I have always been interested in the earliest manifestations of autism in infancy and childhood, especially in the areas of social interaction and communication. In my early work, I compared social behaviors in infants between 6 and 12 months of age who went on to have a diagnosis of autism compared to those whose outcomes reflected typical development. I was intrigued to have discovered the following: at 6 months of age, infants who went on to have autism did not differ from comparison infants in the extent to which they directed looks, smiles, and vocalization toward their caregiver during a face-to-face interaction; yet, by 12 months of age, these same infants were now far less likely to use gaze and gesture (pointing, reaching, showing) to regulate social interactions with another person, and exhibited delays in their development of spoken language (Rozga et al., 2011). Subsequent research revealed that other socio-emotional and communicative behaviors were also affected by 12 months of age, including a decreased tendency to orient and display a change in affect in response to another person’s distress (Hutman et al., 2010), and lower rates of gaze to faces, shared smiles, and directed vocalizations (Ozonoff et al., 2010).
In this early work, I used the usual methods of detailed behavior coding from video — a small army of undergraduate research assistants who carefully scrutinized the videos to annotate them for each behavior of interest, including frame-by-frame changes in gaze direction and affect. These procedures, while highly reliable, were extremely time consuming and remain difficult to do outside a research laboratory setting. Yet, a growing body of research suggests that, given the right behavior targets and screening procedures, we can identify infants and toddlers who show early signs of autism. So how do we scale up our screening efforts? And how do we support clinicians and researchers in their efforts to track and measure early social and communicative development, both to identify children who exhibit early signs of autism and to evaluate treatment progress?
The focus of my most recent research has been thinking about how technology and advances in the field of computational modeling can help support more rapid, large scale, and accurate measurement of behaviors relevant to understanding autism. In my position as a research scientist in the School of Interactive Computing at Georgia Tech, I am collaborating with a number of computer scientists who are interested in developing novel computational methods for measuring and analyzing the behavior of children and adults during face-to-face social interactions. With support from NSF’s Expeditions in Computing, we have started a large multi-site study that aims to develop new computational tools to automatically measure and quantify social and communicative behaviors exhibited by typically developing children, as well as children with diagnoses on the autism spectrum, based on video and audio data collected during various social-interaction protocols. Please visit the Computational Behavioral Science and Child Study Lab websites to learn more about this work.
I also have several related ongoing collaborations with my colleagues in Computer Science and psychologists at the Marcus Autism Center in Atlanta. These include using wrist- and ankle-worn accelerometers to help automatically detect and measure self-injurious, aggressive, and repetitive behaviors in individuals with severe behavior problems, and developing new video-capture technology to help parents collect meaningful video examples of their child’s problem behaviors and developmental progress in the home.
If you would like to hear more about my work, please feel free to contact my by email.