Authenticity in Virtual Reality for assessment and intervention in autism: A conceptual review

Highlights

• Virtual Reality (VR) technologies are ‘potentially transformative’ for education.

• It is assumed that VR is veridical i.e. authentic and realistic.

• It is argued that VR may be especially useful for individuals with autism.

• Challenges to the assumption of veridicality for VR for autism are highlighted.

• Systematic research is needed to reveal the factors that influence responding.

Abstract

Virtual Reality (VR) technologies have shown potential for learning and assessment for children, adolescents, and adults with autism. Much of the research in this area has taken a conceptual stance of veridicality; that is, that VR offers promise because it can provide authenticity and levels of realism alongside stimulus or environmental control, or both, which may first facilitate learning and the generalization of skills to the real world, and secondly can provide experimental contexts with strong ecological validity for assessment. This conceptual review raises questions about the assumption of veridicality of VR for autism research by examining research literature that has used VR to support learning and to investigate social responding. In so doing, it provides a framework for examining the assumed relationship between virtual and real contexts in order to highlight particular features of design and interaction, as well as background characteristics of participants, that may help or hinder learning and understanding in virtual environments. The conclusions suggest there is a need for the field to systematically examine the different factors that influence responding in VR in order to understand when, and under what circumstances, the responses of individuals with autism can be considered appropriately authentic. There are also opportunities for thinking more radically about research directions by focusing on the strengths and preferences of people with autism, and promoting more participatory and inclusive approaches to research.

Introduction

With the acquisition of the Oculus Rift headset by Facebook in 2014, Virtual Reality (VR) technologies have once again been making news headlines. Facebook Founder and CEO, Mark Zuckerberg, claims that ‘Virtual Reality will change our lives’ (Blunden, 2016). More specifically, Albert ‘Skip’ Rizzo, Director for Medical Virtual Reality at the Institute for Creative Technologies at USC Davis, argues that the new, lightweight, and inexpensive headset could be ‘transformative’ for rehabilitation, educators, and psychologists (Robertson & Zelenko, nd), because VR can provide access to highly realistic, and motivating online contexts and experiences for learning and assessment. Such enthusiasm echoes that of researchers considering the potential of VR well over a decade ago for studying and supporting social interactions (e.g. Blascovich et al., 2002).

Fundamental to understanding human behaviour, such studies usually present the dilemma of trading experimental control in the laboratory setting with real world ecological validity. However, VR technologies have the potential to avoid this trade-off, providing both experimental control and realistic scenarios with good ecological validity. As a consequence, VR technologies have been argued to offer particular promise for supporting learning for children and adults with autism who may find social interactions difficult or may respond differently to ‘typical’ social overtures. Indeed, the majority of published interventions in autism educational research are targeted at social-communicative difficulties (Parsons et al., 2011), underscoring why VR may have particular application for this group. However, despite nearly twenty years of research, the potential of VR for autism education still remains an aspiration rather than a reality. This paper provides a conceptual review of the field and argues that a rethink may be needed about paradigmatic assumptions regarding the role of VR technologies for assessment and intervention in autism research.

Blascovich et al. (2002) argue that immersive virtual environment technology (IVET) could transform the study of social behaviours through being able to carefully control and simulate different aspects of social interactions. They suggest that IVET would enable researchers to determine ‘the critical aspects of successful and unsuccessful social interactions’ (p. 121), because the technology provides experimental control in ecologically valid, and replicable contexts. Underpinning these claims of transformation lies the fundamental assumption of veridicality; viz. that the experiences within VR technologies are authentic and realistic such that people will behave and respond in a similar way in virtual worlds as they do in the real world, thereby enabling generalization from the former to the latter (Yee, Bailenson, Urbanek, Chang, & Merget, 2007).

People with an Autism Spectrum Disorder (ASD) comprise a large group of users for whom VR has been argued to be of especial interest (e.g. Parsons & Mitchell, 2002). According to the most recent definition (American Psychiatric Association (APA), 2013), ASD is diagnosed on the basis of pervasive difficulties with social communication and interaction, coupled with restrictive, repetitive, and stereotyped behaviours. Approximately one in 100 children in the UK (Baird et al., 2006), and one in 68 (Baio, 2014) in the US receive a diagnosis of ASD, and the personal and familial impacts of an autism diagnosis are significant. For example, despite over thirty years of intervention research, the social, economic, employment, and accommodation-related outcomes for adults on the autism spectrum remain poor (Billstedt et al., 2005, Eaves and Ho, 2008, Howlin et al., 2004). Consequently, finding more effective ways to educate and support individuals with autism to improve outcomes and decrease reliance on specialist provision remains a research priority for individuals and their families (de Bruin et al., 2013, Pellicano et al., 2014).

There are many research reviews of the field (see Table 1 for a summary): some focusing on interactive technologies in general, but not VR (e.g. Boisvert et al., 2010, Ramdoss et al., 2012, Ramdoss et al., 2011, Shane et al., 2012); others including VR alongside a range of interactive technologies (Grynszpan et al., 2014, Ploog et al., 2013, Reed et al., 2011, Wainer and Ingersoll, 2011, Wass and Porayska-Pomsta, 2014), and others focusing on VR technologies specifically (Bellani et al., 2011, Georgescu et al., 2014, Parsons and Cobb, 2011, Rajendran, 2013). All suggest that technologies have valuable potential for helping to meet this research priority in relation to education and assessment for children and adults with ASD. Indeed, there is some evidence (discussed further below) that VR can provide an effective platform within which social interactions and communication of individuals with ASD can be controlled, explored, examined, and supported. However, existing educational technology reviews (Table 1) have mostly reported on the effectiveness of technologies for supporting learning outcomes, regularly reporting similar limitations in the evidence base as a result of small sample sizes and research design weaknesses. Therefore, another review that considers effectiveness would not contribute much new knowledge to the field. Instead, this paper aims to take a different perspective and argues that the fundamental assumption of veridicality of VR technologies in terms of recreating authentic assessment and learning contexts for people with autism may be open to question. Given the renewed interest in VR technologies, it is timely to revisit this fundamental assumption with a view to shaping the future research directions in this area.

Blascovich et al. (2002) propose that immersive VR technologies offer considerable promise to researchers by overcoming longstanding threats to reliability and validity in social research. They define a virtual environment (VE) as ‘synthetic sensory information that leads to perceptions of environments and their contents as if they were not synthetic’ (p. 105); in other words, the person using the VE has a psychological sense of the representational reality of the experience. In addition, Blascovich et al. (2002) define an immersive VE (IVE) as:

• one that perceptually surrounds the individual. Immersion in such an environment is characterized as a psychological state in which the individual perceives himself or herself to be enveloped by, included in, and interacting with an environment that provides a continuous stream of stimuli (p. 105).

According to Blascovich et al. (2002), these qualities, or affordances, of IVEs mean that the challenge of authentically (re)creating ecologically valid contexts in which human behaviour can be studied can be substantially overcome, viz: 3D scenarios and contexts approximating ‘real life’ can be created; these identical scenes can be experienced by multiple users without necessarily a feeling of pretence or artificiality; collaborative VEs can support the interaction of multiple users (participants, researchers, and confederates) at the same time, and so social interactions can be manipulated and studied; experimental conditions can be replicated for different studies with high fidelity; and participants can be more easily randomly assigned to conditions of the experiment, thereby improving generalizability. The combination of controlled conditions with high ecological validity within VR therefore offers a very powerful tool with which to study human behaviours in real time, dynamic interactions. Although Blascovich et al. (2002) specifically discuss immersive VEs, the affordances noted here could apply equally to non-immersive VEs (such as those presented on desktop or laptop computers). While the degree of immersion may differ between mode of presentation and this may influence the sense of presence that users experience (see below), the fundamental ideas underpinning this review could apply to any hardware configuration that allows users to engage with 3-D VEs.

The purpose of this article is not to review the substantial research in VR that has taken place to explore such interactions. Nevertheless, it is important to emphasize that real world social conventions and behaviours are regularly anticipated, and imported, within interactions in VEs by many different groups of users (Bailenson et al., 2003, Parsons et al., 2007, Schroeder, 2002, Slater, 2009, Yee et al., 2007). However, it is also well-established that such expectations and implementation of behaviours are influenced by the individual characteristics of the users, as well as specific features of the technology, and how information is displayed and interacted with. For example, responses in VR are known to be influenced inter alia by the personality characteristics of the user (Alsina-Jurnet and Gutiérrez-Maldonado, 2010, Hammick and Lee, 2014, Sacau et al., 2008); their prior experience of technology (Richardson, Powers, & Bousquet, 2011), their familiarity with the interface (Waller, 2000), and their expectations of the technology (Garau, Slater, Pertaub, & Razzaque, 2005). Indeed, Waller (2000, p. 316) concludes that the scale of individual differences shown in responses within a VE ‘are large enough to be worthy of careful scientific scrutiny’.

These individual and technological factors contribute to the extent to which users feel a sense of ‘presence’ within the VE, that is, ‘the sense of being caught up in the representations of virtual worlds’ (Jacobson, 2001, p. 653). There is a substantial literature that discusses, explores, and critiques the concept of presence that is beyond the scope of this article, but there is broad agreement that presence is fundamental to how VEs are experienced and, therefore, whether and how behaviours and interactions take place (Riva et al., 2014, Sanchez-Vives and Slater, 2005). For example, Fox, Christy, and Vang (2014) discuss the factors that influence the extent to which individuals experience presence in VEs to the point that they are persuaded by their content, providing examples from health, education, advertising, and work collaboration. The extent to which someone is persuaded by their experiences in VEs depends on the degree of involvement or engagement that they feel with the content (Boucenna et al., 2014, Witmer et al., 2005). Consequently, levels of engagement are ‘inextricably linked with instructional power’ (Mineo, Ziegler, Gill, & Salkin, 2009; p. 185) and, therefore, the likely success of a VE for supporting learning and intervention (Mikropoulos & Natsis, 2011).

In addition to the characteristics of the users, features of the technology contribute to the sense of presence and are known to influence how participants respond and behave within VEs (Bente, Rüggenberg, Krämer, & Eschenburg, 2008). These features include the virtual characters (avatars) within the VE: how anthropomorphic they are; their behavioural realism; the interaction between these two features (Georgescu et al., 2014, Nowak and Biocca, 2003, Vinayagamoorthy et al., 2005); and whether the other characters encountered in a VE are controlled by humans (human-avatars) or by computers (agent-avatars) (Blascovich et al., 2002, Nowak, 2004). Other features are: how the user interacts with the VE, for example through a joystick, head-mounted device, or desktop computer (Mikropoulos and Natsis, 2011, Santos et al., 2009); and the sense of agency that is experienced by participants as a product of these technology features (Riva, Banos, Botella, Wiederhold, & Gaggioli, 2012). Indeed, Riva and Mantovani (2014) emphasize that presence and agency are directly related within experiences of using VEs such that: ‘presence is a core neuropsychological phenomenon whose goal is to produce a sense of agency and control: I am present in a real or virtual space if I manage to put my intentions into action (enacting them)’ (p. 10). Consequently, judgements about, or behavioural indicators of, presence are vital for understanding whether a particular form of media has the potential for supporting realistic actions and influencing future behaviours, i.e. can be useful as a basis for promoting memorable experiences (Sylaiou et al., 2013) and encouraging learning (e.g. Mikropoulos and Natsis, 2011, So and Brush, 2008).

The assumed veridical nature of the relationship between real and virtual behaviours and experiences, as well as strong stimulus control within VEs, has prompted researchers to use VR technologies for physical and psychological educational intervention and rehabilitation, with success: for example, for people with motor disabilities to practise physical movements (Holden, 2005); in psychotherapy for phobias and social anxiety (Gega, White, Clarke, Turner, & Fowler, 2013; Riva, 2005); for treatment of symptoms of psychosis (Freeman, 2008); for patients following a stroke (Weiss, Naveh, & Katz, 2003); and for daily living, cognitive, and social skills of people with intellectual disabilities (Standen & Brown, 2005). Consequently, there is evidence that VEs can, and do, represent authentic, realistic, and plausible scenarios and social encounters that both reflect and support real world conventions, understanding and behaviours for a range of different user groups (cf. Blascovich et al., 2002).

It is perhaps not surprising that the promise of veridicality of VEs has proved alluring for research in the field of ASD. Given the core diagnostic impairments of social communication and interaction (APA, 2013), and the substantial impacts of the disorder, authors have suggested that VR technologies have significant potential for assessment, training and education for individuals with autism (Goodwin, 2008, Parsons and Mitchell, 2002, Trepagnier, 1999), precisely because social scenarios and encounters can be carefully designed and controlled. Indeed, in reviewing research that has explored the use of VR for individuals with autism, Bellani et al. (2011) use a definition of VR that assumes veridicality, while also emphasizing stimulus control (emphasis added): ‘VR is a simulation of the real world based on computer graphics [which] can be useful as it allows instructors and therapists to offer a safe, repeatable and diversifiable environment during learning’ (p. 235). Similarly, Parsons and Mitchell (2002) suggest that VEs might be particularly useful for the field of autism research because of the capacity for role play: ‘responses can be practised in realistic settings in the absence of potentially threatening and frightening real-world consequences’ (p. 438).

However, Parsons and Cobb's (2011) review of the field raised two important questions that relate to the assumption of veridicality: (1) ‘to what extent do 3-D images, and the capability of moving around 3-D space, matter for helping children to learn, and in supporting transfer of learning between virtual and real contexts?’ and (2) ‘…what are the special and unique affordances of these [VR]…technologies for supporting learning for children on the autism spectrum?’ (p. 363). They concluded that there is a need to consider these aspects more directly, and in more detail, if we are ‘…to understand how to use the features of VR to best support learning’ (Parsons & Cobb, 2011; p.362). This review offers a framework below for making more explicit the assumed relationship between virtual and real contexts in order to highlight particular features of design and interaction, as well as background characteristics of participants, that may help or hinder learning and understanding in VEs.