliefs, values,

CE

goals, and contexts when attempting to explain their sexually abusive behavior. It is

even more important to focus on offending behavior rather than lists of risk factors,
AC

which at best are proxies for causal mechanisms. Risk-related phenomena can only

be understood in relation to offending behavior, and behavior can only be understood

when we consider the goals that it is directed by, or more precisely, the outcomes it is

directed towards. The main question that will be addressed in this section is why

some individuals choose to sexually abuse children rather than seeking pro-social and

healthy sexual intimacy with adults.

 ACCEPTED MANUSCRIPT
19

2.1. The Good Lives Model

According to the GLM, offending is a result of obstacles that frustrate

individuals’ attempts to achieve personally meaningful outcomes or values. The GLM

PT
proposes that people share a common set of values, termed “primary human goods”,

which vary in their personal importance, but all are significant to some degree (Ward

RI
& Maruna, 2007). These primary goods are personal values, or aspects of life that

SC
individually and collectively contribute to physical, emotional, and psychological

wellbeing (e.g., mastery and relatedness). According to the GLM, in order to

NU
understand individuals’ antisocial actions it is necessary to view their behavior as

attempts to realize these values– formulated in a good life plan. These attempts or
MA
means are termed “instrumental” or “secondary goods”, and it is problems with these

strategies for goal attainment that cause offending (Ward, et al., 2006). It may be that
ED

the individual shifts towards the use of inappropriate means because he lacks the
PT

interpersonal skills necessary to achieve his goals in prosocial ways, for example,

sexual intimacy with an adult. Personal frustration could lead to selection of

CE

secondary goods (strategies) that encourage him to seek intimacy with a child

instead. In addition to the use of inappropriate means to secure primary goods, a

AC

good life plan may lack scope or contain conflicting goals.

Thus, primary human goods are normal and healthy, but difficulties arise

when they are translated into harmful secondary goods; the goals, strategies, and

implementation of a problematic good life plan. Furthermore, problems can arise in

the form of deviant or harmful goals or alternatively in the psychological processes

that influence the strategies used to achieve them. Therefore, risk is not simply the

presence of a series of dysfunctions, but rather the absence of the internal and

external resources necessary to live a fulfilling life in a healthy and prosocial way.

Explaining how values and their associated goals and strategies are translated into
 ACCEPTED MANUSCRIPT
20

maladaptive means arguably requires an agency-based model. The development of

the Agency Model of Risk (AMR) is an attempt to bridge the gap between theory and

practice, because it directly links dynamic risk factors with offending behavior.

PT
2.2. The Agency Model of Risk

The theoretical roots of the AMR are diverse and draw from dynamical

RI
systems theory (Christensen & Hooker, 2000), gene-culture co-evolution theory (e.g.,

SC
Odling-Smee, Laland, & Feldman, 2003), cognitive neuroscience (e.g., Maise, 2011),

philosophy (e.g., Sterelny, 2012), psychology of action (e.g., Huang & Bargh, 2014)

NU
criminology (e.g., Walsh & Bolen, 2012), and forensic psychology (Andrews & Bonta,
MA
2010; Ward & Maruna, 2007). In drawing from this body of theory and research, we

have formulated three general theoretical assumptions concerning the nature of

ED

human beings and the role of values in explaining human action. These assumptions

provide the overall theoretical justification for the AMR.

PT

2.2.1 Theoretical Assumptions of the AMR

CE

First, we argue that emergent materialism is a powerful and persuasive view of

AC

human functioning, with its commitment to multiple, non-reductive levels of analysis

and its insistence that psychological capacities (such as agency) play a crucial role in

promoting survival and flourishing. According to emergent materialism, there are

distinct levels of analysis corresponding to the different systems that constitute

human beings, and that sustain their functioning. There are ontological (i.e., the

nature of human beings and other organisms) and epistemological (i.e., knowledge

about human beings and other organisms) strands to this view. From an ontological

perspective, each system impacts on the others and yet has its own unique set of

processes that contribute to the overall functioning of the organism. Organisms are
 ACCEPTED MANUSCRIPT
21

dynamic, complex systems embedded within social, physical, and cultural

environments. From an epistemological viewpoint, knowledge concerning the

processes constituting each system level offers a valuable insight into the organism’s

PT
functioning. The salience of the level of explanation will in part depend on the

relevant question and therefore the problem at hand. Agency-level explanations of

RI
human interaction, including crime, with their reference to personhood and

SC
intentional mental states provide a unique and irreducible explanatory perspective.

Attention to individuals’ experiences, values, beliefs, and social/cultural contexts

NU
should be explicitly taken into account when developing theories of normative and

dysfunctional human behavior.

MA
Second, we acknowledge the pervasiveness of normativity in the natural world

as well as in human culture. The influence of values is evident in the norms that
ED

govern the functioning of different action sequences such as predator behavior,

PT

workings of biological systems, the application of human moral systems, and primate

social relationships. In speaking of the ubiquity and essential roles of norms in

CE

biological and social systems, Christensen (2012) views “normativity as inherent in

the organization or form of living systems, specifically in the form that generates their
AC

unity and hence explains their existence” (p.104). Normative principles are natural in

the sense that they specify the functional parameters of biological systems and social

practices, and have their origin in the development of agency in organisms of all

types. Norms are reflected in goals, and the strategies selected to further these goals

are evaluated against these norms, typically in a fluid, dynamic, and immediate

manner, in “real time”. In complex animals, such as human beings, the capacity to

flexibly adjust goal directed strategies and plans in response to changing

environmental contingencies is in part due to cognitive capacity, and to the

availability of social and cultural resources (Sterelny, 2012).

 ACCEPTED MANUSCRIPT
22

Third, internal psychological processes, such as emotions, cognition, drives,

and needs, underpin goals and can influence subsequent actions unconsciously or

consciously (Haung & Bargh, 2014). Goals are activated or selected in response to

PT
external contexts and their cues such as the presence of prey or threats, and by

internal cues such as hunger, fear, sexual desire, or anger. In describing these cues

RI
Christensen and Hooker (2000) state:

SC
“We shall term the signals which a system uses to differentiate an appropriate

NU
context for performing action the system’s explicit norm signals. For example,

hunger signals differentiate blood sugar levels and act to initiate and focus food
MA
search activity. Explicit norm signals provide information about appropriate

action because they differentiate more and less discrepancy between some
ED

current system condition and a reference condition, the norm satisfaction state,
PT

which modulates subsequent performance”. (p11)

CE

Emotions, hunger, and other motivational states function as signals that

organisms needs to deal with relevant challenges, whether they are threats,
AC

opportunities for rewards, or physical needs. One of the notable features of highly

complex organisms such as primates (and especially humans) is that they have to

balance multiple goals at any one time and construct plans and strategies intended to

realize them in order to guide action (Huang & Bargh, 2014). Such models enable the

organism to anticipate or predict environmental outcomes, and the norms (which set

standards of performance and specify desired outcomes) associated with their

constituent goals help to correct action if there is a discrepancy between these and

outcomes (Christensen, 2012). Higher order cognitive systems, language, and

normative codes expressed within language are capable of creating more abstract,
 ACCEPTED MANUSCRIPT
23

symbolic cues that can activate goals and their subsequent plans. Furthermore,

adaptations such as social learning (via modeling, imitation etc.), can facilitate this

process and create external contexts that scaffold the acquisition of complex skills.

PT
For example, in a criminal justice context, individuals sometimes acquire offending

skills from fellow offenders and through reflection on their own offending

RI
experiences (Bourke, Ward, & Rose, 2012).

SC
2.2.2 Dynamic Risk and Protective Factors, and Agency

NU
Dynamic risk factors in the correctional and forensic domains are intended to

predict harm related to reoffending, typically to victims and the community.

MA
Protective factors are features that lessen the chances of risk factors having this

effect, or more generally, if present they reduce the likelihood of offending occurring.
ED

It is a good idea to start with the personal and contextual factors that both dynamic
PT

risk factors and protective factors refer to. In our view, they refer to the components

of agency: goals, plans and strategies, their implementation and evaluation, and the
CE

subsequent revision of goals and plans. Furthermore, in our model there are three

levels of agency, each associated with its own distinct set of goals, plans and
AC

strategies, and each capable of influencing the others types of agency (see below):

systems-level, social role, and personal identity. As proposed earlier, dynamic risk

factors do not exist other than as summaries of different variables, and because

protective factors are defined in terms of harmful (risk associated) outcomes, they do

have not conceptual meaning aside from this context.

Thus, risk factors can be viewed as psychological and social processes (i.e.,

those associated with goals, plans, strategies, and action implementation) that impair

normal functioning and hence disrupt persons’ internal and external relationships to

their social, cultural, and physical environments. This disruption can be at multiple
 ACCEPTED MANUSCRIPT
24

levels and may even be confined to incorrect actions within a single practice (e.g.,

relationship repair). Protective factors, once broken down into their core elements,

work in multiple ways across the various levels of agency to inhibit and/or disrupt

PT
dysfunctional systems, and to restore normal functioning. Sometimes the constraints

exerted by protective factors may be external such as the construction of supportive

RI
social networks around high-risk offenders. Strengths are conceptualized as internal

SC
and external capacities that enable offenders to flourish in certain environments,

which can be used also to motivate them, and which can also be recruited to function

as protective factors in some situations.

NU
Agency may be defined in terms of an individual’s capacity to act in a goal
MA
directed manner with respect to both internal states and external situations,

including interacting with social and physical environments (Ward & Durrant, 2015).
ED

Figure 1 depicts the Agency Model of Risk (AMR), a dynamic interactional model,
PT

which integrates motivational states, contexts, and goal-directed behavior. This

model represents a conceptual shift from investigating statistical correlates of

CE

recidivism to reasoning about the psychological and behavioral processes associated

with individuals’ goal directed, offending behavior. The AMR provides a conceptual
AC

framework within which the relationship between risk-related factors and behavior

can be understood. It is compatible with the GLM as it prioritizes individual agency,

values, goals, and contextual features when attempting to explain sexually abusive

actions. The individual conceptual components of the AMR will now be described.

----------------------------

Place figure 1 about here

------------------------------
 ACCEPTED MANUSCRIPT
25

2.2.3. Levels of agency

As stated above, there are three levels of agency, or agent level descriptors,

within the AMR. A level of agency is associated with a specific set of psychological

PT
and behavioral processes that arise in response to particular clusters of internal and

external cues. That is, each agency level is triggered by unique environmental factors

RI
and associated emotions, such as physical threats or interpersonal rejection. The

SC
three levels are not necessarily comprehensive, nor are they mutually exclusive.

However, one level of agency may be most relevant for a particular value or goal, or

NU
dominant within a particular context. The first level of agency is an individual’s self-

concept or personal identity, essentially what he believes about himself and his
MA
purpose in life. This level of agent description is concerned with formulating a good

life plan, and is heavily influenced by individuals’ beliefs, personal values, and life
ED

priorities.
PT

The social role agency level concerns the self in relation to others from the

perspective of an institutional role, for example parent, partner, teacher, or even

CE

social outcast. Much like personal identity, social roles depend on persons’ self-

concepts, but are also related to other peoples’ actions and perceptions of them. The
AC

social roles incorporate social and professional responsibilities, skills and capacities,

and also reflect what people place importance on. For example, if individuals value

knowledge and mastery they may wish to pass this on through teaching. The final

agency level is the systems-level, in which physical or biological needs are salient.

This level is especially relevant when physical states that influence offending, such as

sexual arousal or altered consciousness (e.g., intoxication), are present. For example,

if the systems-level is activated by sexual arousal in the presence of a victim, the

individual may be motivated to offend in order to experience pleasure. This may

occur even if a person would not normally offend, for example, when his self-
 ACCEPTED MANUSCRIPT
26

regulation capacity is compromised by intoxication or emotional stress. The three

levels of agency directly influence which values are salient at any time. When each

level of agency is activated, their associated values are translated into concrete goals

PT
that, explicitly or implicitly, direct individuals’ subsequent actions (Durrant & Ward,

2015).

RI
2.2.4. Goals

SC
Values, or broad categories of primary human goods, must first be translated

into specific goals in order to effectively direct behavior. These goals are the purpose

NU
of action sequences, and consist of representations of tangible events or outcomes
MA
that the individual (correctly or incorrectly) believes will lead to attainment of the

desired value. Goals are informed by values and beliefs: what an individual wants and
ED

how he believes he can achieve his desires. For example, an individual who highly

values pleasure might believe that the only way to achieve this is by having sex with
PT

children. The goal may be to form a trusting relationship with children, and then

sexually abuse them. The strategies used may be “successful”; but the outcome is
CE

harmful and maladaptive. The goal is flawed because it is underpinned by deviant

AC

sexual interests, distorted beliefs about children and sex, and/or interpersonal

difficulties with peers. Sexual offending is simultaneously motivated by goals

informed by core beliefs, personal identity, interpersonal functioning, and sexual

needs. Once a goal has been formulated, individuals deliberate about how best to

achieve it within their current context.

2.2.5. Strategies for goal attainment

During the process of planning, skills (or at least perceived skills), such as

interpersonal capacities, become relevant. This involves thinking about what sort of

actions will be most successful in achieving a person’s goals. Like other aspects of
 ACCEPTED MANUSCRIPT
27

agency, this task is informed by beliefs and expectations, such as ascertaining

whether or not other people are dangerous or accepting, or if the person believes he is

capable of achieving his goals. Successful planning is highly dependent upon a suite

PT
of knowledge and skills, such as a theory of mind (Ward & Keenan, 1999). If

individuals do not possess the ability to reason about others mental states it will be

RI
difficult to meet their interpersonal goals. Planning also involves cognitive skills, such

SC
as mental time travel (essentially counter factual thinking about possible future and

past events), and the use of scripts and action templates (Ward & Durrant, 2015).

NU
These skills are refined over time and can become deeply ingrained in long-term

memory; at this point, offending action-sequences are frequently implemented

MA
relatively automatically (Ward & Durrant, 2015). An individual may not even be

consciously aware that his actions are directed towards harmful goals (Smallbone &
ED

Wortley, 2004). Offense-related action sequences thus can become a part of persons’
PT

behavioral repertoires, particularly when experienced as successful in achieving

valued outcomes.
CE

2.2.6. Implementation: Behavioral outcomes and self-monitoring

AC

The implementation phase refers to offending, and subsequent outcomes and

appraisals. A person inquires into the success of his actions and whether or not he

should devise a more effective strategy to achieve his goals. Implementation is

seamlessly integrated with the other two components of the action sequence (Ward &

Durrant, 2015). As well as the standard action progression, an offender may form

new goals during strategy formation or in the implementation and evaluation phases,

and unsuccessful strategies might be abandoned or refined. Therefore,

implementation is reliant on the goal-directed strategies formed prior to offending,

but is also influenced by triggers, including the context and the individual’s emotional
 ACCEPTED MANUSCRIPT
28

state. As stated earlier, emotions function as internal cues that allow individuals to

implicitly monitor the progress of their goal-directed actions (Christensen & Hooker,

2000). Effective action-sequences are repeated and can form long-lasting cognitive

PT
and behavioral scripts, guiding subsequent actions in relation to external conditions.

They are “ecologically sensitive” (Ward & Durrant, 2015) and depend on existing

RI
internal and external conditions, what have traditionally been considered acute

SC
dynamic risk factors.

After an initial offense has occurred, other values may be implicated in the

NU
subsequent sequence of offending related actions. For example, in addition to seeking

pleasure, a child sex offender may start to value experiencing mastery of the skills
MA
gained through abusing children (e.g., grooming). If there are insufficient skills or

opportunities for prosocial mastery, he may rely on offending as a behavioral strategy

ED

to realize this value. Therefore, the same strategies (e.g., sex with a child) are
PT

employed to reach different values and their associated goals (e.g., mastery or

pleasure). Unsuccessful strategies directed towards prosocial goals could cause a shift
CE

towards offending behavior (e.g., substituting a child for an adult partner). This

highlights the bi-directional nature of the model: goals and strategies lead to
AC

implementation, but they are also influenced by self-monitoring and action-

appraisal.

2.2.7. Context

When attempting to explain sexual offending and its relationship to dynamic

risk factors, it is crucial to consider contextual opportunities and constraints. Social

and physical environments provide access to other people and situations that are

instrumental in meeting goals, but they may also fail to provide the external

resources necessary for desirable outcomes. For example, if individuals do not have
 ACCEPTED MANUSCRIPT
29

access to appropriate partners it is likely that they will modify their strategies to meet

their needs in more harmful ways– even if the original goal was intimacy with an

adult. Successful agency involves thinking about how best to integrate goals,

PT
strategies, and norms within a coherent action sequence (Ward & Durrant, 2015).

Norms are essentially beliefs and standards that are shared by the community; for

RI
example, that sex between children and adults is inappropriate. This is a widely

SC
accepted belief, reflected in most societies’ laws, and based on moral ideas about the

harm and coercion involved in this sort of crime. Thus an important question is why

NU
an individual would choose to transgress these norms, and commit acts that others

find abhorrent. Norms are part of an individual’s cultural and social context and in
MA
this respect, are interpersonal and relational.

In some cases, social contexts or sub-group norms actively support sexual

ED

offending. For example, networks of individuals with pedophilic interests who share
PT

resources and sets of ideas justifying sexual abuse often initiate and maintain

offending in their members. These sub-groups support distorted beliefs about

CE

children, sex, and the world, and reject widely held norms concerning what

constitutes acceptable sexual behavior. In a sense, sex offenders benefit from

AC

communicating and interacting with others that support, rather than challenge, their

beliefs and attitudes. In some instances, this is their only means of achieving

connectedness or belonging to a group. This social connection is likely to consolidate

offense-supportive beliefs, strengthen deviant sexual preferences, and in some

instances, even offer additional opportunities for offending.

The relationship between agency and context is a reciprocal one. While

seeking out or creating environments that align with priorities and preferences,

persons are also heavily influenced by these contexts. Environments can trigger

underlying motivations, or in some cases, actually cause them directly (Ward &
 ACCEPTED MANUSCRIPT
30

Durrant, 2015). A final key point is that both the agent and context must, to some

degree, support offending. While one factor could appear to contribute most, neither

on their own is sufficient to cause a sexual offense. An individual must have offense-

PT
related personal vulnerabilities or propensities, and also make use of external

resources in order to achieve offense-related goals. From the perspective of the AMR,

RI
sexual offending is a relational phenomenon, and as such, dependent upon

SC
individuals’ vulnerabilities and the social contexts within which they are embedded.

2.3. Agency Model of Risk summary

NU
In summary, the AMR arguably adds value to traditional accounts of dynamic
MA
risk factors by prioritizing features that are missing in contemporary theory and

practice: values, motivation, and agency. Viewing risk factors within the context of
ED

the AMR offers a more individualized and person centered explanation of how they

relate to offending, rather than assuming that because they are associated with
PT

sexual abuse they are automatically causal factors. The actual causal links between

dynamic risk factors and sexual offending are much more complex and nuanced than
CE

portrayed in standard practice. The following section will provide examples of how
AC

the AMR can account for various combinations of risk-related factors. This should

make the above rather abstract description more concrete and make it clearer how

the AMR can inform treatment with sex offenders.

3. The Agency Model of Risk Applied

Throughout this article risk-related phenomena have been deconstructed and

reconceptualized as aspects of a sex offender’s agency and the contexts within which

offending takes place. The AMR identifies individual agency and motivational factors

as underlying mechanisms that, when combined with psychological vulnerability and

 ACCEPTED MANUSCRIPT
31

risky contexts, cause sexual offending. According to the AMR, these vulnerabilities

(i.e., risk factors) are proximate mechanisms that influence an individual’s ability to

act in a prosocial and goal-directed manner. The four domains of risk outlined above

PT
will now be explored using the AMR (see Figure 2). Because each of these broad risk

categories is a composite construct, its facets are distributed across the various

RI
components of the AMR. In addition, it will become clear that each risk-related

SC
domain also contains aspects of each of the others.

------------------------------

NU
Place figure 2 about here

------------------------------
MA
3.1. Intimacy deficits
ED

The dynamic risk category of intimacy deficits indicates problems in achieving

interpersonal goods, such as relatedness and community. In the example depicted in

PT

Figure 2, difficulties in childhood attachment and subsequent rejection have

contributed to the offender’s current personal identity. He sees himself as vulnerable,

CE

entitled to sex, and misunderstood. He emotionally identifies with children rather

AC

than adults, and believes that relationships with children are mutually beneficial. The

emotional states of loneliness and unhappiness are caused by interpersonal conflict

and rejection. These states result in the formation of an initial goal of “intimacy with

others”, but because of the individual’s interpersonal difficulties and incapacity for

emotional intimacy with adults, the goal shifts to “intimacy with a child”. Contextual

factors contributing to this goal include limited access to other adults and a lack of

intimacy in existing social relationships. The offender is able to gain unsupervised

access to children due to his role as a primary school teacher. The strategies he selects

to achieve this goal are dependent on his interpersonal skills and cultural norms. He
 ACCEPTED MANUSCRIPT
32

thinks about how best to gain the child’s trust and what he needs to do to conceal or

justify his behavior. The actual implementation and evaluation of the individual’s

offense related plan relies on the physical context in which the offending takes place

PT
(school room), the victim’s reactions, and his capacity to regulate his feelings and

behavior during the offense process.

RI
3.2. General self-regulation problems

SC
The dynamic risk category, general self-regulation problems, contributes to

difficulties in achieving values such as inner peace (i.e., freedom from distress) and

NU
agency. In order to experience emotional equilibrium, a person must use coping
MA
strategies that are appropriate to specific personal and environmental challenges. He

may lack the ability to successfully act in a goal-directed way due to inadequate
ED

problem-solving skills. For example, he could act impulsively without considering

consequences or abuse alcohol (as a coping strategy) and only offend while
PT

intoxicated. The offender’s inhibitions could be temporarily suspended, leading him

to act impulsively without formulating conscious goals and strategies. His offending
CE

might be reactive and closely related to emotional and physical states (i.e., the
AC

systems-level of agency), rather than as a result of explicit planning. In such cases,

problems exist within the action-sequence, and implementation occurs before goals

and strategies have been fully considered. On the other hand, sexual offending may

be instrumental in that an individual explicitly aims to achieve specific goals such as

relatedness, pleasure or mastery. The offender depicted in Figure 2 appears to have

formulated a coherent, well thought out action plan; the problem is that his goal is

harmful to children.
 ACCEPTED MANUSCRIPT
33

3.3. Cognitive distortions

The category of dynamic risk labeled cognitive distortions, or offense-

supportive attitudes refers to problems in achieving values, such as (accurate)

PT
knowledge and (pro-social) mastery. Achieving these states requires prosocial and

adaptive beliefs that are shared with others; offense-supportive beliefs are

RI
maladaptive and transgress the norms of society. Core beliefs about the self, others

SC
(including victims), and the environment determine the goals people select and the

way they interact with the world. The individual depicted in Figure 2 believes he is

NU
vulnerable and misunderstood; other adults are viewed as cold and rejecting, and the

world seen as a dangerous place. By way of contrast, children are regarded as safe
MA
and accepting, capable of consenting to an emotional and sexual relationship, and

unharmed by sexual contact. These beliefs influence the goals he selects, the
ED

strategies and plans formed, and their implementation. In general terms, once an
PT

individual has decided on the most successful action-sequence, he relies on his

environment to support and strengthen his beliefs, for example, turning to offense-
CE

supportive associates or using pornographic material to both arouse himself and

rehearse his offending.

AC

3.4. Sexual deviance

Generally speaking, the dynamic risk category of sexual deviance is evident

across the action-sequence. It is activated by the systems-level of agency, and is often

associated with the seeking of pleasure. Sexual deviance heavily influences goal and

strategy formulation, as it determines what is experienced as pleasurable, and how

this is sought. In the context of sexual offending, deviance typically refers to

pedophilic interests, or sexual attraction to children rather than to adults. In other

 ACCEPTED MANUSCRIPT
34

cases, sexual deviance can include problematic sexual practices, including sexual pre-

occupation, promiscuity, and impersonal sexual scripts. These problems can cause

sexual arousal in inappropriate situations, such as when children are present. The

PT
individual depicted in Figure 2 is motivated in part by sexual arousal and loneliness,

which have influenced his goal selection- sex with a child. This goal is supported by a

RI
maladaptive belief that children are sexual beings, and leads to strategies targeting

SC
them for grooming and sexual behavior. Sexually deviant beliefs and action-

sequences are supported and maintained by external resources (e.g., child

pornography).
NU
MA
3.5. Summary: The Agency Model of Risk applied

This section has sought to demonstrate how the AMR can provide a theoretical
ED

framework within which to understand the mechanisms underlying risk-related

phenomena. In the AMR, the composite constructs currently relied on to predict and
PT

explain risk, are broken down into aspects of human agency. Risk-related factors gain

meaning when viewed as the presence or absence of the resources required for
CE

successful agency. In combination with values and motivational states, they influence
AC

the goals offenders prioritize and the way they attempt to achieve these. Once the

normative aspects of human motivation and agency are factored in it is possible to

explain the relationship between statistical predictors and offending behavior. Risk

prediction tools work because they tap into aspects of the cognitive and behavioral

processes (action-sequences) that cause offending. However, they do not

acknowledge the roles of human agency and values that motivate all behavior,

offense-related or otherwise. Arguably this is not their purpose, but an unbalanced

picture of risk lacks the ability to effectively inform the management and treatment of

child sex offenders.

 ACCEPTED MANUSCRIPT
35

4. Evaluation and Conclusions

We will now briefly evaluate the potential research and practice contributions of

the AMR. This section will address two questions, 1) What can the AMR offer the

PT
contemporary theoretical literature on the conceptualization of dynamic risk factors?

2) How can this model guide correctional practice?

RI
4.1. Theoretical and research implications

SC
The key purpose in developing the AMR is to shift research and practice

attention to offending behavior rather than its correlates, and to broaden the scope of

NU
risk to include the normative aspects of human functioning. As the AMR is focused
MA
on proximate mechanisms, it has a greater chance of identifying real causal factors,

thereby clarifying the relationship between dynamic risk factors and the onset or
ED

recurrence of sexual offending. One of the AMR’s core assumptions is that all humans

act in a goal-directed manner, which aligns with the GLM’s conceptualization of

PT

offending as instrumental (but maladaptive) in meeting human needs, and is

consistent with contemporary biological models of agency (Christensen & Hooker,

CE

2000). The AMR encourages a broader approach to assessment, and is consistent

AC

with the use of the GLM as a case formulation and treatment framework. In addition,

the AMR can explain behavior according to multiple domains of functioning. While

providing an agency level of explanation for offending behavior, relying on

motivation, values, and beliefs, the AMR also incorporates biological, emotional, and

cognitive variables, as well as interpersonal and cultural contexts. Therefore, it is

arguably able to capture the uniqueness of human experience while taking into

account the domains of risk identified by researchers.

The broad scope and dynamic nature of the AMR are its major strengths. Each

aspect has briefly been discussed, but it is acknowledged that as it is only a

 ACCEPTED MANUSCRIPT
36

preliminary model, significant theoretical work remains. Certainly, it is a further step

towards a theoretical, empirical, and practical focus on the patterns of cognition and

behavior that result in offending. It is recommended that researchers test the utility

PT
of the AMR as a model for explaining sexual offending behavior. This may occur in a

treatment context with individuals, or by considering a number of suggested

RI
typologies from the literature. Investigating the utility of the AMR in practical

SC
contexts is likely to enhance practitioners’ and offenders’ understanding of risk and

its component factors.

4.2. The Agency Model of Risk in practice

NU
MA
Practitioners routinely use professional judgment during structured risk

assessment. However, they do not always demonstrate a robust understanding of the

ED

research and theory underpinning assessment tool items. For this reason, it is

important to link psychological theory and practice, by providing a model that

PT

integrates statistically derived risk factors, psychological theory, and criminal

behavior. The AMR is a behavioral model that can guide professional judgment of
CE

risk, explain why harmful behavior has occurred, and facilitate risk management or
AC

reduction. Therefore, a multi-systemic approach to assessment that relies on the

concepts of the AMR has the potential to add value to traditional methods. For

example, following risk assessment, the AMR can be used to explain how risk-related

factors influenced an individual’s sexual offending and associated behavior. This is

possible because the process of goal formulation, planning, and implementation

(offending) is dependent on the offense-supportive beliefs, capacities, skills, and

preferences used for risk prediction. Risk-related factors and values become pieces of

an explanatory puzzle, together forming a more comprehensive picture of an

individual’s motivation for, and vulnerability to commit, sexual offending. Protective

 ACCEPTED MANUSCRIPT
37

factors are useful in indicating the psychological and environmental resources

necessary to support desistance and successful social integration.

The AMR prioritizes individual patterns of offending behavior, rather than

PT
merely assigning numerical values to characteristics shared with other individuals

who commit sexual offenses. This suggests that it will be more meaningful for

RI
individuals, as it has the potential to enhance their understanding of their own

SC
behavior and triggers for offending. Comprehending human values, and the goals and

strategies they are translated into, is more likely to help uncover the purpose of

NU
sexual offending- for both individuals who commit offenses and others working with

them. Practitioners can utilize personal accounts of offending (i.e., offense-chain

MA
analyses) to discuss individuals’ goals, how they have planned to achieve these, and

how this led to sexual offenses. For example, we suggest that a therapist first consider
ED

individuals levels of agency: What is their personal identity? How do they see
PT

themselves in relation to others? What physical needs are they trying to meet? Once

individuals’ most important values and motivational states initiating goal-directed

CE

action are identified, it is easier to infer the relevant underlying causal mechanisms

and ascertain how best to change their sexually abusive behavior and any related
AC

problems. An AMR approach to rehabilitation prioritizes building internal and

external resources and competencies that assist pro-social agency, rather than

reducing risk correlates.

5. Overall conclusions

The recent preoccupation with prediction and dynamic risk factors has arguably

left the field of criminal psychology theoretically stagnant. The reliance on statistical

correlates of recidivism as explanations for sexual offending is at best misleading, and

at worst has the potential to undermine human agency. Perhaps the most
 ACCEPTED MANUSCRIPT
38

controversial claim of this article is that risk and protective factors do not exist as real

causal entities. They are composite constructs, indicative of problems with agency:

the absence of resources required to act successfully and prosocially towards

PT
achieving valued goods. The AMR accommodates these proximate causes of sexual

offending, and depicts how they consist of motivational (agent), behavioral (action-

RI
sequences), and contextual features. It demonstrates that statistically derived

SC
predictors are deficient in their ability to explain why individuals choose to sexually

offend; they are just one piece of the etiological puzzle.

NU
Risk assessment is reasonably successful because risk factors refer to problems

that are correlated with sexual offending behavior. While currently adequate for
MA
prediction, this conceptualization of dynamic risk factors is not sufficient to explain

offending behavior or to guide interventions to change it. A shift in focus towards

ED

agency conceptions will hopefully promote more individualized pictures of risk in

PT

relation to individuals’ values, capacities, and contexts. The AMR integrates a range

of empirically and theoretically identified features (as well as others specific to the
CE

individual) together within a behavioral model that explains why these features are

problematic within certain contexts. It directs attention back to the individual, and
AC

asks clinicians and researchers to consider whether or not certain characteristics

contribute to offending or facilitate desistence. It provides a conceptual bridge

between etiological theory, risk prediction and treatment, and in this respect, pulls

the various strands of sex offender practice together.

 ACCEPTED MANUSCRIPT
39

Table 1

Child Sex Offender Dynamic Risk Domains

Cognitive Emotional Interpersonal Sexual
Cognitive Skill Deficits Poor Self-Management Social Skill Deficits Sexual Pre-occupation
Cannot identify/solve Risky/impulsive Comfortable around Promiscuity

PT
problems behavior (generally) children Excessive use of
Rigid thinking Substance abuse Incapacity for intimacy pornography/masturbation
Cannot describe goals Uses sex/masturbation with adults High sex drive
Cannot predict negative to cope Immature relationships Loses control when

RI
consequences Non-compliance Aggression sexually aroused

Negative Affect Intimacy Deficits

SC
Offense-Supportive Hostility towards No long-term/co- Sexual Deviance
Beliefs women inhabiting partners Views child pornography
Believes that sex with Lack of remorse No close adult Unusual/harmful interests
children is ok Blames others relationships Paedophilia

NU
Self as uncontrollable Negative rumination Children friends/peers Distorted sexual scripts
and entitled to sex Conflicted desires Social isolation Wide range of previous
Dangerous world sexual offenses
MA
ED
PT
CE
AC
 ACCEPTED MANUSCRIPT
40

PT
RI
SC
NU
MA
Figure 1. The Agency Model of Risk.
ED
PT
CE
AC
 ACCEPTED MANUSCRIPT
41

PT
RI
SC
NU
MA

Figure 2. The Agency Model of Risk: An example of a child sex offender.

ED
PT
CE
AC
 ACCEPTED MANUSCRIPT
42

6. References

Afifi, A. O., & MacMillan, H. L. (2011). Resilience following child maltreatment: A

PT
review of protective factors. Canadian Journal of Psychiatry, 25, 266-272.

Andrews, D. A., & Bonta, J. (2010). The Psychology of Criminal Conduct. New

RI
Providence, NJ: Anderson Publishing.

SC
Beech, A. R., & Craig, L. (2012). The current status of static and dynamic factors in

sexual offending risk assessment. Journal of Aggression and Peace Research,

NU
4, 169-185. DOI 10.1108/17596591211270671

Beech, A. R., & Mitchell, I. J. (2005). A neurobiological perspective on attachment

MA
problems in sexual offenders and the role of selective serotonin re-uptake

inhibitors in the treatment of such problems. Clinical Psychology Review, 25,

ED

153-182. doi:10.1016/j.cpr.2004.10.002.
PT

Beech, A. R., & Ward, T. (2004). The integration of etiology and risk in sexual

offenders: A theoretical framework. Aggression and Violent Behavior, 10, 31-

CE

63. doi:10.1016/j.avb.2003.08.002.

Bourke, P., Ward, T., & Rose, C. (2012). Expertise and sexual offending: A
AC

preliminary empirical model. Journal of Interpersonal Violence, 27, 2391-

2414.

Christensen, W. (2012). Natural sources of normativity. Studies in the History and

Philosophy of Biological and Biomedical Sciences, 43, 104-112.

doi:10.1016/j.shpsc.2011.05.009

Christensen, W. D. & Hooker, C. A. (2000). An interactivist-constructivist approach:

to intelligence: Self-directive anticipative learning. Philosophical Psychology,

13, 5-45. doi: 10.1080/09515080050002717

 ACCEPTED MANUSCRIPT
43

de Vries Robbé, M. (2014). Protective Factors: Validation of the structured

assessment of protective factors for violence risk in forensic psychiatry

(Doctoral thesis, Radboud University, Nijmegen, Netherlands). Retrieved from

PT
http://www.saprof.com/_files/Phd%20Thesis%20Protective%20Factors.%20

Michiel%20de%20Vries%20Robbe.pdf

RI
Durrant, R., & Ward, T. (2015). Evolutionary Criminology: Towards a

SC
Comprehensive Explanation of Crime and its Management. San Diego, CA:

Academic Press.

NU
Hanson, K., & Harris, (2000). Where should we intervene? Dynamic predictors of

sexual offense recidivism. Criminal Justice and Behaviour, 27, 2-35. doi:
MA
10.1177/0093854800027001002.

Hanson, R. K., & Morton-Bourgon, K. (2009). The accuracy of recidivism risk

ED

assessments for sexual offenders: A meta-analysis of 118 prediction studies.

PT

Psychological Assessment, 21, 1-21. doi: 10.1037/a0014421.

Haung, J. Y., & Bargh, J. (2014). The selfish goal: Autonomously operating
CE

motivational structures as the proximate cause of human judgment and

behavior. Behavioral and Brain Sciences, 37, 121-134.

AC

doi:10.1017/S0140525X13000290

Illari, P., & Russo. F. (2014). Causality: Philosophical Theory Meets Scientific

Practice. New York: Oxford University Press.

Maise, M. (2011). Embodiment, emotion, and cognition. Basingstoke, UK: Palgrave

Macmillan.

Mann, R. E., Hanson, R. K., & Thornton, D. (2010). Assessing risk for sexual

recidivism: Some proposals on the nature of psychologically meaningful risk

factors. Sexual Abuse: A Journal of Research and Treatment, 22, 191-217. doi:

10.1177/1079063210366039.
 ACCEPTED MANUSCRIPT
44

Navathe, S., Ward, T., & Rose, C. (2013). The development of the Sex Offender

Relationship Frames Model. Psychiatry, Psychology and Law, 20, 60-72, doi:

10.1080/13218719.2011.619061.

PT
Ó Ciardha, C., & Ward, T. (2013). Theories of cognitive distortions in sexual offending

what the current research tells us. Trauma, Violence, & Abuse, 14, 5-21. doi:

RI
10.1177/1524838012467856.

SC
Odling-Smee, F. J., Laland, K. N., & Feldman, M. W. (2003). Niche construction: The

neglected process in evolution. Princeton, NJ: Princeton University Press.

NU
Schaefer, G. A., Mundt, I. A., Feelgood, S., Hupp, E., Neutze, J., Ahlers, Ch. J.,

Goecker, D., Beier, K. M. (2010). Potential and dunkelfeld offenders: Two

MA
neglected target groups for prevention of child sexual abuse. International

Journal of Law & Psychiatry, 33, 154–163. doi: 10.1016/j.ijlp.2010.03.005.

ED

Seto, M. C. (2008). Pedophilia: Psychopathology & Theory. In D. R. Laws & W. T.

PT

O'Donohue (Eds.), Sexual Deviance (pp. 64-82). New York: The Guildford

Press.
CE

Smallbone, S. W., & Wortley, R. K. (2004). Onset, persistence, and versatility of

offending among adult males convicted of sexual offenses against children.

AC

Sexual Abuse: A Journal of Research and Treatment, 16, 285-298.

Sterelny, K. (2012). The evolved apprentice: How evolution made humans unique.

Cambridge, MA: MIT Press.

Thompson, E. (2007). Mind in life: Biology, phenomenology, and the sciences of

the mind. MA, USA: Belknap-Harvard.

Thornton, D. (2013). Implications of our developing understanding of risk and

protective factors in the treatment of adult male sexual offenders. International

Journal of Behavioral Consultation and Therapy, 8, 62-65.

 ACCEPTED MANUSCRIPT
45

Walsh, A., & Bolen, D. (2012). The neurobiology of criminal behaviour: Gene-brain-

culture interaction. Burlington, VT: Ashgate.

Ward, T. (2014). The explanation of sexual offending: From single factor theories to

PT
integrative pluralism. Journal of Sexual Aggression, 20, 130-141. doi:

10.1080/13552600.2013.870242.

RI
Ward, T., & Beech, A. R. (2015). Dynamic risk factors: A theoretical dead-end?

SC
Psychology, Crime & Law, 21, 100-113. doi: 10.1080/1068316X.2014.917854.

Ward, T., & Casey, A. (2010). Extending the mind into the world: A new theory of

NU
cognitive distortions in sex offenders. Aggression and Violent Behavior, 15, 49-

58. doi: 10.1016/j.avb.2009.08.002.

MA
Ward, T., & Hudson, S. M. (1998). A model of the relapse process in sexual offenders.

Journal of Interpersonal Violence, 13, 700-725. doi:

ED

10.1177/088626098013006003.
PT

Ward, T. & Keenan, T. (1999). Child molesters’ implicit theories. Journal of

Interpersonal Violence, 14, 821-838. doi: 10.1177/088626099014008003.

CE

Ward, T., & Maruna, S. (2007). Rehabilitation: Beyond the risk paradigm. New

York: Routledge.
AC

Ward, T., Polaschek, D., & Beech, A. R. (2006). Theories of sexual offending.

Chichester, UK: John Wiley.