Association for Behavior Analysis International

The occurrence and intensity of target behaviors often reflect the social importance and urgency of intervening on such behaviors. However, accurately capturing behavioral occurrences is time-consuming, prone to human error, and subject to various biases. Additionally, the measurement of behavioral intensity—such as the force exerted during a self-injurious episode—remains largely subjective. Typically, this measurement relies on clinical judgment during the behavior's occurrence or is assessed after the fact by evaluating injuries sustained. These limitations present challenges to developing effective and timely interventions. Inertial measurement units (IMUs), which are small wearable motion capture devices, offer a potential solution. IMUs incorporate three types of sensors: accelerometers, gyroscopes, and magnetometers, all of which can detect and record movement. When these devices are affixed to a specific body segment, they can provide precise and real-time data on human movement. This study aimed to explore the feasibility of utilizing IMUs to objectively quantify both the occurrence and intensity of self-injurious behaviors. The investigation was carried out through three experiments, each designed to test different parameters and configurations of IMU usage. The results offer valuable insights into the potential application of IMUs in clinical settings and suggestions for future research.

Conducting a functional analysis (FA) is considered the gold standard for assessing the function of disruptive behavior and informing function-based treatment plans for individuals with autism and developmental disabilities. However, data collected during FAs are subject to human error. Accelerometers are wearable sensors that capture an individual’s movement and can be used to identify behavioral events. The purpose of this study was to pilot the use of accelerometers to identify the occurrence of self-injurious behavior events during a FA. Three participants with autism, who engaged in self-hitting behaviors, participated in this study. Researchers conducted a FA with the participants while they wore small accelerometer devices. Observational data were collected using (1) live observation (“clinical-grade”), (2) from frame-by-frame video analysis (“research-grade”), and (3) via accelerometers. Researchers established a ground truth data set and calculated interobserver agreement across data sets. Discussion of results and recommendations for practice and future research are included.

Behavioral analysts commonly use visual inspection to analyze alternating treatment graphs for assessment and intervention effects. However, there is high variability and disagreement between independent raters due to the subjectivity of this method. Machine learning models (MLMs) offer a promising solution by providing a more objective, data-driven approach to graph analysis, potentially increasing both the accuracy and consistency of interpretations. This research evaluates a machine learning models' ability to produce high accuracy, low Type I error, and strong power metrics when analyzing these graphs. Models trained on a dataset of simulated graphs were assessed for both performance and generalizability on a set of real-world graphs. The model achieved accuracy comparable to human visual raters while demonstrating the best balance of Type I error and power. These findings highlight the potential of MLMs to serve as effective tools in reducing subjectivity and improving the overall reliability and validity of alternating treatment graph analysis, thus enhancing decision-making processes in behavioral analysis.

Identifying and reliably tracking behaviors associated with emotional states, such as happiness and sadness, can greatly enhance a clinician’s ability to individualize interventions and evaluate the social validity of treatments for children with autism spectrum disorder (ASD). However, emotional states often manifest through subtle, idiosyncratic, and subjective behavioral indicators, making consistent data collection and analysis challenging. These challenges limit the practical application of such assessments in clinical settings. Recent advancements in artificial intelligence, particularly in neural network methodologies, have the potential to address these limitations by improving the accuracy and efficiency of emotion measurement. This pilot study investigates the use of facial emotion recognition technology, driven by advanced neural networks, to monitor indices of happiness and unhappiness in young children with ASD. Video data from 42 autistic children, aged 7 to 48 months, was used to train and test four different models. Among these models, the combination of convolutional and dense neural networks demonstrated the best performance, achieving the highest accuracy and recall in predicting emotional states. Limitations and suggestions for future research are discussed.