Improving Social Communication in Individuals With ASD Using AI

Using Artificial Intelligence (AI) to Improve Social Conversation in Autistic Adolescents and Adults

The purpose of this study is to identify whether researched and commonly used face-to-face interventions can be effectively implemented through artificial intelligence (AI) using an application on the phone or computer. The investigators plan to recruit verbal individuals diagnosed with autism spectrum disorder who demonstrate challenges with socialization. Modules focusing on various difficulties experienced by autistic individuals will provide practice and feedback using voice recognition and feedback. If effective, this intervention can be scaled up to provide cost-effective accessible assistance to individuals, particularly those who do not have access to care or prefer to secure services in the comfort of their own homes.

Study Overview

• Status: Recruiting
• Conditions: Autism Spectrum Disorder
• Intervention / Treatment: Device: AI Modules for Improving Social Conversation

Detailed Description

Social skill intervention has been proven effective to improve interpersonal and communication skills for people with low social awareness or certain neurological conditions such as autism spectrum disorder. However, the application of this type of trainings is limited due to the lack of trained clinicians, the difficulties in finding/creating social settings, and the cost of ongoing treatments, among other factors. In this study the investigators hope to elucidate whether autistic individuals will benefit from the translation of face-to-face interventions delivered by AI application (app) that can be easily accessed on the phone or computer.

For the last decades the world has experienced a geometric increase in the number of individuals diagnosed with autism spectrum disorder, with the CDC reporting that 1 in 36 individuals now being diagnosed with the condition. With such large numbers, individuals in need of intervention services face long clinical wait times along with a shortage of trained providers. From this study the investigators aim to learn whether AI programs that do not require the labor intensive face-to-face clinician hours are effective in improving the social conversation of verbal autistic individuals.

The investigators anticipate that in just 4 weeks of short daily (weekdays) practice sessions with the app, the treatment groups will show improvements in areas that are generally challenging for this population and will report more confidence during social conversation compared to the waitlist control groups. The investigators are requesting that participants complete a minimal number of intervention trials, with the option of completing more if desired. Should participants vary in their improvement based on intervention time, this information should assist with understanding an optimal level of intervention for effectiveness. The investigators will carefully document percentage correct for each module, time spent on the module, and improvement.

The investigators will match adult participants who are within 5 years of age. Adolescents will not be matched but will be divided equally into two groups. The investigators anticipate no change in the wait list control groups.

The purpose of this study is to examine the efficacy of Artificial Intelligence (AI) intervention using an application for adults and adolescents with autism; therefore, human subjects need to be used for this project.

The intervention our Artificial Intelligence (AI) app targets has already been developed and is based on face-to-face research studies. The study will recruit autistic adults and adolescents from the community and will assess the effects of using artificial intelligence to practice social skills.

Our staff will have a single phone/zoom contact with autistic participants (depending on their preference) wishing to enroll in the online training intervention to complete a brief screening before completing informed consent documents electronically through Adobe Sign. Participants must have a primary diagnosis of autism spectrum disorder to participate and must be able to understand and converse in complete sentences.

Parents of potential adolescent participants and potential participants will complete a password-protected 10-20 minute phone/zoom screening (Zoom is safe for PHI) to determine whether their adolescent/they are eligible for the study.

For adolescents, during the last five minutes of the phone/zoom screen, research personnel will invite the adolescent to join in a brief conversation. The researcher's goal is to assess behavior and compliance. Research personnel conducting the screen will ask everyday conversational questions. During the Zoom screening, research personnel will answer any of the potential participant's and their parent's questions. The parent's and potential participant's responses will be recorded in REDCap (database is safe for PHI). Potential participants who are not able to carry on a conversation or who express a lack of interest in the study/intervention will be excluded from the study.

Participants who are not excluded during the phone/zoom screen will be consented. After consenting, in lieu of conducting autism diagnostic assessments, the investigators will request a confirmation of an autism diagnosis or release of information from the medical doctor, psychologist, or other specialist that provided the formal diagnosis of autism spectrum disorder. This can be uploaded securely on REDCap by the participant into RedCap in lieu of testing. Additionally, the investigators will request information on participant characteristics age, age of diagnosis, co-occurring diagnosis, ethnicity, and gender (for the latter two categories there will be an option of "prefer not to answer"), employment status, and number of social events engaged in per week for each participant.

Participants with common co-occurring conditions such as depression and anxiety will not be excluded. This information may help us to understand if any co-occurring conditions may affect the results of the study.

The investigators will match adult participants who are within 5 years of age and then will randomly draw one name for the treatment group and one for the wait list group. Adolescents will not be matched but will be divided equally into either the experimental group or the wait list control group.

Baseline Assessment:

Following informed consent/assent, the investigators will establish eligibility based on review of records confirming the diagnosis of autism and inclusion criteria (i.e., ASD diagnosis, willingness to participate, conversational, not needing intervention based on passing scores during the conversation sample).

Eligible participants will be asked to chat while the investigators audio record A 15 to 20 minute conversation sample (depending on how much talking the potential participant engages in).

Conversation Sample (to be collected pre and post intervention). For both the treatment and wait list control groups. The conversation sample will include a natural interaction with someone who is aware of the experiment and will probe for targeted areas (e.g., for the Empathy module the examiner will say statements such as "I'm feeling a little under the weather today" to evoke an empathetic response). At least 5 opportunities will be provided for the targeted area (e.g., empathy, taking the right amount, asking relevant questions).

The social conversation sample will be transcribed using Sonix.ai and transcriptions will be given a code name (with no identifying participant information). Each response relevant to the study will be scored. Participants who respond at 60% or higher during the conversational probe sample will not be included in the study as they are not in a range that warrants intervention.

Social Validity Scale. Prior to the start of intervention, the investigators will ask each participant who scores below 60% and is eligible for participation to fill out a brief survey relating to their comfort in social situations (see pre- post questionnaire attachment - AI scale).

DESIGN. This is a four week randomized clinical trial. Two groups of 30 (one for the adolescent group (age 11-18) and a second for adults (18-35) will be included. For each module, fifteen will participate in intervention immediately after completing baseline and fifteen will be placed in a wait list control group. Both groups will repeat the baseline measures (social conversation sample and social validity scale) again after 4 weeks.

Each participant in the treatment group will be asked to participate for 4 weeks, at least 10 trials a day, for at least 5 days per week. The investigators anticipate that each module of 10 will take about ten minutes (unless the participant completes additional practice modules). Our AI system will record the number of trials and the length of time each participant engaged in the modules.

Intervention: The modules will be delivered through the investigators' software. These modules have been developed to provide feedback on specific areas following voice or text input (user preference will be recorded in case this effects the results).

The investigators currently have developed AI modules for common social communication challenges exhibited by individuals with ASD. Examples include:

Empathy (general and in the work setting), Asking Relevant Questions, Listen and Respond, Talking the right amount, giving the right amount of personal information, Tactful Responses, Giving compliments, and Staying on topic during Conversation. These are all areas that can be challenging for individuals diagnosed with autism spectrum disorder.

AI modules consist of thousands of total training examples. Participants may use the investigators' software on their phones or computers.

Each module will contain groups of 10 training examples and the participants' verbal responses will be analyzed and feedback provided.

Post Intervention:

Social Conversation Sample. Following four weeks of intervention an identical conversational sample will be collected in exactly the same manner as the baseline. Again, the conversation samples collected via zoom will be audio recorded and computer transcribed, after which audio recordings will be erased.

Social Validity Survey. Following intervention participants in the treatment and wait list control group will be asked to complete the same a survey relating to their comfort in social situations. This includes the questions asked initially along with four additional questions related to the intervention.

For any participants who do not reach the 80% criterion during the four weeks of intervention and who do not meet an 80% criterion during the conversational sample, and who wishes to receive additional intervention, an additional four weeks will be provided and the language sample will be repeated after an additional four weeks. If no improvement occurs after the 8 weeks, they will be considered "non responders."

• Study Type: Interventional
• Enrollment (Estimated): 540
• Phase: Not Applicable

Contacts and Locations

• Study Contact: Name: Lynn K Koegel, PhD; Phone Number: 8056899493; Email: lynnk@stanford.edu
• Study Contact Backup: Name: Monica Lam, PhD; Phone Number: 6502240595; Email: lam@stanford.edu

Study Locations

• United States
  • California
    • Stanford, California, United States, 94305
      • Recruiting
      • Stanford University School of Medicine
      • Contact: Lizzie Ponder, MA; Email: lizzieP@Stanford.edu
      • Contact: Lynn Koegel, PhD; Email: lynnk@Stanford.edu

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Child; Adult
• Accepts Healthy Volunteers: No

Description

Inclusion Criteria:

• Primary diagnosis of autism spectrum disorder (ASD)
• Age between 11 to 35
• Ability to engage in social conversation using full sentences for a 20 minute period
• Verbal communication difficulties in an individual targeted area (e.g., empathy) as measured on the conversation sample

Participants with a co-occuring diagnosis of ADHD, depression, or social anxiety will not be excluded since these are very common in this population and are often related to or a bi-product of difficulties with social conversation.

Exclusion Criteria:

• Non-documented or self-diagnosis of ASD
• At or above 60% correct responding on targeted area of social conversation during conversation probe
• No access to phone or computer for intervention sessions
• Inability to carry on a conversation during the conversation probe
• Apparent intellectual impairment that interferes with the ability to carry on a conversation (e.g., responds only in single words or does not respond to conversation, lack of understanding of questions or content during conversation probe)
• Non English speaking
• Serious medical or psychiatric issues that may interfere with conversation or the ability to complete the program
• Lack of interest in participating.

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Treatment
• Allocation: Randomized
• Interventional Model: Parallel Assignment
• Masking: Single

Number of Arms

2

Arms and Interventions

Participant Group / Arm	Intervention / Treatment
Experimental: Intervention Group; This group will receive the treatment (AI conversation modules) available through the application.	Device: AI Modules for Improving Social Conversation; The application if the modules will be delivered using software. These modules have been developed to provide feedback on specific areas following voice or text input (user preference will be recorded in case this effects the results). The investigators currently have developed AI modules for common social communication challenges exhibited by individuals with ASD. Examples include: Empathy (general and in the work setting), Asking Relevant Questions, Listen and Respond, Talking the right amount, giving the right amount of personal information, Tactful Responses, Giving compliments, and Staying on topic during Conversation. These are all areas that can be challenging for individuals diagnosed with autism spectrum disorder. AI modules consist of thousands of total training examples. Participants may access the modules on their phones or computers.
No Intervention: Control Group; This waitlist control group will not receive treatment until the experimental group completes their intervention.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Social Conversation Sample	Measured in terms of what percentage of responses (0% - 100%) during the investigators' social conversation sample are answered appropriately. Each of the participants' responses will be evaluated by a MA or PhD-level expert. Social conversation samples will be collected via zoom, audio recorded and computer transcribed, after which audio recordings will be erased. An initial sample will be collected at baseline, after four weeks of intervention, and at post-intervention.	Approximately 9 weeks per participant (4 weeks of intervention + pre and post measures + 1-month follow-up)

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Social Validity Scale	Likert scale that ranges from 1 - 5. 1 is the minimum value (the worst outcome) and 5 is the maximum value (the best outcome). Prior to and following intervention, participants in the treatment and wait list control group will be asked to complete a self-evaluated survey relating to their comfort in social situations. Participants in the treatment group will also be asked to fill out a Likert scale relating to their satisfaction with the intervention.	Approximately 5-6 weeks

Collaborators and Investigators

• Sponsor: Stanford University

Study record dates

Study Major Dates

• Study Start (Actual): October 15, 2023
• Primary Completion (Estimated): June 30, 2027
• Study Completion (Estimated): June 30, 2050

Study Registration Dates

• First Submitted: July 26, 2023
• First Submitted That Met QC Criteria: August 3, 2023
• First Posted (Actual): August 14, 2023

Study Record Updates

• Last Update Posted (Actual): May 22, 2024
• Last Update Submitted That Met QC Criteria: May 20, 2024
• Last Verified: May 1, 2024

More Information

Terms related to this study

• Keywords: autism; social skills; autism spectrum disorder; AI; LLMs; GPT; autism support
• Additional Relevant MeSH Terms: Mental Disorders; Neurodevelopmental Disorders; Child Development Disorders, Pervasive; Autism Spectrum Disorder
• Other Study ID Numbers: 70604

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No