Doan Tam, Sullivan Brittany, Koerber Jeana, Hickok Kirsten, Soares Neelkamal
Western Michigan University Homer Stryker M.D. School of Medicine, Kalamazoo, MI USA.
Great Lakes Center for Autism Treatment and Research, Portage, MI USA.
Behav Anal Pract. 2024 May 23;17(4):1147-1159. doi: 10.1007/s40617-024-00936-y. eCollection 2024 Dec.
Collecting data and logging behaviors of clients who have autism spectrum disorder (ASD) during applied behavior analysis (ABA) therapy sessions can be challenging in real time, especially when the behaviors require a rapid response, like self-injury or aggression. Little information is available about the automation of data collection in ABA therapy, such as through machine learning (ML). Our survey of ABA therapists nationally revealed mixed levels of familiarity with ML and generally neutral responses to statements endorsing the benefits of ML. Higher certification levels and more years of experience with ABA were associated with decreased confidence in ML's ability to accurately identify behaviors during ABA sessions whereas previous familiarity with ML was associated with confidence in ML, comfort with using ML, and trust that ML technology can keep client data secure. Understanding the perceptions of ABA therapists can guide future endeavors to incorporate ML for automated behavior logging into ABA practice.Applied behavior analysis (ABA) therapists perceive some value in utilizing machine learning (ML) in data collection during ABA sessions, but the majority of therapists are not familiar with the concept of ML.In our survey, ABA therapists with greater familiarity with ML were more likely to be comfortable using ML in their practice.Surveyed ABA therapists with higher certification levels and more experience with ABA were less likely to be confident in ML's ability to identify behaviors accurately.Awareness of ABA therapists' perspectives about ML, especially regarding privacy and security, and partnership with computer scientists can further the development of ML technology to augment data collection during ABA therapy.Educating ABA therapists about the potential of ML, especially the potential to reduce the burden of behavior logging while simultaneously intervening for aggressive and self-injurious behaviors, will be necessary for successful implementation of ML in ABA therapy settings.
The online version contains supplementary material available at 10.1007/s40617-024-00936-y.
在应用行为分析（ABA）治疗过程中，实时收集患有自闭症谱系障碍（ASD）的客户的数据并记录其行为可能具有挑战性，尤其是当行为需要快速响应时，比如自我伤害或攻击行为。关于ABA治疗中数据收集自动化的信息很少，例如通过机器学习（ML）实现的自动化。我们对全国ABA治疗师的调查显示，他们对ML的熟悉程度参差不齐，对支持ML益处的陈述总体反应中立。更高的认证水平和更多年的ABA经验与对ML在ABA治疗期间准确识别行为能力的信心下降有关，而以前对ML的熟悉程度与对ML的信心、使用ML的舒适度以及对ML技术能够保护客户数据安全的信任有关。了解ABA治疗师的看法可以指导未来将ML用于自动行为记录纳入ABA实践的努力。应用行为分析（ABA）治疗师认为在ABA治疗期间利用机器学习（ML）进行数据收集有一定价值，但大多数治疗师并不熟悉ML的概念。在我们的调查中，对ML更熟悉的ABA治疗师在实践中更有可能愿意使用ML。接受调查的认证水平较高且有更多ABA经验的治疗师对ML准确识别行为的能力不太可能有信心。了解ABA治疗师对ML的看法，特别是关于隐私和安全方面的看法，并与计算机科学家合作，可以推动ML技术的发展，以增强ABA治疗期间的数据收集。向ABA治疗师介绍ML的潜力，特别是减少行为记录负担同时干预攻击和自我伤害行为的潜力，对于在ABA治疗环境中成功实施ML至关重要。
在线版本包含可在10.1007/s40617-024-00936-y获取的补充材料。
