Comparative Outcomes of VR and In-person Interventions for Preschool-aged Children With Autism Spectrum Disorder

Autism Spectrum Disorder (ASD) is a neurodevelopmental condition characterized by differences in social interaction and communication, along with restricted or repetitive behavior patterns (American Psychiatric Association, 2022). These developmental differences typically emerge in early childhood and significantly impact early developmental trajectories and continue to affect individuals throughout life (Lord et al., 2018). Over the past two decades, the global prevalence has risen considerably, based on 2022 surveillance data from the U.S. Centers for Disease Control and Prevention (CDC) indicating that one in 31 eight-year-old children received an ASD diagnosis, almost a fivefold increase since 2002 (Shaw et al., 2025), not only reflecting improved diagnostic practices and greater societal awareness of early developmental differences (Maenner et al., 2023), but also highlights the need to support worldwide autistic individuals (Zeidan et al., 2022).

Social communication difficulties represent one of the most impactful challenges for autistic children (Kasari \& Rotheram-Fuller, 2007), and are associated with downstream consequences in academic functioning, peer relationships, and mental health (Bellini et al., 2007). A substantial proportion of autistic children have age-appropriate cognitive or language abilities, yet still display marked difficulties in social cognition, play skills, and emotion regulation (Lee \& Park, 2007; Liss et al., 2001). Differences in social engagement between autistic children with age-appropriate cognitive abilities and their typically developing peers become increasingly apparent during early childhood, including lower frequency of peer interaction, challenges in interpreting social cues, and difficulties maintaining conversations (McConnell, 2002; Barry et al., 2003). These challenges often become more pronounced upon school entry, when social demands increase and unstructured peer interactions become more complex (Koegel et al., 2013). Thus, preschool social skills intervention is viewed as a critical strategy for preventing later social isolation and school adjustment difficulties (Barnhill, 2007; Dawson et al., 2010; Landa, 2008).

Group-based social skills interventions are frequently used in early-intervention and clinical settings because they offer natural peer contexts, opportunities for observational learning, and embedded practice within dynamic social routines. (Barry et al., 2003; Capes et al., 2019; Gates et al., 2017). Systematic reviews and randomized trials report moderate effects for group-based social interventions for young autistic children (Freitag et al., 2016; Gates et al., 2017; Wolstencroft et al., 2018). Despite these benefits, many existing programs rely on discrete, theme-based instruction structures that target one skill at a time and often incorporate didactic teaching (Kasari et al., 2015; Radley et al., 2020; Tripathi et al., 2022). Such formats may limit opportunities for repeated, contextualized practice and may reduce flexibility in responding to children's moment-to-moment social engagement, constraining generalization and maintenance of gains (Babb et al., 2020). Access to early social-communication intervention is also inconsistent across regions and service systems, particularly in settings facing clinician shortages or limited availability of structured group programs. (Stahmer et al., 2005; Vismara \& Rogers, 2010). These challenges underline the need for feasible, ecologically valid intervention models that can increase children's opportunities for naturalistic social learning.

Two emerging approaches may address these challenges. Virtual reality (VR), particularly through head-mounted displays (HMDs), has become a promising platform for delivering social and communication skill interventions to autistic individuals (Karami et al., 2021). VR provides immersive, structured, and engaging environments in which autistic individuals can practice social scenarios with reduced real-world unpredictability (Parsons \& Cobb, 2016; Wallace et al., 2017). A systematic review has demonstrated that VR-based interventions can enhance attention, social skills, emotional skills, and communication abilities (Mesa-Gresa et al., 2018). In parallel, Naturalistic Developmental Behavioral Interventions (NDBIs) have become one of the most well-established evidence-based approaches for young autistic children, integrating developmental principles and applied behavior analysis within child-led, context-embedded activities that support social engagement, emotional regulation, and flexible learning across related skill domains (Sandbank et al., 2020; Schreibman et al., 2015). NDBIs have been adapted into group formats to enhance ecological validity and cost-effectiveness, while providing more opportunities for peer interaction and naturalistic social learning (Tiede \& Walton, 2019; Vivanti et al., 2017). Integrating NDBIs principles into VR may be particularly valuable, as VR environments can embed naturalistic interaction opportunities while maintaining consistency and controllability. However, empirical applications of NDBI-informed approaches within VR remain limited (Dechsling et al., 2022). To address this gap, our interdisciplinary research team developed an NDBI-informed VR social-skill intervention program for autistic preschool children without intellectual disabilities (ID), featuring immersive play-based scenarios and dynamic prompting tailored to children's behavioral cues (Wu et al., 2023; NSTC 112-2218-E-008-016). Preliminary single-group findings suggested improvements in children's social behaviors (Lin et al., 2024).

Although VR offers consistency and high practice density, its interactions remain constrained by preprogrammed structures. In contrast, in-person group programs allow clinicians to provide highly flexible, moment-to-moment adjustments in response to children's spontaneous behaviors and peer dynamics. Recent reviews indicate the scarcity of direct comparisons between VR-based interventions and equivalent in-person social-skills group programs, leaving it unclear whether the two modalities produce comparable therapeutic benefits (Dechsling et al., 2021). Understanding these differences is critical for determining how alternative delivery formats support engagement and learning in early intervention contexts, particularly as service systems explore more flexible or scalable models of intervention (Simacek et al., 2021). Building upon this rationale, the present study compares two delivery modalities: (1) the VR-HMD system and (2) an in-person, clinical psychologist-led group intervention directly adapted from the VR curriculum. The in-person version retained the session structure and therapeutic principle of the original VR curriculum, allowing a direct comparison of delivery format. In addition, because social-communication skills may be expressed differently across home and school contexts, the present study incorporated both caregiver and teacher reports to capture cross-contextual outcomes and informant-specific patterns, and to compare outcomes and clinical utility between VR-mediated and in-person group interventions to inform flexible and scalable service models for autistic preschoolers without ID.

This study adopted a randomized two-arm design to compare two delivery formats of a social-emotional skills intervention for autistic preschoolers. Participants were recruited from the pediatric and child psychiatry departments of a university-affiliated medical center in Southern Taiwan. Thirty-two children between 4 and 6 years 11 months of age who were clinically diagnosed or suspected to be autistic were enrolled following referral by licensed pediatric and child psychiatrists with DSM-5 criteria. Eligibility was determined through a multi-step diagnostic evaluation, including (1) clinical judgment by a licensed psychiatrist, (2) scores ≥15 on the Checklist for Autism Spectrum Disorder-Chinese version (CASD-C), and (3) meeting ASD classification on the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2). Cognitive functioning was evaluated using the Wechsler Preschool and Primary Scale of Intelligence, the Fourth Edition (WPPSI-IV), and only children with Full-Scale IQ and Verbal Comprehension Index (VCI) scores above 70 were included. Exclusion criteria included (1) significant sensory or motor impairments (e.g., visual or hearing impairments, severe physical disabilities), (2) known genetic or neurological conditions (e.g., epilepsy, fragile X syndrome, traumatic brain injury), or (3) behavioral profiles that would preclude participation in a group-based setting (e.g., severe aggression). After enrollment, participants were randomly assigned to either the VR-based intervention group or the in-person group intervention, with 16 children in each condition. The two intervention formats were matched in terms of session number, duration, content, and implementation of NDBI-informed strategies.

The outcome measures, including social and adaptive behaviors, were administered at baseline (Time 1, T1) and immediately after completion of the full-course intervention (Time 2, T2).

The intervention comprised eight weekly sessions, each lasting approximately 50 minutes, delivered through two modalities: a virtual reality (VR)-based format and an in-person group format. Both formats followed the same NDBI-informed curriculum and session scripts, ensuring equivalence in therapeutic goals, activity structure, prompting hierarchies, and reinforcement strategies. All sessions were organized around integrated, play-based activities, in which multiple core social-emotional competencies were targeted simultaneously, including social initiation and response (e.g., eye contact, calling peers by name, and seeking attention appropriately), behavioral regulation (e.g., turn-taking, impulse control), and emotion recognition and regulation (e.g., identifying feelings, responding to distress).

Across both modalities, sessions used developmentally appropriate, game-based activities that embedded opportunities for naturalistic social learning. Activities included ball-toss games, card games, block-building tasks, and a final board-game review. A dice-rolling routine was used to select activities to increase engagement and give children a sense of autonomy. Naturally occurring events, such as disagreements, rule violations, waiting, or emotional frustration, were used as teachable moments to model emotional awareness, flexible thinking, and simple problem-solving strategies. Systematic prompting and prompt fading were applied across all activities to support increasing independence. Caregivers received a structured home-practice worksheet after each session to facilitate skill generalization across daily routines.

Children in the VR intervention group participated in an immersive head-mounted display program originally developed by Wu et al. (2023) and adapted for the present study. The system presented cartoon-based classroom- and playground-like environments in which children interacted with peer avatars during structured social tasks. All instructional prompts, feedback, and graded support were preprogrammed based on NDBI principles. The system incorporated multimodal sensing functions, including eye-tracking to monitor children's visual attention and respiration monitoring to support deep-breathing practice during emotion regulation exercises. These signals inform system-generated prompts and graded challenge levels. Sessions were supervised externally by a trained graduate student in clinical psychology, whose role was limited to assisting children in operating and adapting to the VR equipment and ensuring safety. All instructional cues, prompts, and feedback during the activities were provided solely by the VR system. At the end of each session, caregivers received a brief summary of the session goals to support their understanding of the targeted skills by the graduate student. This feedback was concise and focused on the activity objectives, as teaching prompts and behavioral supports were fully delivered by the VR system.

For the in-person intervention group, the same core curriculum was delivered, with each session involving four preschoolers and two trained group leaders: a licensed clinical psychologist and a well-trained graduate student in clinical psychology, both receiving weekly supervision from a senior licensed clinical psychologist who also served as the primary developer of the VR curriculum, ensuring consistency between the VR-based and in-person adaptations. Group leaders implemented individualized scaffolding based on children's performance levels, using a range of naturalistic teaching strategies including live modeling, verbal and visual prompts, praise, and contingent feedback. They also provided moment-to-moment adjustments based on children's spontaneous behaviors and peer dynamics. At the end of each session, caregivers received individualized feedback from group leaders regarding their child's participation and social-emotional behaviors during the session. This feedback also allowed caregivers to raise questions, and group leaders provided practical suggestions to support skill generalization at home.