Keywords

Introduction

There has been a lot of ink spilt over the last 12 years or so concerning the comparative merits of the Risk-Need-Responsivity (RNR) and the Good Lives Models (GLM) of offender rehabilitation (e.g., Andrews, Bonta, & Wormith, 2011; Ward, Yates, & Willis, 2012). The proponents of the RNR) and ) GLM have critically engaged each other along a number of theoretical, empirical, ethical, and practice dimensions, (a) typically finding fault with their critics’ formulation of their own model and (b) pointing to putative conceptual confusions and logical flaws in the other’s model. The trouble is that while much heat has been generated in this debate, there has been little progress in developing an integrated approach to offender rehabilitation that incorporates the best from the GLM and RNR. Relatedly, there has been a notable lack of any real understanding of exactly what are the core differences between the two approaches and if in fact they amount to anything of theoretical or practice significance. In our view, writing yet another paper that compares the core values, assumptions, and practice implications of the two models in a comprehensive way is unlikely to change the repetitive and somewhat acrimonious nature of the debate. It certainly will not move the field further forward, which is a pity as theoretical and practice innovation is sorely needed in the correctional and sexual offending fields (Ward, 2014). Furthermore, there have been comprehensive recent summaries of both the RNR and the GLM in the general correctional and sexual offending literature that can be consulted by interested readers (e.g., Andrews & Bonta, 2010a, 2010b; Laws & Ward, 2011; Thornton, 2013; Willis, Ward, & Levenson, 2014; Yates, Prescott, & Ward, 2010). We do not intend to provide such a summary here.

In our opinion, there are important theoretical differences between the RNR and the GLM that would profit from sustained analysis and which have been somewhat neglected so far. This is a pity as at least one of the neglected issues may be key in understanding what we have got right so far in correctional treatment, why we have become side tracked, and where we need to go. The topic we refer to is the nature of the need principle and its associated concepts of dynamic risk factors, risk prediction and management, and causal explanation. According to Andrews and Bonta (2010a), the need principle proposes that potentially changeable variables associated with reductions in recidivism (i.e., dynamic risk factors or criminogenic needs) should be targeted in treatment to create safer communities. The need principle in conjunction with the principles of risk and responsivity constitutes the theoretical core of the RNR . In recent years, sex offender researchers and practitioners have increasingly recruited the need principle to explain offending and to structure treatment (Mann, Hanson, & Thornton, 2010; Thornton, 2013). The assumption is that dynamic risk factors referred to by the need principle directly track causal processes and as such should be used to develop sex offender treatment programs. Certainly, preliminary evidence suggests that programs that incorporate the principles of the RNR are likely to be more effective than those that do not (Hanson, Bourgon, Helmus, & Hodgson, 2009; Marshall & Marshall, 2012).

We argue that the assumption that dynamic risk factors track causal process in any straightforward sense is incorrect and therefore the RNR is unable to provide a comprehensive guide for treatment on its own. We are assuming that comprehensive rehabilitation frameworks/theories should be theoretically coherent and not crucially depend on problematic ideas or false assertions. By “coherency,” we mean that the concept of dynamic risk factors and its expression in the need principle should not refer to incompatible causal processes or be formulated in logically inconsistent ways. We would also add that important distinctions implicit in the concept of dynamic risk factors should be carefully drawn out rather than run together. On the other hand, the GLM with its basis in agency theory is able to incorporate the concept of dynamic risk in a theoretically coherent manner and apply these insights directly to treatment and offender rehabilitation. The key and pivotal theoretical difference between the RNR and the GLM is that the former is based on shaky conceptual foundations while the latter is not. And given that the need principle is arguably the distinctive theoretical idea in the RNR, its lack of coherency means that the rehabilitation framework collapses into a patchwork of guidelines and practices. While the specific principles and intervention suggestions of the RNR are useful, they need to be underpinned by additional theory if it is to provide a theoretically sound practice framework.

In this chapter, we critically examine the need principle and its associated concepts of dynamic risk factors, risk management, and causal explanation. Concluding that the RNR need principle is theoretically incoherent because the concept of dynamic risk factors does not refer to genuine causal processes in any straightforward sense, we turn to the GLM. In our examination of the GLM’s ability to conceptualize dynamic risk factors, we present the agency model of risk, a recent theoretical innovation, to explain how this composite construct can be employed to explain sexual offending and, ultimately, to guide sex offender treatment. Finally, we conclude the chapter with some brief comments on the comparative empirical and theoretical status of the RNR and GLM and the implications of this standing for future research and practice.

The RNR Need Principle: Dynamic Risk Factors and Causal Explanation

The RNR Basic Principles

The principal architects of the RNR model of offender rehabilitation are the Canadian researchers James Bonta, Don Andrews, and Paul Gendreau (e.g., Andrews & Bonta, 2010a; Andrews, Bonta, & Wormith, 2006; Gendreau & Andrews, 1990). Exactly what constitutes the RNR rehabilitation model is not entirely clear, but typically researchers and practitioners have understood and implemented it according to its three primary principles of risk, need, and responsivity and their associated assumptions (Ward, Yates, & Melser, 2007). In brief, the risk principle suggests offenders at higher risk of reoffending will benefit most from more intensive levels of intervention, including high intensity treatment. The need principle proposes that changeable features of the offender reliably associated with reductions in recidivism (i.e., dynamic risk factors or criminogenic needs) should be targeted in treatment programs in preference to those that have no demonstrated empirical relationship to crime. The responsivity principle states that correctional programs should use empirically supported treatment models (i.e., cognitive behavioral therapy) and be tailored to offender characteristics such as learning style, level of motivation, and the individual’s personal and interpersonal circumstances. The first two principles (risk and need) are used to select treatment intensity and targets, and the whole set of principles are employed to guide the way practice is actually implemented.

The need principle is the central theoretical component of the RNR as it defines intervention targets in terms of risk of reoffending and specifies what kind of factors should be considered dynamic risk factors or criminogenic needs. That is, according to the RNR, dynamic risk factors are changeable features of offenders and their life circumstances that are good predictors of reoffending. In terms of the four key concepts associated with causality identified by Illari and Russo (2014), causal inference, explanation, prediction, and control, the concept of dynamic risk factors emerged from the context of risk prediction and has over time extended its conceptual reach to include all four. That is, researchers and practitioners applying the need principle and the concept of dynamic risk factors to the sexual offending field use them to make causal inferences, explain sexual offending, predict reoffending, and control offense-related propensities and situations. The formulation of the need principle in the RNR and its subsequent role as the core intervention guideline has meant that it has increasingly been used to inform correctional policy (Mann et al., 2010; Thornton, 2013; Ward, 2014; Ward & Beech, 2015).

The other two principles rely on the need principle conceptually and practically. They rely on it conceptually because risk is partly defined in terms of dynamic factors and/or refers indirectly to dynamic factors by way of static variables. That is, static variables are viewed as indicators or pointers to features of offenders that need to be modified if their chances of reoffending are to be reduced. The risk and responsivity principles are practically dependent upon the need principle because their application depends on its acceptance. The risk principle assumes that offender risk bands have been identified, based on a combination of dynamic and static risk variables. Without identified risk factors, the principle is unable to be employed. Relatedly, the responsivity principle concerns the way treatment for moderate- to high-risk offenders is conducted and practically depends on (a) the listing of intervention targets and (b) the development of interventions to alter them in ways that are likely to reduce recidivism rates. The validity and applicability of the need principle with its concept of dynamic risk factors is presupposed by the risk and responsivity principles.

Dynamic Risk Factors and the Sexual Offending Domain

In the sexual offending area, there has been a recent surge of research, clinical, and policy interest in risk assessment, predication, and management (Ward, 2014). Third-generation risk assessment measures and protocols have been developed that incorporate both static and dynamic risk factors, and practitioners are turning to these measures and assessment guidelines for help in formulating cases and planning interventions with sex offenders (Beech & Craig, 2012; Brouillette-Alarie, Babchishin, Hanson, & Helmus, in press; Hanson et al., 2009; Hanson & Morton-Bourgon, 2005; Mann et al., 2010; Thornton, 2013). Ward (2014) recently commented on the increasing reliance on dynamic risk factors and risk management strategies to structure and deliver treatment and raised some concerns:

The status of theory construction has fallen significantly and there is very little cooperation between researchers working on the conceptualization of risk factors and those seeking to explain the causes of sexual offending. In addition, assessment and case formulation seems to revolve largely around the detection of dynamic risk factors and the classification of offenders and their problems amounts to formulating risk profiles. (p. 30)

The concept of dynamic risk factors and its encapsulation in the need principle is arguably a major reason for this shift in theoretical preferences and associated practice. Research on dynamic risk factors in the sexual offending domain has converged on a list of empirically supported factors which are reasonably good predictors of sexual reoffending (Beech & Craig, 2012; Mann et al., 2010; Russell & Darjee, 2013; Thornton, 2013). The empirically supported dynamic risk factors include sexual preoccupation, any deviant sexual interest, sexual interest in children, sexualized coping, sexualized violence, pro-offending attitudes, pro-child molestation attitudes, pro-rape attitudes, generic sexual offending attitudes, emotional congruence with children, lack of sustained marital-type relationships, conflicts in intimate relationships, general self-regulation problems, grievance thinking, impulsivity/recklessness, noncompliance with supervision, antisocial personality disorder, poor problem solving, employment instability, negative social influences, and violation of conditional release.

A notable feature of the above list is the sheer number of dynamic risk predictors, which raises the grain problem. What level should the predictors be categorized at? Do they cluster together into natural groups (kinds) or should they be lumped together at finer levels of resolution and considered separately? Opting for the larger grain solution, Thornton (2013) recently grouped the above risk factors into the four general dynamic risk domains of sexual interest (deviant), distorted attitudes, relational style (problems), and self-management (difficulties). The question of how best to categorize dynamic risk factors is theoretically important once the issue of their coherency is raised. We will return to this issue later in the chapter.

A final general comment on the concept of dynamic risk factors concerns terminology. While most authors seem to use the terms “dynamic risk factors” and “criminogenic needs” interchangeably, others prefer to flag their potential causal role and refer to them as offense-related propensities, psychological traits, vulnerability factors, or dispositions. For example, in a recent paper, Mann et al. (2010) identify a number of what they call psychologically meaningful risk factors which they believe to be prima facie causes of sexual offending and validated predictors of recidivism. They propose that to qualify as psychological meaningful risk factors, (a) there should be plausible reasons for regarding the factor in question as a cause of sexual offending and (b) strong evidence should exist that it actually predicts sexual offending.

Critical Comments

The concept of dynamic risk factors and its utilization in sexual offending research and practice contexts is an important innovation and has lead to significant advances in risk assessment and treatment. What remains to be determined is its degree of theoretical coherency as an explanatory concept as opposed to its utility in risk prediction contexts. The shift from risk prediction to explanatory (etiological) and practice domains is a major one that should be matched by conceptual analysis and if necessary theoretical refinements. The danger in not subjecting such an important concept to theoretical investigation with respect to its ability to function in explanations of sexual offending is that its subsequent use in clinical assessment and case formulation is unjustified. We shall see later on that there are problems with the concept of dynamic risk factors currently relied on in the sexual offending field that render its unmodified incorporation into clinical explanation and research contexts unwarranted (Ward, 2015). A critical question is then: do dynamic risk factors refer to, or pick out, the processes and their associated component structures that cause sexual offending/reoffending in an acceptably transparent and coherent way? Or is the general concept, and the specific examples of dynamic risk factors (e.g., deviant sexual interests), characterized by confusion, incompatible causal elements, and vagueness? In the following discussion, we will be referring to the concept of dynamic risk factors and make the assumption that the concept and its theory and practice utilizations are intended to refer to real processes. That is, we adopt a realist view of scientific theory and accept that our scientific theories and their constituent concepts (which refer to entities, properties, processes, etc.) represent objective, offense-related phenomena and their causes; they exist independently of our individual viewpoints and perspectives (Haig, 2014).

A first general point is that once you start to talk about dynamic risk factors in causal terms, you are obligated to provide a theoretical account of them at some point. While theoretical entities are initially formulated in relatively vague terms such as psychological meaningful causes or offense-related propensities, sooner or later it is expected that a more refined theoretical account is produced, that is, an account that (a) spells out the nature of the underlying causes (structure and processes), (b) details the processes by which they create offense-related problems and outcomes, and (c) describes their relationships to each other. While researchers such as Mann et al. (2010), Beech and Ward (2004), and Thornton (2013) understand this requirement, at this stage they have not provided any such analysis. An encouraging sign has been the recent work by Brouillette-Alarie et al. (in press) on the latent constructs underpinning the Staic-99R and Static-2002R actuarial measures of reoffending risk in sex offenders. In this study, they discovered that the three factors of persistence/paraphilia, youthful stranger aggression, and general criminality could be viewed as potential psychological meaningful constructs, or cutting to the chase, as possible causes of sexual offending. However, Brouillette-Alarie et al. did not unpack the constructs in a theoretically coherent way, and the level of categorization was so broad it is difficult to know whether they are best conceptualized as summaries of predictive factors or references to putative causes (see below).

A second problem with the concept of dynamic risk factors concerns their degree of coherency. At the most general level of dynamic risk factor formulation, some of the conceptual subcomponents appear to be inconsistent with one another. 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 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. This issue highlights the challenges when shifting the zone of application of concepts from one domain to another without making the appropriate conceptual adjustments. In the case of dynamic risk factors, the shift has occurred from the arena of risk prediction to those of explanation and treatment planning.

The problem of construct coherency remains even if you shift from the broad categorization to the specific list of dynamic risk factors identified by researchers such as Mann et al. (2010). To recall, this list includes specific dynamic risk factors such as deviant sexual interests, sexual interest in children, sexualized coping, pro-offending attitudes, emotional congruence with children, lack of sustained marital-type relationships, conflicts in intimate relationships, general self-regulation problems, grievance thinking, and impulsivity/recklessness (to take one example, that of emotional congruence with children). Mann et al. (2010) state that emotional congruence

refers to feeling that relationships with children are more emotionally satisfying than relationships with adults. The offender who is emotionally congruent with children may find children easier to relate to than adults, may feel he is still like a child himself, and may believe that children understand him better than adults do. He often feels himself to be “in love” with his child victims, as if the relationship was reciprocal…… (p. 201)

The above passage comes from the Mann et al. (2010) description of this dynamic risk predictor and is intended to capture the various facets of the emotional congruence construct such as feeling less anxious around children, feeling psychologically like a child, believing children are more understanding or compassionate, and experiencing greater emotional fulfillment with relationships with children. These are all potentially distinct causal processes and may in fact be incompatible as a group. There is no problem with the “composite” nature of the construct in the context of a paper on dynamic prediction; arguably it improves its performance as a predictor to sample diverse aspects of emotional congruence. The difficulty is that when you present the dynamic risk factor of emotional congruence as a possible causal factor and use it to formulate treatment plans and guide treatment, you need to be clear (a) which of the above senses of the concept is applicable, (b) what exactly do you mean by the facet or process in question, and (c) link the causal processes and structures to the outcome variable of interest (offending, relationships etc.). So far, no researcher or theorist has done this with dynamic risk factors except in a very rudimentary sense (see Brouillette-Alarie et al., in press; Mann et al., 2010; Thornton, 2013; Ward & Beech, 2015). The other dynamic risk factors share this problem, and therefore, we conclude that there is a degree of indeterminacy and possible incoherence evident in the concept of sexual offending dynamic risk factors which make its routine use in explanatory and treatment planning domains unjustified.

A third problem is related to the issue of vagueness and incoherency described above (Durrant & Ward, 2015; Ward, 2015). The specific dynamic risk factors are composite constructs in an additional sense as well; they include both trait-like (enduring) and state (temporary) aspects. The stable dynamic factor of general self-regulation includes negative emotionality (a mental state) and poor problem solving (a trait or enduring psychological feature). Another example is the dynamic risk factor of poor cognitive problem solving. Deficient problem solving may involve (a) trait-like features such as lack of relevant knowledge, dysfunctional core beliefs, difficulty integrating information, and problems anticipating future possibilities and/or (b) current states such as feeling anxious, having trouble focusing on relevant features, or experiencing conflicting motivation. The only way to clarify what type of factor is involved, the degree to which it is a cause or an effect, or whether it is an internal or contextual feature is to present a detailed theoretical account of the dynamic risk factor in question and its underlying causal properties and their impact on the person and situation. To date this has not been provided by sexual offending researchers.

A final difficulty with the concept of dynamic risk factors and its incorporation into sexual offending research and treatment is a lack of theoretical attention to their interaction with one another. When researchers and theorists attempt to dig beneath the surface and explain why—and how—an offender sexually abused a child or raped an adult, they are engaged in the causal problem of explanation. The task is to provide an account of how a cause actually produces a specific effect rather than resting content with the demonstration that there is a statistical relationship between two (or more) factors. In an attempt to describe the relevant causal processes in detail, researchers often construct mechanistic explanations that depict how certain entities interact to produce an outcome. According to contemporary theories, the causes of a phenomenon of interest typically consist of a number of interacting processes, none of which are necessary or sufficient for the outcome to occur. In other words, it is more accurate to conceptualize the causes of something like sexual offending as plural in nature, consisting of background conditions, triggers, and interacting mechanisms, that it makes sense to think about causal fields rather than specific causes. In applying causal thinking to case formulation and treatment planning, practitioners need to think in terms of an array of causes and their interaction (Hart, Sturmey, Logan, & McMurran, 2011). The integration of information on risk variables, psychological problems, and situational factors is likely to require the availability of theories that explain how dynamic risk factors exert causal influence and how they combine to create a propensity to offend.

Conclusions

In the above analysis, we identified a number of conceptual problems in the concept of dynamic risk factors. These problems apply to all uses of the concept within the correctional domain and in the sexual offending field. In brief, we argue that the concept of dynamic risk factors is a composite construct and, as such, is valuable within risk assessment contexts. However, once extended beyond this area, it fails to deliver on its explanatory promises. More specifically, the concept is vague, refers to incompatible and/or distinct causal processes, and does not distinguish between trait and state factors. Given that the concept of dynamic risk factors is theoretically problematic, any theory or theoretical framework that depends on this concept is substantially weakened. The RNR presupposes the validity and theoretical cogency of the concept of dynamic risk factors and is therefore markedly weakened by these conceptual flaws. We conclude that the RNR rehabilitation framework, or theory, is not a coherent rehabilitation framework and collapses into a loose patchwork of practices, guidelines, and norms.

The GLM, Dynamic Risk Factors, and the Agency Model of Risk

The theoretical dependence of the RNR on the concept of dynamic risk factors has undermined its coherency as rehabilitation theory. One of the points of contrast between the RNR and the GLM is the latter’s emphasis on personal agency and goal-directed behavior. In our view, this view of human functioning and motivation provides a way of conceptualizing dynamic risk factors that enable it to accommodate their composite or hybrid nature. We will now briefly outline the GLM and then describe the agency model of risk recently developed in a number of publications (Durrant & Ward, 2015; Ward, 2015; Ward & Beech, 2015). We conclude that because the GLM is able to satisfactorily integrate the important concept of dynamic risk factors into its structure by way of the agency view of risk, it is in this respect a more coherent rehabilitation theory than the RNR.

The Good Lives Model of Offender Rehabilitation

The Good Lives Model (GLM) is a strength-based approach to offender rehabilitation because it is responsive to offenders’ particular interests, abilities, and aspirations (Ward & Maruna, 2007; Ward & Stewart, 2003; Willis et al., 2014). It also asks practitioners to explicitly construct intervention—good lives plans—plans that help offenders acquire the capabilities to achieve personally meaningful goals. From the perspective of the GLM, sexual offending results when individuals lack the internal and external resources necessary to realize their values in their everyday lives using pro-social means. In other words, criminal behavior represents a maladaptive attempt to secure valued outcomes (Purvis, Ward, & Shaw, 2013; Ward & Stewart, 2003; Yates et al., 2010). Rehabilitation plans should therefore aim to equip offenders with the knowledge, skills, opportunities, and resources necessary to satisfy their life values in ways that do not harm others. Related to its strong focus on offenders’ core commitments and lifestyles that reflect these, there is a corresponding stress on agency. That is, because of the assumption that offenders like the rest of us actively seek to satisfy their life values through whatever means available to them, any rehabilitation plan should be pitched at the level of agency, goals, planning, and facilitative environments. In this sense, it is an ecological model and always keeps in mind the relationship between the environments in which persons live and the capabilities and resources they need to live meaningful and crime-free lives (Ward & Stewart, 2003).

Criminogenic needs or dynamic risk factors are conceptualized within the GLM as internal or external obstacles (i.e., flaws within a good life plan) that make it difficult for individuals to secure primary goods in personally meaningful and socially acceptable ways. These flaws take the form of insufficient attention to the range of goods required for individuals to have a chance at fulfilling lives, lack of internal and external capabilities, the use of inappropriate and counterproductive means to achieve personal goals, and conflict within a person’s good life plan (Ward & Maruna, 2007; Yates et al., 2010). There is no assumption that dynamic risk factors are anything other than individual, social, and environmental problems that are causally related to sexual offending. It is understood they are composite constructs developed in the domain of risk prediction that are expected to break apart when recruited to perform explanatory roles.

According to the GLM, there are two ways rehabilitation programs can reduce dynamic risk factors. First, the establishment of the internal and external capacities needed to achieve a primary good (or more broadly, implement a good life plan) in socially acceptable and personally fulfilling ways can directly modify dynamic risk factors. For example, learning the skills necessary to become a mechanic might make it easier for an offender to develop the skills for concentration and emotional regulation, thereby reducing impulsivity, a criminogenic need. Second, the reduction of risk can occur indirectly when an offender is strongly motivated to work hard in treatment because of his involvement in projects that personally engage him.

The Agency Model of Risk

Dynamic risk factors in the correctional domain are intended to predict harm related to reoffending, typically to victims and the community. 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. As argued above, dynamic risk factors have no reality apart from prediction contexts and do not refer uniquely to causal processes that result in sexual offending. They are composite variables best conceptualized as predictive devices rather than explanatory constructs. In this sense, we agree with Borsboom (2005, p. 158) that

If term is treated as referential but has no referent, then one is reifying terms that have no other function than that of providing a descriptive summary of a distinct set of processes and attributes. For instance, one then comes to treat a name for a group of test items as if it were a common cause of the item responses. That of course is a mistake.

In our opinion, one useful way of thinking about dynamic risk factors in sex offenders is by conceptualizing them in terms of the components of agency, that is, viewing dynamic risk factors as composite constructs that are useful predictors because they cover important aspects of goal-directed actions, within an offending context. The capacity for agency is inherent in all living things; however, in human beings, the level of sophistication is ratcheted up several notches because of their ability to intentionally structure learning and physical environments (i.e., niche construction). The key components of agency are (1) goals, plans, and strategies; (2) implementation of plans and their evaluation; and (3) the subsequent revision of goals and plans in light of outcomes. Furthermore, in our recent Agency Model of Risk, there are three levels of agency, each associated with its own distinct set of goals, plans, and strategies and each capable of influencing the other types of agency. The levels of agency in the AMR are the system level (goals related to physical integrity and functioning), social role (goals concerned with specific social roles such as being teacher), and personal (concerns individuals overall sense of identity and core normative commitments). The type of goals offenders possess and the plans they construct to achieve their goals and to evaluate their effectiveness are partly a function of the contexts in which they live and the resources available to them. Goals are activated or selected in response to external contexts and their cues such as the presence of threats and by internal cues such as hunger, fear, sexual desire, or anger. Dynamic risk factors can be viewed as flaws in individual functional capacities, social supports, and opportunities. Thus, dynamic risk factors once broken down into their causal elements are seen 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 occur at multiple levels or can be confined to incorrect actions within a single practice (e.g., relationship repair task). Protective factors, once stripped down into their core elements, work in multiple ways across the various levels of agency to inhibit and/or disrupt dysfunctional systems and to restore normal functioning. Sometimes, the constraints exerted by protective factors are external, such as the construction of supportive social networks around high-risk offenders.

The implications of this depiction of dynamic risk factors and their division into criminogenic needs and lifestyle destabilizers are far reaching. For example, the dynamic risk factor (criminogenic need) of intimacy deficits in sex offenders can be understood as (1) maladaptive beliefs and norms concerning relationships (e.g., adults are untrustworthy); (2) interpersonal and emotional regulation strategies that damage relationships with adults (e.g., do not talk about feelings and avoid social contact with adults) and that isolate individuals from social support; and (3) the active search for, and construction of, social environments in which such individuals feel comfortable and where their needs seem to be met (e.g., pedophilic networks and spending a lot of time with vulnerable children). The idea is to break down dynamic risk factors into several causal elements that in certain environments create, and maintain, antisocial values or behavior. Furthermore, it then becomes much easier to dig beneath the surface to redirect research and practice to relevant targets.

The GLM—and arguably other strength-based treatment approaches (see Marshall, Marshall, Serran, & O’Brien, 2011)—is able to theoretically ground a conceptualization of dynamic risk factors in the AMR because of its strong emphasis on offender agency and the central role that values, goals, strategies, and environmental variables play in non-offending and offending spheres. The point of creating the AMR was to provide detail on possible ways dynamic risk factors could be causally related to offending. The conceptual link between the GLM and an agency view of dynamic risk factors is its assumption that individuals translate important values into concrete goals and actively strive to realize them in their everyday lives. Therapeutically, the construction of good lives plans around offenders’ most heavily weighted primary goods (and their associated personal goals) encourages practitioners to build desistance elements into rehabilitation initiatives (Laws & Ward, 2011; Scoones, Willis, & Grace, 2012). After all, a good life plan is a plan for living a different kind of lifestyle, and this means understanding the dependence of human beings on their relationships with others and the environment if significant change is to be maintained once they leave prison. Relevant desistance factors include access to social models that promote a non-offending lifestyle, employment, a stable emotional relationship, good social support, cognitive competencies, development of an adequate self-concept, and the acquisition of a sense of meaning in life.

GLM and RNR Empirical Research

Our major aim in this chapter has been to examine the way the RNR and GLM conceptualize dynamic risk factors. Of course, we understand that empirical considerations are also critical in evaluating rehabilitation theories and their associated treatment programs: are the constructs valid and treatment based on them effective in reducing reoffending rates? Understanding the causes of offending and reoffending should help practitioners to predict, explain, and control aspects of crime. We will now briefly describe recent work on the RNR and GLM and identify areas where empirical research is required to assist theorists and practitioners to determine the strengths and weaknesses of the RNR and GLM.

Risk-Need-Responsivity Model

There is considerable meta-analytic support indicating that adherence to the three major RNR principles is associated with reductions in sexual reoffending and also that adhering to more principles is associated with greater reductions in reoffending (Hanson et al., 2009). According to the RNR model, changes in dynamic risk factors should be associated with reductions in reoffense risk (Andrews & Bonta, 2010a, 2010b). Some research supports this assumption. The Violence Risk Scale: Sex Offender version (VRS:SO; Olver, Wong, Nicholaichuk, & Gordon, 2007; based on the Violence Risk Scale, VRS; Wong & Gordon, 2006) incorporates within-treatment change into sex offender risk assessment. Included are clinician-rated static and stable dynamic risk scales, and change in each of the dynamic domains is measured using a modified application of the transtheoretical model of change (Prochaska, DiClemente, & Norcross, 1992). Olver et al. (2007) found that posttreatment pro-social change was significantly related to reductions in sexual recidivism after controlling for static risk, pretreatment dynamic scores, and follow-up time (Olver et al., 2007).

Several researchers/clinicians have highlighted the importance of the responsivity principle and its poor adherence among sexual offending treatment providers (e.g., L. E. Marshall & Marshall, 2012). Narrow operationalization of the RNR principles and even narrower assessment of their adherence likely obscure detection of the most effective sexual offending treatment programs. Missing from large-scale meta-analyses are consideration of therapist characteristics, group cohesion, use of approach goals, assessment of agency capacity, etc., which have all been demonstrated to enhance treatment effectiveness (Marshall et al., 2011). Finally, one of the weakest aspects of the RNR is its theoretical looseness and tendency to concentrate on the application of the RNR principles to correctional practice at the expense of establishing its coherency as a rehabilitation framework (Ward & Maruna, 2007).

The Good Lives Model of Offender Rehabilitation

The GLM was first proposed by the lead author in 2002 (Ward, 2002) and since then has undergone several theoretical and practical developments (e.g., Laws & Ward, 2011; Purvis et al., 2013; Ward & Maruna, 2007; Ward & Stewart, 2003; Willis, Yates, Gannon, & Ward, 2013; Yates et al., 2010). Emerging research supports the conceptual underpinnings of the GLM when applied to individuals who have sexually offended (Barnett & Wood, 2008; Willis et al., 2013). Substantial variation has been observed in terms of how the GLM has been operationalized in practice (Willis et al., 2014). It is therefore not surprising that empirical support for the GLM has been somewhat mixed, and studies investigating recidivism outcomes have not yet been conducted. However, studies to date suggest that closer adherence to the model is associated with better outcomes (see below). In addition, research suggests that widening the net of risk assessment practices to include assessment of offenders’ strengths might further enhance the predictive validity of sex offender risk assessment (e.g., Scoones et al., 2012).

The three group-based studies to date that have explored the effectiveness of adopting the GLM as an overarching rehabilitation framework in sexual offending treatment programs compared to traditional risk-oriented approaches have provided preliminary, although mixed, evidence concerning its superiority (although it was never inferior to traditional approaches—see Barnett, Manderville-Norden, & Rakestrow, 2014; Harkins, Flak, Beech, & Woodhams, 2012; Simons, McCullar, & Tyler, 2006). For example, Simons et al. (2006) found that offenders who received the GLM approach were more likely to complete treatment, remain in treatment longer, and be rated by therapists as more motivated to participate in treatment compared to clients who received traditional relapse prevention treatment. In addition, clients who received the GLM approach demonstrated significantly better coping skills posttreatment, and no such gains were observed for clients who received the RP approach.

Compared to the RNR, there has been a lack of good quality treatment outcome studies using the GLM as a rehabilitation framework with which to construct treatment plans. Part of the problem is that the GLM was always intended to include the RNR principles, and therefore, treatment derived from these theories that simply compares the two is missing the point. The superiority of the GLM was seen to reside in its greater degree of theoretical coherency and ability to incorporate aspects of desistance and treatment that the RNR struggles to find room for (e.g., treatment alliance, agency, approach goals, core values, personal identity, etc.). What is required is further research into the two models that (a) explicitly evaluates them as rehabilitation frameworks and (b) provides detailed specification of the differences between strict RNR treatment programs and those augmented or underpinned by GLM principles (see Willis et al., 2014).

Conclusions

The RNR consists of a loose coalition of rehabilitation principles, theoretical assumptions concerning the relationship between psychological variables and offending, and concrete practice guidelines. It is not an integrated theoretical framework because the foundational concept of dynamic risk factors does not travel well away from prediction contexts. It is a composite or hybrid construct and as such a poor candidate for recruitment into etiological theories of sexual offending and reoffending. The problem means practitioners should not rely on dynamic risk factors without recourse to a “translation” model such as the AMR when they set about constructing case formulations and delivering treatment. On the other hand, the GLM is able to incorporate the concept of dynamic risk factors into its theoretical structure, by accepting that they are unable on their own to explain sexual offending and therefore should not be relied on in assessment and treatment planning. Another reason the GLM is friendly to dynamic risk factors is that it does not have a strong commitment to them as explanatory concepts or psychological causes and insists on taking them apart when setting out to understand why offenders sexually abuse others. The stress the GLM places on agency and the associated requirement to construct meaningful good lives plans for sex offenders make it much easier to deconstruct dynamic risk factors into their multiple causal elements. If researchers and practitioners within the sexual offending field want to continue referring to dynamic risk factors in an explanatory sense or, more broadly, structure practice according to the RNR, they will need to supplement their theoretical resources with something like the AMR. Another alternative is to use the core normative, etiological, and practice assumptions of the GLM and the AMR and to embed the RNR principles and allied concepts within this theoretical framework. What they should not do is to carry on as usual. In our view, such a decision is likely to push the field rapidly into theoretical rehabilitation dead ends (Ward & Beech, 2015).

What does all this mean for the RNR as a rehabilitation theory? An obvious question is: why has there been so little discussion about problems with importing dynamic risk factors into etiological, assessment, and treatment domains or more expressed concern about their composite nature? Perhaps, the issue is merely a semantic one and too trivial to bother about. Sooner or later the promissory note issued by the RNR theorists will be made good, and a reworked, more powerful, and coherent account of dynamic risk factors will be presented. We think this is all beside the point and that the arguments outlined in the chapter speak for themselves: the RNR suffers from a fatal case of conceptual incoherence and therefore fails as a rehabilitation theory. The GLM does not suffer from the same problem (although it may be subject to other difficulties—see Andrews et al., 2011) and, in this respect, is a stronger rehabilitation model. If researchers and practitioners in the sexual offending and the broader correctional fields are committed to the ideal of evidence-based practice, this means becoming theoretically literate as well as being familiar with the empirical evidence for assessment measures and treatment programs (Gannon & Ward, 2014). Theories, models, concepts, and principles are cognitive tools, and if we are to do our jobs as well as we can, they are necessary to sharpen our practice.