Perceptions of Machine Learning among Therapists Practicing Applied Behavior Analysis: A National Survey

Area of Science:

• Behavioral Science
• Artificial Intelligence
• Clinical Psychology

Background:

• Real-time data collection in Applied Behavior Analysis (ABA) therapy for autism spectrum disorder (ASD) is challenging, particularly for rapid behaviors like aggression.
• Limited research exists on automating ABA data collection using machine learning (ML).

Purpose of the Study:

• To investigate ABA therapists' perceptions of using ML for automated data collection during therapy sessions.
• To identify factors influencing therapists' confidence, comfort, and trust in ML technology for ABA practice.

Main Methods:

• A national survey of ABA therapists was conducted.
• Data collected included familiarity with ML, confidence in ML accuracy, comfort with ML use, and trust in data security.

Main Results:

• The majority of ABA therapists are unfamiliar with ML.
• Therapists more familiar with ML reported higher confidence, comfort, and trust in its application.
• Increased ABA certification and experience correlated with lower confidence in ML's ability to accurately identify behaviors.

Conclusions:

• ABA therapists see potential value in ML for data collection but require education on its benefits and security.
• Addressing privacy concerns and fostering collaboration between therapists and computer scientists are crucial for successful ML implementation in ABA.
• Future development should focus on enhancing ML's perceived accuracy and utility to reduce therapist burden and improve client interventions.