The impact of emotion on offender decision-making: advancing our understanding through virtual re-enactment

ABSTRACT

This article reviews approaches to the study of emotion in offender decision-making research. It then draws together theoretical models from psychology and neuroscience to propose an innovative method with which to move this field forward. Theories in this field have long called for a deeper understanding of the impact of emotion on criminal behaviour. While recent research has made some fruitful gains, it has been hampered by an overly narrow view of emotion and by limited methodologies. Using the expertise paradigm as an interdisciplinary and integrative framework, this article advocates incorporating into it established models of emotional experience from affective neuroscience to aid this endeavour. It further argues that employing the currently evolving virtual enactment method as a measurement tool will provide an unparalleled insight into the study of emotion as it unfolds during the offending act. Both environmental crime prevention and desistance from crime are likely to benefit considerably as a result.

KEYWORDS:

• offender decision-making
• emotion
• expertise
• affective neuroscience
• virtual reality

Introduction

One of the first models to clearly focus on the offender’s viewpoint on crime, some four decades ago, was the rational choice perspective (RCP, Cornish & Clarke, 1986). It focused on decisions and behaviour at and around the crime scene and provided a strong theoretical basis for a wide range of situational perspectives. These perspectives included models to reduce the opportunities for crime in communities and neighbourhoods such as routine activities model (Cohen & Felson, 1979); situational crime prevention (Heal & Laycock, 1986), and crime pattern theory (Brantingham & Brantingham, 1975). Decades of research on the offence, the offender’s experience and crime prevention ensued as a consequence of this new theoretical perspective. Scholars have built on the RCP framework, aiming to deepen both its explanatory value and its efficacy in real world applications (e.g. work from the perceptual deterrence perspective, Nagin, Solow, and Lum (2015); the hot/cool perspective, Van Gelder (2013); and the expertise perspective, Nee and Ward (2015).

One could argue that these recent interdisciplinary theoretical developments, which incorporate robust findings from allied sciences, have opened a new and fruitful pathway to understanding the actions and decisions of the offender (Topalli et al., 2020). They have, however, failed to fully translate this new knowledge into practical, effective methods of preventing crime in communities and to support desistance in individuals who have already committed offences. This article aims to highlight an aspect of human experience that has been undervalued in the development of these explanations: that of the impact of emotion. Psychological sciences have seen prodigious developments in the study of human affect in recent decades (Dukes et al., 2021) but this has not been reflected in the study of offender decision-making. Thus, using the expertise paradigm as an integrative conceptual framework, I argue that incorporating theoretical models and findings from affective neuroscience can potentially make a significant contribution in enriching our understanding of offence related decision-making. Additionally, I argue that the virtual enactment methodology (using virtual reality) that is evolving alongside the expertise paradigm (Meenaghan et al., 2018; Van Gelder et al., 2022), provides an unparalleled way of measuring emotion as it unfolds during offending cognition and behaviour, by addressing the ethical and methodological limitations of traditional research methods. Accordingly, the integration of models of emotion from affective neuroscience into the expertise model, in conjunction with the adoption of virtual reality technologies, can open up new avenues for theoretical and empirical research. Ultimately, this could contribute significantly to the prevention of crime.

The appeal for a clearer understanding of the emotional influences on criminal decision-making is not new and has been gathering momentum for a number of years (Nagin, 2007; Roche et al., 2020; Van Gelder et al., 2014a). This has been fuelled in part by a deeper appreciation of the nature of, and relationship between, cognition and emotion. Researchers in psychology, neuroscience and philosophy have (1) moved away from a view that brain and body are distinct entities, towards a recognition that our whole bodies are involved in the acquisition of knowledge and experience (a.k.a. embodied cognition) and (2) acknowledged that emotion is a crucial and fully integrated aspect of information-processing and perception.

The article is divided into four sections. In the first section, I give an overview of recent theoretical developments in research on offender decision-making, explaining why the expertise paradigm may be the most productive of these with which to incorporate the study of emotion, and trace how these have begun to fill explanatory gaps in traditional models. Second, I examine what we have learnt about emotion in criminology so far and reinforce the long-held view in the discipline that we need to know more about how emotion contributes to criminal behaviour. I argue that we need to go beyond relying on qualitative offender studies and quantitative vignette studies of the general population, to study the impact of emotion on decision-making in a more systematic and controlled way. Third, I demonstrate how recent findings in affective neuroscience and philosophy can provide a theoretically exciting adjunct to investigate the emotional mechanisms that scaffold behaviour. Finally, I propose that a mixed-methods approach to research methodology, using virtual reality (VR), physiological and behavioural measures alongside self-report measures, is a fruitful way to capitalise on these new insights. The sections on expertise and virtual enactment mostly reference empirical work done on residential burglary, as much research has been undertaken on this topic thus far. It is important to emphasise, however, that the proposals for theoretical and empirical work on the study of emotion given in this article are by no means limited to this crime. Emotion is an integral part of all decision-making, and therefore impacts all types of offending behaviour.

Recent theoretical developments in offender decision-making

At its origins, RCP shone a light on a previously neglected and essential part of the offending puzzle (Cornish & Clarke, 1986). It provided a valuable foundation by describing the relationships between offenders and their environments and noting that criminal decisions were constrained by satisficing (arriving at a ‘good enough’ plan of action), limited information and the use of decision-making heuristics. While alluding to factors that are not always conscious (e.g. bounded rationality), it did not specify the nature of these processes. Consequently, the practical value of applied models of crime reduction that initially arose from RCP were constrained by these limitations.

This situational perspective on crime prevention i.e. attempts to prevent and reduce the opportunities for crime by changing the physical environment that facilitates it, saw a partially successful change in addressing offending behaviour such as burglary (Farrell & Pease, 2014), street crime (Sherman, Gartin, & Buerger, 1989) and robbery (Cook & Macdonald, 2011). Over time though, evaluations showed they had a limited impact on offending rates (Telep et al., 2014) and possibly resulted in a displacement of crime to poorer neighbourhoods (Tseloni & Thompson, 2018). The worldwide drop in official burglary rates in the 2000s was largely attributed to the situational crime perspective (Farrell et al., 2011), but this has since plateaued at an unacceptably high rate.¹ In recent years, however, RCP has received considerable updating with major advances in the study of information-processing and decision-making. These efforts have gone some way to addressing three salient gaps in the RCP model, adding to the paradigm’s explanatory value. First, strong evidence for dual processing models (i.e. the divide between conscious and less conscious thinking) has been utilised from cognitive and social psychology research. This can be seen particularly in the fields of perceptual deterrence (Apel, 2022); offender expertise research (Nee & Ward, 2015) and the hot/cool processing model (Van Gelder, 2013). These research findings have provided a deeper understanding of the attentional mechanisms, knowledge structures and automatisation processes at play in generating offending behaviour. Second, as well as examining the offender’s appraisal of the negative consequences of the crime (risk versus gain), more recent explanations focus on important integral and incidental processes that drive decision-making and behaviour ‘in the moment’, as the crime unfolds (Barnum, Nagin, & Pogarsky, 2021; Jacobs & Cherbonneau, 2019). Third, models which demonstrated how improved cognitive functioning as a result of practice and experience can render offenders more proficient in the commission of crimes (Nee & Ward, 2015; Van Gelder et al., 2017) move away from a solely negative focus on how cognitive limitations hamper decision-making.

To capitalise on this growing theoretical movement, I suggest we need a more incisive way to understand the role of cognition in offending behaviour. Below I focus on the expertise model in preference to other dual processing models of criminal behaviour I have mentioned above for the following reasons. As the first part of its name suggests, the perceptual deterrence literature has embraced developments in the dual information-processing world. This has been especially regarding the offender unconsciously updating their decision-making as the crime progresses in terms of the risk of getting caught (known as perceptual updating; Paternoster, 2018). However, by its very nature, this field is focussed on the negative consequences of crime and their deterrent value. The influence of positive emotion in driving and sustaining (and perceptually updating) the unfolding process of a crime in commission, not to mention the sense of agency it affords the offender, cannot be overstated and is missing from current explanations. The perceptual deterrence research has also been associated with somewhat limited methodologies, which will be discussed later. The hot/cool processing model (Van Gelder, 2013), developed from the work of Loewenstein and Nagin (e.g. Loewenstein et al., 1997) suggests that rational criminal decisions about the costs and benefits of crime (cool decisions) will be imbued with ‘hot’ automatic emotional memories regarding what is immediately rewarding and what is immediately aversive about the situation. This work underlines the importance of seeing cognition and emotion as a unified concept rather than epiphenomenal. It does not explain though, how much automatic, unconscious processing is cool and unemotional (e.g. medical decision-making; Norman et al., 2006). The model undoubtedly moves us forward in terms of theoretical thinking about decision-making, but so far, experimental work in this vein also focusses solely on negative emotions such as anger and shame (Van Gelder et al., 2014b).

Alternatively, the expertise model sets out a broader framework drawn from a wide range of experimental social and cognitive psychology. It gives a more complete explanation of the mechanisms underlying mental processing (such as the role of memory and behaviour in developing cognitive schemas and the automisation of decision-making – see below); it embraces positive emotion rather than disregarding it; and it acknowledges competencies in the individual leading to an increased sense of agency, thus supporting more offending (and conversely assisting desistance; Nee & Vernham, 2017). All in all, it encompasses a broader spectrum of human experience and is an integrative framework that can accommodate and assimilate additional explanations, in line with an integrative pluralistic approach to science (Kendler, 2005; Mitchell, 2003).

Below I demonstrate how the expertise model, and the VR methodology that has developed alongside it, arguably provide the most valuable conceptual resources in moving this research forward.

To do this, a brief introduction to the main features of the expertise model are given next. A more in-depth description is given in Nee and Ward (2015).

A snapshot of expertise theory using the example of burglary behaviour

Expertise refers to the skills and knowledge an individual develops through learning and concerted practice in a particular domain that result in consistently superior performances compared to those new to that domain (i.e. a novice; Ericsson, 1996). Expertise is hypothesised to lie on a continuum with novices at one end and masters at the other (Chi & Bassok, 1989). Most individuals have the ability to develop expertise in some domains (for instance, many of us are able to develop expertise in relation to driving a car, or learning an additional language), but it is rare for individuals to reach the extreme end of proficiency, unless they engage in continual, deliberate and challenging practice (Ericsson, 1996). Expertise in any given field is dependent on the acquisition of time-saving representations in memory and their subsequent behavioural manifestations.

Reviews have consistently identified processes which unify experts, irrespective of their domain of practice (Palmeri et al., 2004). These processes include: (1) preconscious attention (below the threshold of consciousness) associated with a heightened situational awareness of cues relevant to the domain of expertise (Bargh, 1994); (2) the development of complex knowledge networks (cognitive schemas) stored in long-term memory (Fiske & Taylor, 1991) as a result of this; and (3) automaticity (habitual, quick, intuitive) of both decision-making and behaviour (Logan, 1988), the latter of which is triggered by scripts in long-term memory.

An outline is given below of examples of burglars undertaking more ‘efficient’ burglaries, with more knowledge and skill than control groups, despite their impoverished educational backgrounds. It illustrates how using knowledge from allied disciplines can deepen our understanding of (and aid disruption of) offender decision-making.

Selective preconscious attention

Research is littered with examples of the heightened burglary cue awareness that supports preconscious scanning of the neighbourhood. Wright and Decker’s burglars (1994, p. 80) describe ‘half-looking’ and ‘scoping’ potential properties to burgle. Cromwell, Olson, and Avary's (1991) ‘journeyman’ burglar searched out opportunities based on systematic recognition of environmental cues signifying gain and low risk. Nee et al.’s (2019) experienced burglars demonstrated economical processing of a large number of burglary-related cues compared to controls.

Cognitive schemas

Burglars report considerably more knowledge and consistency regarding target selection in interviews and experiments than control groups. These include: preferences for corner properties; rear access and vegetational cover; types of windows, doors and locks; (lack of) occupancy cues and ability to burgle less familiar properties (Bennett & Wright, 1984; Clare, 2011; Nee & Taylor, 2000; Wright & Decker, 1994). This is supported by both crime survey data (Tseloni et al., 2017) and studies of location and topography of burgled properties (Bernasco & Luykx, 2003; Coupe, 2017). As well as extending our knowledge considerably (see below), recent VR re-enactments have seen burglars replicate these preferences in real-time (Nee et al., 2015, 2019). These studies highlight the ease and efficiency with which these offenders scan, scope and behave while carrying out the crime in relation to other offenders and non-offending samples. Findings suggest the development of, and instantaneous access to, rich and inter-connected schemas about how to do burglaries with minimum risk and maximum gain.

Automaticity

Evidence concerning the ease and flow of the burglary episode, from scanning a neighbourhood, to target choice, to completing the burglary, goes as far back as 1984 in Bennett and Wright’s ground-breaking series of studies on experienced burglars. In studies using ‘expert’ burglars, control groups of novices (students and householders; Wright et al., 1995) and other offenders without burglary experience (Logie et al., 1992; Nee et al., 2019) there is evidence for tried and tested, heuristically driven navigation strategies to areas of high value, more thorough and systematic searching, and a more discriminate choice of items to steal (Wright & Decker, 1994). These studies all suggest a lighter cognitive ‘touch’, and more automated information-processing using scripts from long-term memory. Backing up these quantitative indications are compelling examples of qualitative, often spontaneous verbalizations indicating the automatic nature of burglary decision-making and behaviour (see Meenaghan et al., 2020; Nee & Meenaghan, 2006).

While burglary has had a considerable amount of work done in the expertise vein, levels of skill and automaticity have been identified when preparing for and carrying out many other crimes, allowing offenders to detect vulnerabilities associated with their victims and evade detection. Work has been carried out on child sexual offenders (Bourke et al., 2012); sexual burglars and sexual robbers (Reale et al., 2022); rapists (Chopin et al., 2022); street offenders (Topalli, 2005); identity thieves (Vieraitis et al., 2015); domestic violence perpetrators (Bonomi & Martin, 2021), cyber-paedophiles (Chopin & Décary-Hétu, 2023) and carjackers (Topalli, Jacques, & Wright, 2015) to name a few. Understanding the power of emotion during the often unconscious and instantaneous decisions proximal to these crimes will surely be crucial in addressing them more effectively.

The expertise paradigm, like other dual processing models, has improved the explanatory power of offender decision-making models in recent years, but none as yet have explored the contribution of emotion to offender decision-making in any great depth (Nagin, 2007; Roche et al., 2020; Van Gelder et al., 2014a). This is a significant omission as there is a developing consensus in the field of cognitive neuroscience and decision-making that emotion (1) motivates and facilitates everyday decision-making and (2) that cognitive and emotional processes are deeply intertwined.

Recent criminological studies of emotion

A growing movement in criminology has called for a more systematic examination of the role of immediate emotion in the development of prolific offending behaviour. From early work in the 1990s on the distorting effect of emotion on risk perception (Loewenstein et al., 1997) to Van Gelder et al.’s (2014b) research on how immediate anger reduces anticipated shame, work that focuses purely on the impact of the consequences of crime on decision-making has been called into question. The deterrence perspective has had a natural focus on how the perceptions of anticipated punishment will affect online decision-making, but more recent studies from this perspective too have incorporated an appreciation of the impact of immediate emotion on decision-making. Pogarsky et al. (2017) criticise RCP for failing to take into consideration affective ‘signals’ in the ongoing information-processing during an offence. Two national surveys using large MTurk samples and online vignettes, showed that exposure to emotion-laden information caused changes in perceptions of formal sanction risk. For instance, individuals who received positive emotional messages about texting were less likely to think they would be arrested if texting while driving compared to those who received neutral or negative messages (Pogarsky et al., 2017). This was labelled as the ‘affect heuristic’. Later, explicitly examining the effect of fear of apprehension on risk-taking, Pickett et al. (2018) using a similar methodology, found that the inclusion of measures of fear increased the explanatory power for both criminal propensity and situational offending intentions. In their analysis fear was a stronger predictor of situational intentions to offend, than either self-control or prior offending. Roche et al. (2020) built on this work by factoring in the effect of perceived control over apprehension on criminal propensity. Perceived levels of control over sanctions reduced fear which increased criminal propensity. Barnum and Solomon (2019) induced integral (current) emotions using provocative vignettes. They neatly highlighted the interaction between individual differences and situational determinants. They showed that the more emotion experienced, the more the individual’s attention was steered away from the outcomes and consequences of the potential violent behaviour. Although not explicitly measuring emotion, Barnum et al. (2021) employed a complex design aiming to expose the mechanisms underlying perceptual deterrence. They used videos in a study of speeding in which traffic conditions, speed and levels of priming were manipulated in a large MTurk online experiment. Despite wide ranging personal differences (such as risk seeking and driving experience), on average, as speed increased, perceived risk of apprehension and perceived lack of safety (i.e. sense of danger and its associated feelings) increased. On the flip side and interesting from an expertise point of view, risk of apprehension and sense of safety were negatively correlated with the intention to speed. This resonates with Apel’s (2022) long-held finding that the experience of escaping apprehension reduces its perceived risk, and vice versa.

These noteworthy recent studies have shed light on the impact of emotion in the criminal act, especially by emphasising perceptual updating at the moment of the crime through past experience (Paternoster, 2018). That said, they would benefit from further empirical testing to improve their ultimate effectiveness in the reduction and prevention of crime. The studies often use hypothetical written scenarios (such as lotteries, or imagined offending behaviour in vignettes) that bear little resemblance to the real assessment of risk involved in serious offending behaviour (Van Gelder et al., 2022). Consequently, they focus on overly narrow aspects of the decision-making process which likely fail to capture the complex information-processing characteristics of decision-making in real life (e.g. long-term and working memory, incidental and situationally-induced emotions, processing of internal and external cues etc.). Additionally, the studies focus almost exclusively on the negative aspects of the emotions surrounding offending behaviour to the exclusion of the important positive aspects which may in part constitute the experience of the offender, such as the perception of reward/mastery. Recent qualitative studies on emotion and crime involving offenders rather than the general public (Jacobs & Cherbonneau, 2017, 2019; Lindegaard & Jacques, 2014; Meenaghan et al., 2020) strongly support the role of positive as well as negative emotion as an important component of criminal behaviour, corroborating the qualitative studies reviewed earlier. Crime-related skills such as developing the ability to hold one’s nerve during criminal activity, and exhibiting negative emotions such as anger and threat in ‘appropriate’ situations, are consistently reported by offenders and add to their sense of agency, reputation and criminal capital (Nguyen, 2020). Mastery, agency, a feeling of belonging, and the positive emotions associated with these are not only crucial to the understanding of why people commit crime, but also to its desistance.

A final limitation in this recent work, and perhaps the most crucial issue to address, is the reliance on samples from largely non-offending backgrounds. These samples differ from offender populations on numerous demographic measures such as gender, ethnicity, education, income, housing etc. This is not to say that valuable research cannot be done on such samples. Important discoveries have been made and replicated using, for instance, student samples across the decades, for theory testing and pilot-testing new methods (e.g. the Milgram obedience experiments, Milgram (1963); or the Zimbardo prison experiments, Haney et al. (1973), or more recently lie detection (Vrij, 2008)). But in the pursuit of understanding the proximal cognitions and emotions that drive offending behaviour, the myriad of factors that affect both brain development and life experience in the majority of offenders, indicates that their perception and experience of what risk is, will differ considerably from non-offenders. It is well-known that a wide range of criminogenic risk factors affect most offenders from conception onwards (Liu, 2011) including exposure to toxic substances, lack of secure boundaries, malnutrition, sleep deprivation, traumatic brain injury, and specific learning and behavioural disorders which worsen educational failure and the likelihood of unemployment and substance misuse. All of these issues disrupt the ability to develop effortful control over cognition, emotion and consequent behaviour (Gupta et al., 2011), resulting in someone who is considerably more impulsive than their typically developing counterparts. Paradoxically, despite these neurological and behavioural challenges, a young person who has embarked on a life of prolific offending still has the capacity to quickly build up, through practice, a repertoire of automatic mental schemas (knowledge networks) around their illegal activities, allowing them to undertake crimes efficiently and systematically. This difference in the ability to carry out crimes between those experienced in offending and those not experienced in offending (e.g. students, householders, community samples, police officers) or offenders who are not experienced in a particular type of crime, has been demonstrated empirically many times (Logie et al., 1992; Nee & Taylor, 2000; Nee et al., 2015, 2019; Topalli, 2005; Van Sintemartensdijk et al., 2020). Thus, it would be unwise to comprehensively generalise about perception of risk and deterrence from non-offender samples given their general lack of experience with real offending situations.² Indeed, many criminologists point this out (Apel, 2022; Nagin et al., 2015; Paternoster, 2018). Offenders will, at the very least, be desensitised (or sensitised in a different way) to criminogenic situations and are likely to perceive reward, risk and threat in a different way to non-offenders. Their different take on risk necessitates the replication of the risk perception work with prolific offenders to ascertain its applicability to the population for which it is, one could argue, most important.

Many researchers above have acknowledged this limitation in relation to their own work (e.g. Pickett et al., 2018; Pogarsky et al., 2017) accepting that (1) those who perceive a high likelihood of apprehension for committing a crime may nonetheless be unafraid of getting caught, and (2) many prolific offenders are deeply desensitised to the fear of apprehension (Brezina et al., 2008; Carroll & Weaver, 1986; Treiber, 2014; Wright & Decker, 1994) due to lack of ambiguity, perceived ability to cope, or biological factors. Golman et al. (2021) observe that lay people take more risks when feeling good and are more risk-averse when feeling bad, yet the opposite has been noted in offenders (Van Gelder et al., 2009; Wright & Decker, 1994).

To summarise the argument so far, the investigation of offender decision-making suggests that individuals can develop considerable levels of offence specific expertise that improves their ability to elude detection and increases their chances of subsequent offending. Preliminary research on emotions and their impact on offence related cognitive processes indicates that positive and negative emotional states are integral components of decision-making by offenders. Emotions seem to play a strong facilitative, motivational role in offending and not simply a disinhibitory one. Ultimately, to consolidate these initial findings related to offending behaviour, we need to supplement earlier studies with more systematic and rigorous experimental work with offenders. However, in addition to using offender samples and comparison groups in our work, I suggest we need to update our models of emotion, by incorporating advances in understanding from allied disciplines such as neuroscience, when developing explanatory models of crime.

How affective neuroscience can help us understand offender decision-making and behaviour

There have been numerous references in recent decades to the idea that emotion strongly affects decision-making in psychology, criminology, and neuroscience. However, traditionally cognition and emotion have been (1) seen as epiphenomenal (i.e. separate processes), and (2) studied with a focus on the end result (i.e. when a participant reports the labelling of a distinct emotion or set of emotions they have felt), rather than examining the dynamic and complex interplay of processes which results in these experiences (Scherer & Moors, 2019). Emotions are often described as impacting decision-making once it has ensued (e.g. Clore’s (1992) ‘affect as information’; Slovic et al.’s (1991) ‘affect heuristic’), despite many researchers providing evidence that the elicitation of emotion by various triggers initiates and drives the entire cognitive and behavioural episode (see Bechara and Damasio’s (2005) ‘somatic markers’; or Panksepp’s (1992) ‘valenced states’).

Criminological and correctional psychological explanations of offending behaviour have been hindered by the two traditional flaws noted in the previous paragraph. More recently, experimental studies using advanced fMRI brain scanning (Todd et al., 2020), have seen a movement towards studying cognition and emotion as a unified, embodied process. Emotion is seen as the interface between the environment and the individual’s experience (Mulligan & Scherer, 2012), with all events filtered through a lens of whether they are good or bad for us (Todd et al., 2020), also known as affective framing (Maiese, 2011). The emotional episode has become increasingly viewed as a complex, multi-staged process in the individual, and this process involves: an elicitation phase in which a stimulus is appraised; autonomic action preparation involving a shift in attention and orientation plus neural responses such as endocrine secretion; expressive behaviour in response to the stimulus (e.g. anger or happiness); a non-verbal feeling state; and finally, a verbal description or categorisation of feelings. All of these stages clearly involve cognition too, which is mostly automatic and only conscious towards the end of the process, if then (Scherer & Moors, 2019). These developments have shifted neuroscience research away from the reporting of and study of discrete emotions and their underlying neural substrates, towards the idea of valence (i.e. a continuous scale of positive (e.g. excitement) to negative (e.g. agitation) emotion, inviting either approach or avoidance behaviours).

These ideas have underpinned a theoretical movement in affective neuroscience towards an ‘enactive’ stance (Colombetti, 2014; Dent & Ward, 2022) meaning that humans are viewed as embodied organisms that exist in a needful, dynamic relationship with their environment (physical, cultural and social), driven by both first-order desires (reflexive bodily sensations) and second-order desires that are more intentional (Maiese, 2011). It is this embeddedness in their physical and cultural environment that will clearly differentiate (for instance) the perception of risk in an offender who has spent years transgressing normative moral codes to commit crime compared with a person from a well-educated, well-supported, crime-free background. These complex insights into the structure of emotion and cognition provide an innovative and fertile foundation from which to understand more clearly what is driving (criminal) behaviour. Taking the expertise perspective on residential burglary as an example, for some time we have known that burglars have heightened recognition of environmental cues to do with layout, access, security and potential gain when selecting properties (Clare, 2011; Nee, 2015; Wright & Decker, 1994). However, this information has not been precise enough to translate into effective ways to reduce crime, as noted in the statistics earlier. Learning how emotion is triggered by visual cues and frames critical decisions as a burglar (1) appraises streets, (2) chooses their target, and (3) undertakes their burglaries could be the key to understanding the tipping point resulting in a crime. In other words, the point at which reward outweighs risk at each stage of the crime. This is not only important for better predictive validity in our theories and knowledge, but can clearly underpin a more systematic neighbourhood crime prevention approach for burglary and many other crimes. It can also improve the success of our rehabilitation programmes by supporting those who are desisting from crime in recognising emotionally loaded triggers, early on in the offence decision-chain. The expertise paradigm embraces the concept of positive as well as negative emotion in the instigation and maintenance of offending behaviour (rather than simply updating current decisions to offend with memories of bad consequences). Accordingly, the neuroscientific approach to emotion provides a valuable addition to the expertise framework, with which to explore offending behaviour. What is needed are research technologies that are capable of providing insight into the complex cognitive and emotional processes involved in crime. I propose that new models of emotional valence, assimilated into the expertise model and measured using virtual reality technology can fulfil such a role.

How to incorporate this perspective into offender decision-making research: with virtual re-enactment

Arguably, the strongest current method for systematically, accurately and ethically studying the emotional component of offender decision-making is virtual re-enactment. This is because we know that many of the processes surrounding offending behaviour are not fully conscious and will not be triggered (in order to be studied) unless an individual is enacting the behaviour in question. A recent review of the ‘learning-by-doing’ literature indicates that VR has two unique attributes that facilitate unmatched triggering of mental schemas in memory, compared to other learning techniques (Johnson-Glenberg, 2018). These are: the sense of presence, i.e. the feeling of being there (Slater & Wilbur, 1997); and the sense of embodiment and agency associated with movement in three dimensions noted in the situated-action literature (Barsalou, 2008). This is because the development of episodic memory (memory for everyday events and the associated emotions) is deeply entwined with locomotion from infancy onwards. More precisely, the ability to encode and retrieve information from our memory is driven by the actions we can perform upon our environment and therefore the limitations/opportunities provided by our bodies (Glenberg, 1997). Re-enacting an event provides the optimal environment for experiencing all the visual, kinaesthetic and spatial cues needed to trigger episodic memory (Wilson, 2002).

We also already have evidence that by re-enacting a behaviour in real-time, VR methodology minimises the cognitive burden of memory retrieval and markedly improves engagement and disclosure in offenders and other marginalised groups (Kip et al., 2019; Meenaghan et al., 2018). Of particular relevance to the current discussion, VR has been shown to accurately evoke emotion. Slater et al. (2009) and Martens et al. (2019) report a number of studies which compare virtual experiences with real experiences of the same event, and find identical physiological and verbal reports. Van Gelder et al. (2019, 2022) have replicated their own studies which originally used written vignettes, in virtual reality. Comparing the two in these recent experiments, participants reported higher levels of realism and stronger feelings of fear and anger in VR than in response to the vignettes. This opens huge potential for theoretical development that predicts criminal behaviour more accurately at various stages of the crime.

Interviews have been shown to yield valuable information in a wide variety of fields, including investigative and therapeutic settings. But in the context of studying complex, automatic decisions and actions of offenders around the scene of the crime, they are subject to the well-established drawbacks of flawed memory retrieval (inferential errors, telescoping forwards, overestimation and bias; Tulving, 2002) and social desirability (Nee, 2010). We have also seen that without reinstating a valid environmental context, participants fail to accurately predict their behaviour when they imagine it in an emotionally ‘cool’ state (Loewenstein et al., 1997). Additionally, while verbalizations provide rich enhancements to our knowledge about behaviour, they are considered the very final, reflective stage of a series of complex cognitive and emotional steps (Ericsson, 2003; Scherer & Moors, 2019), many of which are automatic and below the threshold of consciousness (Treiber, 2014). If we are serious about gaining a much deeper insight into how the environment and individual interact to produce criminal behaviour, we have to reinstate the context of that behaviour, re-elicit that behaviour, and measure the emotional processes as well as the decisions that feed into actions.

It could be argued that ethnographic interviews at the scenes of recent crimes might reinstate the context even better than a virtual environment, but they are not experimentally controllable settings and do not facilitate the re-enactment of the offending behaviour, which allows the embodied cognition and emotion to unfold naturally in reaction to the environment. Controlled measurement of non-verbal emotional states would also be nigh impossible in the field. Reinstating the context of the crime in virtual reality, however, removes all of these problems. It allows for the ethical immersion of an individual (or group) in a realistic criminogenic environment, facilitates the triggering of ecologically valid behaviours (Martens et al., 2019) and the re-enactment of the crime in real time. This enactive approach gives us the potential to study for the first time, complex processes including appraisal, emotional states, shifts in attention and orientation, bodily action preparation and the resulting behaviour(s) of interest, in a tightly controlled manner as they happen (Cornet & Van Gelder, 2020). It therefore provides a sea-change in how we could study criminal behaviour, moving our enquiries onto a new plane that has the potential to provide unparalleled levels of insight. As noted earlier, in many cases it is not possible to verbally describe an action or experience until towards the end of that action or experience, and often times one is not aware at all that one has experienced something. No other method has the potential to let us examine these hitherto inaccessible processes in criminology (Meenaghan et al., 2018). As technology improves, our insights are likely to improve with them, but the final section below elucidates what is possible now.

Evidence from recent empirical studies with prolific offenders already indicates that VR methodology facilitates insights into the mechanisms underpinning criminal behaviour. For instance, convicted domestic abusers have been able to ‘embody’ their victims in VR scenarios and have consequently increased empathy towards them and understood their own aggressive behaviour more clearly (Seinfeld et al., 2018, 2023). Personality traits have been linked to intentions to aggress in response to virtual threats (Van Gelder et al., 2022). VR burglary studies have revealed new insights into expertise that were simply inaccessible without re-enactment. For example, never before were we able to observe (1) the appraisal and navigational strategies used by burglars during the crucial scouting phase in neighbourhoods, demonstrating efficient covering of ground and recognition of particular vulnerability cues; (2) the unique navigation and searching behaviour of the experienced burglar once inside the property, and the marked contrast in the choice of stolen items in comparison to other groups; (3) note the obvious, audible dismay of experienced burglars as they entered a baby’s nursery in the virtual house³ (Nee et al., 2019); or (4) observe burglars’ actual preferences towards different types of neighbourhood guardians (Van Sintemartensdijk et al., 2020), in comparison to control groups.

In order to more accurately measure the physiological aspects of emotions throughout the criminal episode, we can now add to the VR methodology for the first time, evidence-based cardiovascular and brain activity measures designed to assess not only arousal (which can be a rather blunt measure) but also emotional valence. In other words, we can begin to distinguish between positive, reward-laden emotions such as pleasure and excitement, and negative, risk-laden emotions such as anxiety, as offenders enact crimes in response to visual and auditory cues in the environment. Moreover, these measures are portable, allowing them to be used in various correctional settings as we go forward. For instance, visual fixations (which can be measured via eye-trackers integral to VR headsets) are a measure of attention, motivation and preference and in the early stages are outside conscious manipulation (Calvo & Lang, 2004), and spontaneous eye blinks are a relatively robust proxy measure for increased dopamine (the neurotransmitter most closely associated with pleasure and reward) in the brain (Jongkees & Colzato, 2016). Photoplethysmographic signals (changes in blood volume using an optical sensor) can be recorded via a simple ear or finger clip as an accurate and unintrusive measure of positive and negative emotion using cardiovascular arousal (Lee et al., 2019). Pilot work on these measures has started to provide proof of concept of the most efficient way to carry out this research with offenders, who are a vulnerable population and often difficult to access. Nevertheless, the evidence from previous VR studies already indicates that the gains in terms of protecting ordinary citizens, reducing costs in criminal justice systems, and rescinding criminal careers, will be reliant on new methods and technology. These will considerably outweigh what was possible in the past.

I argue that this methodological approach represents a unique opportunity to understand the emotional valence of automatically processed cues during a much larger segment of the offence decision-chain than ever before. Previous research methods in the field of offender decision-making have unearthed broad categories of cues used by offenders (mostly limited to target selection). However, by adding an understanding of emotional valence to the situational preferences of the experienced offender, we have the potential to move the field forward with considerable implications for crime prevention policy and rehabilitation practice. Such emotional states are also arguably proxies for core personal values (that can be addressed in rehabilitation) given their relationship to perceptions of well-being and possible threat. By capturing decision-making and behaviour as it happens, the virtual enactment method will facilitate insight into the value of different signals in the environment as they are processed by the offender. We can have clearer access to what really drives the offender (through positive emotion) and deters the offender (by increasing anxiety and reducing automaticity). We can parse out these key attraction/deterrent cues during: the (1) scoping/preparation phase of the offence; (2) target selection, and (3) the commission of the crime. This will give us the tools to improve situational crime prevention strategy by putting in place more effective deterrents and remove attractors at different stages of the criminal offence chain. For instance, with respect to burglary we might: reduce vulnerability at the neighbourhood/community level while the burglar is scouting the area; make residences look more risky and daunting to enter; and put in place effective, anxiety-inducing deterrents while the burglar is inside the property. Moving forward into a broader criminal justice arena, virtual re-enactments are currently being undertaken with convicted child sex offenders, to understand their automatic cue recognition and navigation around vulnerable environments. By adding emotional valence to these studies, we can envisage altering environments such as adjusting rewarding signage (to play areas/schools) and visibility of children in these areas to strangers. Similarly, understanding the hostile reconnaissance of the terrorist using virtual environments could lead to reducing the vulnerability of public places to terrorist events. By gaining insight into the most compelling and deterrent aspects of different phases of the crime, we can more effectively deter the sex offender, the terrorist, the street robber. Moreover, we can test out these hypothesised interventions in VR before rolling them out into communities.

We will also make much needed gains in offender rehabilitation, by understanding significantly more about the intersection of automatic cue processing and emotion, and aiding the desister to increase awareness of the signals in the environment (and the associated feelings and consequences) which are likely to result in a criminal act. This will afford them the ability to potentially rescind problematic choices at an earlier, less emotionally loaded, stage of the offence decision chain (Nee & Vernham, 2017). For instance, using mindfulness techniques within a virtual neighbourhood, a child sex offender could become more aware of the negative consequences of enacting the ‘seemingly unimportant decisions’ (Ward & Hudson, 2000) that will lead them to walk near a school when children are being released at the end of the day. Relatedly, the close relationship between emotions and values constituting personal identities will enable researchers and practitioners to identify and track shifts in values during the desistance process.

The virtual re-enactment method, and VR in particular, is of course, not without its limitations that could detract from the validity of VR research on emotion and crime until they are resolved. First and foremost, a virtual re-enactment is not identical to undertaking a real crime and can never represent the same stakes as committing a crime in real life. That said, it goes much further in reinstating the context in an ethical (for the victim) way in order to study online decisions and feelings than other methods (Meenaghan et al., 2018). Only time will tell precisely how valuable findings gleaned using this method will be in reducing crime. The ethics of asking offenders to undertake mock versions of crimes, especially if they are not in custody, is trickier and needs close monitoring as we go forward. Concerns about encouraging offenders to recidivate, or re-traumatising remorseful desisters should be taken seriously. Thus far research has found no indication of an increased desire to commit crime, but only indications of remorse in some (Nee et al., 2019) and increased victim empathy in others (Seinfeld et al., 2018, 2023), but research is in its infancy and researchers must continue to scrutinise this aspect. Another issue is that of the ability to identify individuals through their tracking data under some VR conditions. It is crucial that researchers adhere to strict protocols protecting individuals’ security given the vulnerability of offender populations (Van Gelder, 2023). Other issues such as cost of equipment and cybersickness are being resolved as technology improves.

Conclusion

In the 1980s the rational choice perspective furnished researchers with a valuable theoretical foundation with which to begin to study offender decision-making and behaviour at and around the scene of the crime. In order to improve its efficacy in terms of actual crime prevention, there have been attempts in recent years to enrich and update the perspective from a criminological and psychological point of view. The core of these changes has centred around identifying and examining more deeply the complex, multi-staged cognitive and emotional processes underpinning decisions and behaviour during the criminal episode as it happens in the moment, and so moves away from purely negative, consequentialist impacts on decision-making. This concurs with the enactive approach in affective neuroscience which emphasises that human behaviour is both embodied (environment interacting with both body and brain) and embedded (in a physical, cultural and social environment) with its emphasis on studying emotion and cognition as a unified process. The article aimed to highlight the critical extra explanatory value that can be added to our understanding of offender decision-making by incorporating breakthroughs from other disciplines, namely cognitive science and affective neuroscience, and calls for a more systematic, enactive approach which incorporates the impact of emotion as well as cognition, and which only virtual reality can ethically facilitate. The example of the development of the expertise model in offending behaviour and the enactive methodology that has grown beside it provides us with an invigorating and fertile way to explore criminal cognition, emotion and behaviour as a more interactive process. Doing this will provide a step-change in our understanding of what is driving criminal decision-making and help us to disrupt that process through the development of more successful crime prevention and reduction strategies. Moreover, the advent of virtual re-enactment will obviate the issues that restrained traditional research methods in criminology. It allows us to ethically observe criminal behaviour in progress by sufficiently reinstating the context of the crime so we can more directly examine the emotional and cognitive correlates of criminal behaviour.

Acknowledgements

The author would like to thank Zarah Vernham and Tony Ward for helpful comments on previous drafts of this manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 For example at about two million per annum in the US (Statistica, 2023); 450 k p.a. in England and Wales (Office for National Statistics, 2022).

2 There is evidence of considerable levels of petty offending behaviour in university populations (e.g. Garwood, 2011), but it is very unlikely that significant proportions would be the type of prolific offender that is found in prisons.

3 This had previously been put down to the lack of valuable goods in such rooms (e.g. Nee & Meenaghan, 2006) but was revealed in Nee et al. (2019) to be linked to a moral code not to enter young children’s rooms.