Virtual Reality Training of Safety and Social Communication Skills in Children with Autism: An Examination of Acceptability, Usability, and Generalization

Abstract

Individuals with autism spectrum disorder (ASD) can struggle to acquire social, communication, and safety skills. Many of these skills can be targeted in individualized behavior analytic instruction. However, some skills can be challenging to teach given the difficulties associated with reconstructing a real-world scenario within a learning session. Virtual reality (VR) has emerged as a promising technology that can help people with ASD practice these types of skills in an immersive environment. VR is an emerging technology, and more research is needed to determine its efficacy as well as its impact on variables such as client indices of happiness and social validity. In this study, we successfully taught three children with ASD three different skills using a VR treatment package that consisted of VR, prompts, and reinforcement. Prior to teaching these skills, we included a cooperation phase with the intent to increase acceptance of VR equipment as needed. We found that each of the three participants accepted the equipment and VR sessions without the need for additional training. In all cases, the skills the participants acquired in the VR platform were maintained and generalized to the natural environment. Participants demonstrated indices of happiness when engaged with the VR software and parents and clinical staff ranked the VR software positively. Results are discussed in terms of the use of the VR treatment package in intervention and future research for similar technologies.

According to estimates from the Centers for Disease Control’s Autism and Developmental Disabilities Monitoring (ADDM) Network, about 1 in 36 children are identified with autism spectrum disorder (ASD). Individuals with ASD often demonstrate deficits in social communication and interactions (American Psychiatric Association, 2013). These deficits include but are not limited to difficulties reading social cues (e.g., eye contact, gestures, body language), challenges with holding a conversation (e.g., back-and-forth exchanges; initiating, continuing, and maintaining conversations; changing topics and interrupting if necessary), and other social engagement (e.g., peer play, engagement with peers, cooperation with tasks). As a result, these individuals are at an increased risk of social isolation (Bauminger & Kasari, 2000), bullying, wandering, and poor interactions with authority figures and community members.

The science of behavior analysis can be used to help individuals with ASD develop these skills. Behavior analysts can use techniques such as discrete trial training, modeling, video modeling, behavioral skills training, social stories, social scripts, and role-playing to establish social, communication, and safety skills in people with ASD (Lerman et al., 2016). Although often effective, these interventions may lead to limited generalization to everyday life experiences (Bottema-Beutel et al., 2018; Gates et al., 2017). Skills are taught in isolation in a therapeutic context that may not incorporate outcomes influenced by relational dynamics (Bolis et al., 2017; Pasqualotto et al., 2021) or situational variables. Even when these techniques are utilized and generalization is specifically targeted, the arrangements may require extensive resources to implement. For example, when teaching children abduction prevention skills, instructors may need dozens of individuals to be involved in the training to establish a skill that eventually generalizes to the natural environment (Levesque-Wolfe et al., 2021). In recent years, virtual reality (VR) technology has emerged as a promising tool for enhancing various aspects of education and therapy, particularly in the context of ASD. The use of VR may provide naturalistic arrangements for more effective role play-based training (Beaumont & Sofronoff, 2008; Kourtesis et. al., 2023), particularly when multiple trials are needed, and the scenarios are challenging to construct without the help of this type of technology. Applied behavior analysis has demonstrated the effective adoption and integration of technology advancements into practice, such as including biological measures (Crowley-Koch et al., 2013) and distraction goggles (Leopardi et al., 2023), but few studies have demonstrated how a VR treatment package can be used as a supplemental teaching tool to facilitate the fast and safe acquisition of skills that generalize to the natural environment.

The goal of skill building with the use of VR is to provide a safe space in a simulated environment where the user can practice and develop skills without the added complexities that may occur when practicing in real-world contexts (Mesa-Gresa et al., 2018). Some integration of the use of VR includes VR-based social interaction training (Beach & Wendt, 2014; Clay et al., 2021; Didehbani et al., 2016; Ke & Im, 2013; Kinsella et al., 2017; Mitchell et al., 2007), VR social skills and executive functions (Kourtesis et al., 2023), the effectiveness of using VR in teaching street crossing (Josman et al., 2008), and joint attention (Ravindran et al., 2019). Although there is evidence that using the VR platform has been effective in teaching individuals with ASD various skills, it is still considered an emerging technology, warranting further research to better understand its efficacy, usability, and acceptability in individuals with ASD (Bradley & Newbutt, 2018; Glaser & Schmidt, 2022; Kourtesis et al., 2023; Mesa-Gresa et al., 2018).

The purpose of the current study was to, first, determine the effectiveness of a VR treatment package in teaching various skills to children with ASD; second, examine indices of happiness of children while using the VR equipment; third, assess the maintenance of skills established with the use of the VR treatment package; fourth, assess the generalization of skills acquired in the VR platform to the natural environment, and last, assess the social validity of VR among clinical staff and parents.

Method

Participants and Setting

Participants were children with ASD who were currently receiving services through a multistate ABA provider organization and were between the ages of 5 and 8 years old. Individuals were excluded if they had any of the following preexisting medical conditions: a history of seizures or photosensitive response on EEG, eye movement impairments such as strabismus, migraines, or susceptibility to motion sickness. Two females and one male participated in the study. Jayden was a 5-year-old male diagnosed with ASD. Jayden spoke in short sentences and could make comments about events and people in his environment and answer short “wh-” questions. Arya was an 8-year-old female diagnosed with ASD. She understood most words and phrases as a listener and could speak in short sentences. Basma was a 7-year-old girl diagnosed with ASD. Much of her vocal verbal behavior was delayed echolalia, consisting of conversations in the mirror. She followed simple and multistep instructions and responded to spoken words as a listener.

All study sessions were conducted in the location (home, center, school, or community) where ABA services were typically provided for that participant. Natural environment generalization assessments were conducted in places where the participant did not typically receive services (such as the park, playground, or shopping center).

Materials

We used a VR platform called Floreo (https://floreovr.com), specifically designed for learners with ASD, attention deficit hyperactivity disorder, and anxiety. The Floreo system and configuration used in this study utilized an iPhone and iPad both running Floreo's iOS application. The software application on the iPhone leverages the Google Cardboard VR standard and was placed in a Google Cardboard compliant goggle to create a stereoscopic display for its user and enabled a three degree of freedom VR experience. When executed on the iPad, the software application displays the Floreo coaching application.

When used to deliver a VR therapy session, the iPad user selects a lesson and starts the lesson, which begins an interaction over Wi-Fi to pair with the corresponding iPhone to create a linked experience enabling the iPad user to view the VR user's environment as well as provide orchestration abilities.

Each participant had one subscription and one learner profile. Other materials consisted of preferred items and activities, Excel spreadsheets, and data sheets. Three lessons were chosen for all three participants based on their goals for intervention and skill deficits: safely crossing the street, listener identification of animals given two features, and joining a conversation.

Response Measurement

The primary dependent variable for skill acquisition was the percentage of correct responses to steps within the task analysis. We defined correct responses as the participant producing a correct vocal, gestural, or other response as required by the step in the lesson within 3 s of the stimulus presentation, without verbal or other prompts from the interventionist. Incorrect responses were defined as any response other than the correct one or no response within 3 s of stimulus presentation within the VR system.

Secondary dependent variables during the skill acquisition phase consisted of indices of happiness, defined as the participant (1) laughing or smiling; (2) communicating approval about the VR session via vocal statement or sign language; and (3) refraining from engaging in challenging behavior or negative vocalizations; and indices of unhappiness, which were scored if the participant (1) grimaced or frowned; (2) communicated disapproval about the VR session via vocal statement or sign language; (3) engaged in challenging behavior in the form of aggression, self-injury, or property destruction; or (4) engaged in negative vocalizations. We based the definitions for the indices of happiness and unhappiness on an evaluation conducted by Phipps et al. (2022).

Dependent variables analyzed throughout the cooperation phase of the study included equipment acceptance, defined as allowing the interventionist to place the VR headset properly on the participant’s head and secure the strap without engaging in challenging behavior or verbally or physically refusing the equipment. VR-session acceptance was defined as keeping the VR headset on for the entire duration of the VR session (3-min minimum per trial), and VR session engagement was defined as the number of seconds the participant was engaged in the VR session before removing the headset or asking for the VR session to end.

Interobserver Agreement (IOA)

The VR system automatically collected data via the iPad while the coach (i.e., the person conducting the session—either a board certified behavior analyst [BCBA], registered behavior technician [RBT], or technician) entered data in real time during the session. An independent observer recorded primary data collection using an Excel data sheet either live or from a video recording. Secondary (IOA) data collection occurred for 89.9% of all sessions across all phases and participants. The second independent observer recorded data on the participant’s behavior either during the live session or via video recording. During data recording, observers recorded independently, blind to the other observer’s data collection. We calculated total count IOA for an average of 97% of sessions (Participant 1, Jayden), 83% of sessions (Participant 2, Arya), and 85% of sessions (Participant 3, Basma) across all conditions. An agreement was scored if the same responses were recorded by both the primary and secondary observers on the same trial. Agreement was calculated by dividing the number of trials with agreements by the total number of trials and multiplying by 100. Mean IOA was 96.6% for Jayden (range: 80%–100%), 97.8% for Arya (range: 89%–100%), and 100% for Basma.

Procedural Fidelity

We collected data on procedural fidelity across all phases and for all participants. The steps of procedural fidelity can be found in Appendix B. We collected procedural fidelity data for an average of 97% of sessions (Participant 1, Jayden), 83% of sessions (Participant 2, Arya), and 85% of sessions (Participant 3, Basma) across all conditions. Mean procedural fidelity for Jayden was 99.5% (range: 81%–100%), 100% for Arya, and 99.9% for Basma (range: 98%–100%).

Design

We used a multiple baseline across behaviors (i.e., skill targets) design (Kazdin, 2011) to evaluate the effects of the VR treatment package on the acquisition, maintenance, and generalization of safety and social skills across three participants.

Procedure

Cooperation Phase

We included a cooperation phase with the intent to increase acceptance of VR equipment as needed. We designed the cooperation phase to consist of a baseline and an intervention phase. However, the cooperation intervention phase was never needed, given the participant’s positive baseline responses. One session consisted of three trials. Trial duration varied per activity (3 min minimum per trial).

Cooperation Baseline

We began the cooperation baseline session by pressing play on the VR lesson prior to placing the headset on the participant. They were able to see the VR visual immediately upon placement of the equipment. We provided the general and informative instruction: “We’re going to try something new today and use this headset to practice learning some new skills. I’m going to place it on your head.” If the participant verbally protested (e.g., “No thank you.”) or physically stopped the interventionist from placing the headset on them (e.g., pushed the headset away), we would have replied, “That’s okay, we don’t have to” and moved on to the next trial. However, all participants accepted the VR headset during each session (i.e., vocal protest or physically stopping the individual from placing the headset on did not occur). After the participant allowed the headset to be placed on their head, we started a VR practice session that consisted of calming and emotional regulation lesson choices (e.g., taking deep breaths while watching a train move on a track, imitating an animated yoga instructor, watching snow fall in an animated forest) and allowed them to wear the headset for as long as they desired. We collected data on the number of seconds the client engaged in the VR session before removing the headset or asking for the VR session to end. A break was provided between each VR session by removing the headset and re-presenting the instruction, “we are going to try something new” before presenting the next VR session. We conducted at least two baseline sessions with each participant. Participants moved into the skill-acquisition baseline following two consecutive three-trial sessions in which acceptance of the equipment and the VR session were both at 100%. All participants accepted the VR equipment and sessions (i.e., they engaged with the activity without removing the equipment for a minimum of 3 min).

Rule Statement (Basma only)

Basma requested alternative VR activities frequently during the cooperation baseline phase, even though she met the criteria for equipment acceptance (i.e., allowing equipment to be placed on her head) and VR-session acceptance (i.e., engaging with an activity for at least 3 min). We recorded the duration of her engagement with the activities; her average engagement with the pre-selected lesson was 3 min before she requested an alternative VR activity. We noted that we would need her to engage with the lesson for longer periods of time to participate with the three targeted skills and that her vocal requests for alternative VR activities may interfere with learning. Therefore, prior to the skill acquisition baseline phase, we conducted a separate probe session. During this probe session, we presented the rule: “We will take turns. First complete the entire activity for ‘my choice’ then you may select an activity to complete for ‘your choice.’” We then presented an activity from the cooperation phase to see if she would engage with the entire activity before requesting to finish and make a choice. She completed an entire activity on the first trial, so we then moved to the baseline phase for the targeted skills. The rule alone was successful in increasing her time with the activity and minimizing verbal requests for alternative VR activities prior to completing targeted skill acquisition activities.

Skill Acquisition Phase

The skill acquisition phase was used to evaluate the effects of the VR treatment package. The skill acquisition phase consisted of baseline and intervention. For both phases, one session consisted of one lesson or one full progression of the steps in the task analysis for each skill. Each step of the task analysis was considered one trial (task analyses ranged from 22 to 33 steps).

Skill Baseline

Before collecting baseline data, the first author viewed the targeted lesson and related materials and created a task analysis of the steps involved in the successful use of the skill targeted in that lesson (see Appendix Table 3). This task analysis was created through the first authors’ viewing of the lesson, and utilization of the description of the lesson and the behavioral objectives located within that lesson and created by the VR platform. These detailed steps were accessible within the learner’s profile on the VR system dashboard, “reports.” As an example of a lesson titled “Invite a peer to sit,” the steps included: Step 1: Wave to peer to invite him over, Step 2: Invite peer to sit at the cafeteria table. Upon initiation of the session, the interventionist obtained all materials needed for the lesson (iPad, headset, iPhone, preferred items) and presented the instruction: “We’re going to practice learning some new skills. I’m going to place this headset on your head, and we’ll get started.” We then started a lesson on the iPad. The interventionist answered the questions as prompted by the VR system within the lesson (e.g., Did the learner wave to the peer to invite him over?). We did not provide any additional prompts or reinforcement for correct or incorrect responses during the baseline phase.

Table 3.

VR Protocols

Street Crossing	Conversation	Listener: Two Features
1. Individual is oriented toward the street 2. Looks to the left 3. Looks to the right 4. Looks at the signal (crosswalk sign) 5. Identifies “don’t walk.” 6. States “push the button.” 7. Looks at the button 8. Presses the button 9. Waits for the signal to change 10. Looks at the signal 11. Identifies “walk” 12. Looks to the left 13. Looks to the right 14. Looks back to the left 15. Looks back at the signal 16. Looks across the street (until device moves learner across the street) 17. Turns around 18. Individual is oriented toward the street 19. Looks to the left 20. Looks to the right 21. Looks at the signal 22. Identifies “don’t walk.” 23. States, “push the button.” 24. Looks at the button 25. Presses the button 26. Waits for the signal to change 27. Looks at the signal 28. Identifies “walk” 29. Looks to the left 30. Looks to the right 31. Looks back to the left 32. Looks back at the signal 33. Looks across the street (until device moves learner across the street)	1. Looks at the individual speaking 2. Waits quietly while the individual is speaking 3. Looks at the second individual speaking 4. Waits quietly while the individual is speaking 5. Looks at the third individual speaking 6. Waits quietly while the individual is speaking 7. Responds to the question asked by the instructor 8. Looks at the individual speaking 9. Waits quietly while the individual is speaking 10. Provides an on-topic statement 11. Looks at the individual speaking 12. Waits quietly while the individual is speaking 13. Provides an on-topic statement 14. Looks at the individual speaking 15. Waits quietly while the individual is speaking 16. Provides an on-topic statement 17. Makes an exit statement (e.g., “nice talking with you, bye”)	1. Looks at Emma 2. Continues looking at Emma while she is speaking 3. Waits for the instruction “go” before looking away from Emma 4. Orients toward the first item requested to identify 5. Orients toward the second item requested to identify 6. Repeat steps 1–5 until 5 trials are completed (10 items are identified)

Skill Treatment

Prior to treatment, participants must have met the mastery criteria for the cooperation phase. Before each session, we obtained all materials needed for the lesson (iPad, headset, iPhone, preferred items). We then presented the instruction: “We’re going to practice learning some new skills. I’m going to place this headset on your head, and we’ll get started.” Next, we began a lesson on the iPad. We read the prompts provided within the VR system (see Appendix Table 2 for specific prompts per skill targeted). We provided 3 s for the participant to respond to the stimulus in VR (e.g., a person walks over close to the lunch table, allowing the opportunity for the learner to invite them over to sit). If the learner did not respond within 3 s or responded incorrectly, we used least-to-most prompting to prompt the correct response. For vocal responses, the hierarchy included partial-echoic to full-echoic prompts; for nonvocal responses, the hierarchy included vocal, model, partial-physical, and full-physical prompts. If the learner responded correctly, we provided praise. We answered any questions as prompted by the VR system to record data on the lesson within the platform (e.g., Did the learner wave to the peer to invite him over?). We moved through each step in the lesson until the lesson was complete. At the end of each lesson, the therapist removed the headset, and provided a 1-min break before initiating the next lesson. Each skill (lesson) was considered mastered when the participant provided 100% correct responses to all steps in the task analysis for that skill across a minimum of two consecutive sessions.

Table 2.

VR Prompts Assessed for Fidelity

Street Crossing	Conversation	Listener: Two Features
1. “We have to be careful when crossing the street.” 2. “Are cars coming left?” 3. “Are cars coming right?” 4. “Look at the signal, does it say walk or don’t walk?” 5. “How can we change the signal?” 6. “Look at the button” 7. “Press the button” 8. “Let’s wait” 9. “Look at the signal, does it say walk or don’t walk?” 10. “Are the cars stopped?” 11. “Let’s look both ways to make sure it’s safe.” 12. “Does the signal still say walk.” 13. “Look across the street” 14. “Let’s walk” 15. Repeat steps 1–14 to cross back across the street	1. “Let’s practice joining a conversation, first listen to the topic.” 2. “Let’s listen to the conversation.” 3. “Do you know what they are talking about?” 4. “They are talking about ___.” 5. “Make a comment to keep it going.” 6. When an off-topic statement is made prompt, “They are confused, they are talking about ___. Let’s try again." 7. Repeat step 5 for the next opportunity to make a statement or step 6 (if off-topic statements are made) until the learner has attempted 3 statements (whether correct or incorrect) 8. “Okay, it’s time to leave.”	1. “Find Emma” 2. “Listen to her instructions.” 3. Repeat Emma’s instructions immediately after she is done speaking (e.g., “find the one with a long neck and find the animal with stripes”) 4. “Ready to play again?” 5. Repeat steps 1–4 until 5 trials are completed (10 items are identified)

Maintenance

We conducted maintenance probes for every new skill established for each participant. Maintenance probe procedures were identical to baseline procedures but took place at least 7 days after they met the mastery criteria for the skill.

Generalization

We conducted in vivo assessments of skills in the natural environment for all skills except street crossing. We did not assess street crossing in the natural environment given the high risk any errors might pose. Generalization assessments for all other skills were conducted following the mastery of each skill and were procedurally identical to baseline procedures. Generalization assessments took place at least 7 days after they met the mastery criteria for the skill.

Social Validity

We created a social validity scale for parents and clinical staff members who used the VR platform. The social validity scale consisted of nine questions using a Likert scale, with options for evaluating their agreement with various statements about using the system. Scores of 1 indicated strong disagreement, scores of 2 indicated disagreement, scores of 3 demonstrated neutrality, scores of 4 indicated agreement, and scores of 5 indicated strong agreement. We distributed the social validity survey to parents and staff after the completion of the study. The content of the social validity scale can be found in Appendix A.

Results

Cooperation Phase

Table 1 shows the cooperation baseline results for all participants. Column one shows the percentage of equipment acceptance and column two shows VR-session acceptance. When given a choice between three activities for calming and emotional regulation, participants often selected the same activity. The data represent four sessions with 12 total activities across those sessions. For each session, the participant picked three different VR activities to engage with. Each participant engaged in 100% acceptance of both the equipment and VR sessions.

Table 1.

Cooperation Results for All Participants

Participant	Percentage Equipment Acceptance	Percentage VR-session Acceptance
Jayden	100%	100%
Arya	100%	100%
Basma	100%	100%

Intervention for Participant 1 (Jayden)

Figure 1 shows baseline and intervention for all skills targeted for Participant 1, Jayden. During baseline tests for street crossing, Jayden produced 17% correct responses on average (range: 10%–24%). He often looked around at the sky or down the street to watch the cars drive by rather than engaging in the required steps to cross the street. Once in the street, Jayden waved to the cars and stopped to stare at them rather than facing the direction required to continue moving forward. Upon implementation of the treatment package, Jayden followed the instructions to look both ways, look forward in the correct direction to cross the street, and answer questions related to the task. He required additional sessions to identify the street crossing signal (walk versus don’t walk), continue looking straight to cross rather than stopping to say “Hi” to cars in the middle of the street, and consistently look each way and cross slowly enough for the VR system to identify his directional orientation. Upon intervention, his percentage of correct steps completed independently improved from 54 to 100%. We did not test this skill in the natural environment given the safety concerns associated with doing so but his correct responding maintained in the VR environment at the 1-week, 2-week, and 1-month assessments.

Fig. 1.

Baseline, Intervention, Maintenance, and Generalization Phases for Participant 1, Jayden

Jayden displayed an average of 2.5% correct responding (range: 0%–5%) during baseline probes for the skill to join a conversation, only looking at an individual when they spoke on one occasion during the two probes. During the initial phases of the intervention, he followed instructions to look at individuals speaking but did not accurately identify the topic of conversation or make on-topic statements during the initial sessions of the intervention. He had a 55% rate of correct responding during the first intervention session. He demonstrated a gradual increase in accurate responding across sessions, meeting the mastery criteria in the 11th session. The last step established was making an on-topic statement. This skill was then generalized when conversing with staff at the center and maintained at the 1-week, 2-week, 1-month, and 2-month assessments, with scores varying from 95% (one step inaccurate) to 100%.

During baseline probes for listener identifying animals based on two features, Jayden engaged in correct responses 12% of the time, on average (range: 0%–23%). He often missed steps related to listening for both provided features before looking and waiting for the instruction to identify the animals (“go”). Performance during the intervention phase ranged from 62% during the first session to 95% (missing 1 step) during the last few days of intervention as well as the maintenance and generalization assessments. The decision to allow this skill to be mastered at 95% rather than 100% was based on the step being missed (turning away only 1 s before the instructor said “go” on one trial out of the five trials presented by the instructor). The overall skill assessed in this activity was listening to both instructions before beginning the response and identifying the correct animals based on the features provided, both of which were demonstrated to mastery in all sessions in which the learner performed at 95% accuracy. We assessed generalization with toy animals spread out in a playroom.

Intervention for Participant 2 (Arya)

Figure 2 shows baseline and intervention for all skills targeted for Participant 2, Arya. Arya demonstrated an average of 28% correct responding (range: 24%–33%) during baseline assessments. When instructed to cross the street safely, Arya frequently laughed and did not consistently stay oriented toward the road. The automated “ding” provided by the VR system resulted in some motions in the correct direction; however, Arya often appeared more interested in looking at buildings, the sky, or in the direction of sound (e.g., cars beeping, the clicking of the walk sign). Upon intervention implementation, she required additional prompts to orient toward the walk signal, identify the symbols of the signal, and answer questions about the safety of crossing the street. Orienting toward the signal was the last step established before meeting the criterion. Data during intervention ranged from 57% (first intervention session) to 100% (last intervention session) correct responding. Arya mastered the skill in 16 intervention sessions. We conducted one maintenance assessment at the 1-month mark, and Arya demonstrated maintenance of the skill with 97% correct responding. We did not conduct generalization to the natural environment for this skill given the safety concerns associated with doing so.

Fig. 2.

Baseline, Intervention, Maintenance, and Generalization Phases for Participant 2, Arya

Like Jayden, Arya did not stay oriented to the speaker, listen to both instructions before responding, or wait for the instructor to say “go” during the baseline phase of listener identifying animals by two features. Arya displayed an average correct response rate of 9% during the baseline phase (range: 4%–16%). Upon intervention implementation and the delivery of instructions (“Listen to the instructions before finding the animals.”), Arya engaged in correct responding 52% of the time during the first two intervention sessions. In subsequent sessions her accuracy steadily increased until she met the mastery criteria in the 14th session. One generalization (toy animals placed around the room) and one maintenance assessment (in the VR environment) were conducted at the 1-week mark after the intervention. The second maintenance assessment was completed at the 1-month mark. During maintenance and generalization, she reached 96% accuracy at the 1-week mark and 100% at the 1-month mark.

During baseline assessments for joining a conversation, Arya did not always orient toward the individual speaking, and instead looked around the room, laughed, incorrectly identified the topic, or did not make on-topic statements during the conversation. She demonstrated an average correct response rate of 28% during the baseline phase (range: 20%–35%). During the intervention, when prompting and reinforcement were provided, Arya’s correct response rate was between 75 and 90%. She followed the instruction to orient toward others when they spoke, answered the topic questions accurately, and attempted to make related comments, but often required echoic prompts to complete the statement (e.g., she might say “vanilla ice cream” and the instructor prompted “I like vanilla ice cream” to assist her in making a full statement). During sessions in which Arya completed 90% of the steps correctly, she only missed steps in which statements were required. She demonstrated the steps of looking at the correct speaker, waiting her turn to speak, and identifying the topic being discussed. She correctly made an exit statement before ending the session (e.g., “Thank you, goodbye.”). A 1-month assessment was conducted due to the pause in sessions for the 1-month vacation. Of the three skills targeted, this is the only skill she did not maintain (she performed at 36% accuracy).

Intervention for Participant 3 (Basma)

Figure 3 shows baseline and intervention for all skills targeted for Participant 3, Basma. She had a 43% average accuracy rate in her responses (range: 24%–52%) during baseline assessments for listener identifying animals by two features. She often looked before the instructor said “go” and identified the wrong animals when presented with a feature. Once the intervention was introduced, she met the mastery criterion in five sessions. Her percentage of correct responses during the intervention ranged from 88% (first session) to 100% (last session). Generalization was assessed in the playroom of the center with toy animals dispersed around the room. We assessed generalization at the 1-week and 2-week marks and maintenance was assessed at the 2-week and 1-month marks. She maintained her performance across all sessions.

Fig. 3.

Baseline, Intervention, Maintenance, and Generalization Phases for Participant 3, Basma

During the baseline phase for street crossing, Basma often looked around and asked questions about how to perform the task (“Am I supposed to press this?”) and made statements about the difficulty of the task (“This is hard, I need help.”). Her baseline correct response rate was between 15 and 42%. In the first intervention session, she had 57% correct responses, which increased over the course of the intervention sessions to 90% or higher. She met the mastery criteria for this skill in the sixth session. Maintenance assessments were completed at the 1-week and 1-month marks, during which she continued to meet the mastery criteria for the task. Generalization was not assessed for this skill given the safety concerns associated with doing so.

During the baseline phase for joining a conversation, Basma occasionally looked at the correct person speaking but often made comments when the person was still speaking, made off-topic comments, and attempted to end the conversation (“Okay, okay, I’m done talking about this.”). She demonstrated an accuracy rate of 25%–30% in her responses during the baseline phase. Basma met the mastery criteria in only two intervention sessions. The addition of prompts and reinforcement was immediately effective. Maintenance and generalization assessments were completed at the 1-week and 1-month marks. Generalization assessments were conducted with three staff members in the center across two different topics.

Social Validity

Figure 4 shows the average social validity scores completed by four clinical staff members (top graph) and three parents (bottom graph). On average, staff members consistently agreed (score of 4) or strongly agreed (score of 5) with all the positive assessments of using VR with their clients, except for the question “using VR produced negative reactions from my client,” which received an average rank of 4 (agree). Parents also consistently agreed or strongly agreed with all positive statements regarding the use of VR. On average, parents disagreed that VR produced negative reactions and were neutral in their assessment of their ability to use VR with their child without assistance.

Fig. 4.

Social Validity Results for Clinical Team Members and Parents

Indices of Happiness

 Figure 5 shows the percentage of trials with indices of happiness across all participants and phases and the total number of sessions terminated by each participant. Two of the three participants (Jayden and Arya) demonstrated happiness in the form of smiling, laughing, making positive comments about the activity (e.g., “I like the blue car.”), and requesting to play it again during every session and every activity presented. Basma displayed signs of happiness during all sessions for two of the targeted skills (identification of animals by feature and joining a conversation). During street crossing, she showed signs of happiness during 98.7% of overall sessions. During sessions when signs of happiness were not displayed and occasionally during sessions when they were, she stated “This one is hard” or “I’m going to need help” when completing the activity. This demonstration of potential nervousness, frustration, or disinterest occurred either before starting the activity or during the one step she needed additional prompting to establish (identifying when it was okay to cross). When the rule statement was presented, she continued with the activity with no further signs of unhappiness. Basma never attempted to remove the VR equipment.

Fig. 5.

Percentage of Trials with Indices of Happiness Across All Participants and Phases and Total Number of Sessions Terminated by Each Participant

Discussion

We investigated the effects of a VR treatment package on the development, maintenance, and generalization of skills across various domains with three children with ASD. We found that each of the three participants accepted the equipment and VR sessions, and then acquired each of the three targeted skills through the implementation of the intervention. In addition, the acquired skills generalized to the natural environment in all targets in which generalization assessments occurred across all three participants (no generalization assessments were conducted for street crossing). Two of the three participants demonstrated happiness in the form of smiling, laughing, and making positive comments about the activity during 100% of sessions whereas the third expressed some initial signs of unhappiness during the street-crossing lesson. The clinical treatment team utilizing the software consistently agreed or strongly agreed that the children demonstrated the acquired skills in the natural environment, that they would use VR again, and that they see the potential for long-term benefit. They also expressed general enjoyment of using the equipment, both for themselves and their clients. Among staff members, the average ranking for the statement “VR produced negative reactions from my client” was 4 (somewhat agree). We suspect this score was high given Basma’s initial reaction to specific lessons and verbal expression of fear and anxiety in some situations and that the higher average scores reflect team members who worked with her directly. Parents also consistently agreed or strongly agreed with all positive statements regarding the use of VR. On average, parents disagreed that VR produced negative reactions and were neutral in their assessment of their ability to use VR with their child without assistance. This neutrality suggests that if VR is utilized as a parent teaching tool, some training and orientation may be required to support parents’ confidence in using this type of technology.

Our results are consistent with previous research showing that a VR treatment package can be used to teach skills such as social interactions (e.g., Clay et al., 2021) and joint attention (e.g., Ravindran et al., 2019). We extend the existing research with five important findings. First, children easily accept both the equipment and the VR session and do not require any additional interventions to engage with this type of technology. Second, children exhibit behavioral indicators of happiness while using the equipment and engaging in the lessons. Third, skills can generalize from the VR world to the natural environment. Fourth, parents and staff reliably rate the use of VR positively, and finally, apart from one skill and one participant, skills acquired in the VR world tend to maintain over time.

This study demonstrates that a treatment package consisting of VR, prompts, and reinforcement can be effective in teaching new skills and that those skills are maintained after mastery and generalize to the natural environment. It should be noted that when a skill was tested in the natural environment, participants demonstrated it, suggesting that VR is not only effective in establishing new skills but also can lead to the practical application of those skills in real-life situations. Establishing new skills in the VR world is particularly important when safety or difficulties replicating an experience in the environment prevent instructors or parents from teaching them. For example, it can be extremely laborious to teach a child how to respond appropriately to strangers because the teaching trials involve dozens of “actors,” some of whom are strangers (Levesque-Wolfe et al., 2021). This type of arrangement is required when attempting to construct scenarios because the child must experience multiple examples of interacting with both strangers and people who are familiar to them. Although most authors of previous research studies do not report the specific amount of time it takes to teach skills in these mock scenarios without VR technology, we hypothesize that VR may offer a more efficient way to teach these types of skills. We were typically able to target all 3 skills within 1 VR session that lasted approximately 5 min to 20 min. Overall acquisition across all participants took approximately 3 months. In addition to the time efficiencies involved in the use of VR, safety and practical considerations must be made when targeting skills such as flying on an airplane, interacting with police officers, safely exiting a building on fire, and social etiquette in bathrooms or locker rooms. In our study, crossing the street safely was deemed an important target skill for all three participants, but practicing daily and in the real environment could pose risks and require resources that were not typically available (e.g., extra staff, traveling to different locations). By targeting these skills in the VR world, we were able to reduce the risk of practicing in the natural environment and easily conduct the multiple teaching trials required to attain mastery.

Our study has many implications for teaching children with ASD. VR can be used as a supplement to ABA, speech therapy, or other related teaching sessions to increase efficiency and improve the overall learning experience. Although our primary treatment package relied on behavioral principles, we used VR to simulate interactions and conditions that were difficult to construct in the real world. We found that children accepted the equipment easily, meaning that VR could be used in an intervention with no additional training required to help children engage with the equipment. This study is the first to examine social validity and indices of happiness when using VR. Although previous research has demonstrated the effectiveness of VR, our study has shown that VR is not only effective but also that instructors and parents find it enjoyable. In addition, most participants always engaged in behavioral indicators of happiness while engaged with the platform.

The current study has some limitations. First, we did not compare the rate at which participants acquired skills with a VR treatment package to the rate at which they acquired skills using basic behavior analytic teaching procedures. Future researchers may wish to choose two tasks of equal difficulty to target simultaneously, one with VR and one with basic teaching procedures, to determine which strategy is fastest. Next, we did not collect baseline data on the skill in the natural environment prior to teaching. This lack of baseline data was primarily due to the complexity and/or safety concerns related to artificially creating the situation, but future researchers may wish to gather baseline data in the natural environment before teaching and generalization tests. We also did not complete a component analysis to determine which components of the intervention package were responsible for change. Future researchers may wish to do so and may discover a simpler package that delivers the same results. Finally, we did not test for generalization of the street crossing skill due to concerns related to safety. Future researchers may wish to establish creative ways to test for generalization of these types of skills. For example, researchers could stage a safe situation to test street crossing skills by utilizing a quiet neighborhood with blocked outside access and with staff positioned to drive safely.

Our findings suggest various avenues for future research that can augment our understanding of the application of VR within behavior analysis. First, future researchers may wish to investigate the long-term retention of skills gained through VR. Our results show short-term maintenance, but given the skills targeted (e.g., crossing the street) would need to be utilized for the participant’s lifespan, future research should investigate longer periods of time postacquisition to determine maintenance and generalization in different contexts as well as whether periodic refresher sessions in VR are beneficial. Second, future researchers should explore the generalization of skills learned in the virtual world to real-world scenarios more extensively. Our study shows that certain skills can be transferred, but systematic generalization protocols will be necessary to determine the conditions under which transfer of skills occurs most effectively. Third, future studies should consider individual differences in responsiveness to VR training. Research focused on identifying the characteristics of those who respond best to VR could lead to more targeted and individualized interventions. The incorporation of physiological measures, such as eye-tracking and heart rate variability, in conjunction with objective behavior measurement can provide more detailed information about how children interact with and respond to VR environments. Such data could inform the tailoring of VR scenarios to match individual levels of engagement which could optimize skill acquisition. Finally, the cost-effectiveness of VR interventions compared with traditional behavior analytic strategies should be explored. Understanding the economic implications of VR training can support its integration as a regular component of intervention packages for children with ASD.

The investigation underscores the importance of bridging technology and therapeutic interventions to foster skill development in a population that often encounters difficulties in conventional learning environments. VR appears to be an effective tool for teaching children with ASD new skills, offering a promising avenue for future research and intervention development in ASD. We recommend that a VR treatment package be used as a supplement to instructional sessions, particularly for skills that require many resources or pose risks when taught in the natural environment.

Appendix A

Social validity scale for parents and clinical team members.

Thank you for participating in our pilot of virtual reality to teach skill acquisition. Your child participated in learning how to cross the street safely, how to make an on-topic statement in a conversation, and how to identify animals based on their features. Now that your child has completed these skills using virtual reality and has demonstrated the ability to use these skills in the center or home environment, we would like to assess your opinion about the virtual reality procedures and results. Please complete the following survey. Participation is voluntary and you may decline to complete this survey or discontinue the completion of the survey at any time. *The word, “child” was changed to “client” for use by clinical team members.

I liked using virtual reality (answer if applicable).

1 Strongly disagree

2 Somewhat disagree

3 Neither agree nor disagree

4 Somewhat agree

5 Strongly agree

I could easily use virtual reality with my child without assistance (answer if applicable).

1 Strongly disagree

2 Somewhat disagree

3 Neither agree nor disagree

4 Somewhat agree

5 Strongly agree

I am satisfied with my child’s response to using virtual reality.

1 Strongly disagree

2 Somewhat disagree

3 Neither agree nor disagree

4 Somewhat agree

5 Strongly agree

Virtual reality was an appropriate way to teach my child skills.

1 Strongly disagree

2 Somewhat disagree

3 Neither agree nor disagree

4 Somewhat agree

5 Strongly agree

I think my child performs better on the skills learned using virtual reality.

1 Strongly disagree

2 Somewhat disagree

3 Neither agree nor disagree

4 Somewhat agree

5 Strongly agree

My child will benefit in the long run from the procedures implemented in this study.

1 Strongly disagree

2 Somewhat disagree

3 Neither agree nor disagree

4 Somewhat agree

5 Strongly agree

Using virtual reality produced negative reactions from my child (e.g., crying, tantrum, discomfort, etc.)

1 Strongly disagree

2 Somewhat disagree

3 Neither agree nor disagree

4 Somewhat agree

5 Strongly agree

I would use virtual reality to teach additional skills to my child.

1 Strongly disagree

2 Somewhat disagree

3 Neither agree nor disagree

4 Somewhat agree

5 Strongly agree

The skills learned during virtual reality were demonstrated in the natural environment (home, school, community).

1 Strongly disagree

2 Somewhat disagree

3 Neither agree nor disagree

4 Somewhat agree

5 Strongly agree

Appendix B

Procedural Fidelity

Cooperation Procedural Fidelity Steps:

1. The instructor presented the SD as written (a variation of “We are going to try something new and use this headset to practice learning some new skills”).

2. The instructor asked if the participant was ready to participate.

3. The instructor informed the participants the device would be placed on their head prior to placing it on them.

4. The instructor adjusted the straps to fit around the participant’s head.

5. The instructor presented a video from the “emotional regulation” series and the video was already on when the device was placed on their head.

6. After each trial (approximately every 3 min), the instructor asked if the participant was ready to continue.

7. If the participant verbally protested (“No thank you”) or physically prevented the placement of the headset on them, the instructor said, “That’s okay, we don’t have to,” ended the trial and moved on to the next trial.

Skill Acquisition Fidelity Steps:

1. The instructor presented the SD as written (a variation of “We are going to try something new and use this headset to practice learning some new skills”).

2. The instructor asked if the participant was ready to participate.

3. The instructor informed the participant the device would be placed on their head prior to placing it on them.

4. The instructor adjusted the straps to fit around the participant’s head.

5. The instructor presented the device to the participant with the video already on.

6. The instructor asked if the participant was ready to continue.

7. If the participant verbally protested (“No thank you”) or physically prevented the placement of the headset on them, the instructor said, “That’s okay, we don’t have to,” ended the trial and moved on to the next trial.

8. Prompts and instructions were not provided during baseline.

9. Every few minutes, the instructor asked if the participant would like to continue (e.g., “Do you want to keep going?”).

*in addition to the 9 steps listed above, the instructor was also evaluated on accurately presenting the prompts from the Floreo system (see Appendix Table 2)

Funding

No funding was associated with the current study. Our organization purchased subscriptions and licenses for the use of the software as any customer of the organization would. We purchased subscriptions and licenses for participants associated with this study and for participants not associated with the study.

Data Availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethical Approval

All procedures were performed in accordance with the ethical standards of the institutional review committee and with the 1964 Helsinki declaration and its later amendments. This study was formally approved by the organizations’ Research Review Committee prior to its’ start.

Conflicts of Interest

The authors of this article declare no conflicts of interest regarding this article.