5 Minutes with Karen Adolph

Using robots to analyze infant function? As part of this month’s Women in Data Science series, we catch up with Karen Adolph, Professor of Psychology and Neural Science

Karen Adolph is a Professor of Psychology and Neural Science at New York University, an affiliated faculty member at the Center for Data Science, and director of the Databrary project. Her work focuses on perceptual-motor development, particularly in infants. Adolph earned her doctorate at Emory, and taught at Carnegie Mellon University before joining NYU in 1997.

1. What research projects are you most excited about this year?

I’m super excited about our work on infants’ spontaneous locomotor exploration. Before infants are independently mobile, caregivers cart them from place to place. But after infants learn to crawl or walk, they decide when and where to move. We’ve discovered that infants rarely move toward recognizable destinations. Using head-mounted eye trackers, an instrumented floor, and high-tech video, we’ve found that mobile infants rarely get up and go to see something they spied while stationary. Instead, they mostly take steps in place or stop in the middle of the floor beyond arms’ reach of any person, place, or thing. In fact, even with no play partner, infants move just as much in a totally empty room as in a room filled with toys. Infants are “moving to move”—and they move a lot (4000 steps/hour in an empty room!)

Their paths are incredibly variable — in length, direction, speed, and shape — and cover most ground surfaces and available area. Using simulated robot soccer (Robocup), we showed that this type of variability is a feature, not a bug. Robots trained on natural infant paths win more matches in Robocup compared with robots trained on geometric paths. And, robots trained on more variable infant paths win more matches than those trained on less variable paths.

2. Why the particular focus on infants in your research? What makes them fascinating as a research subject?

Infants are fascinating because the changes in their bodies, environments, and behaviors are so dramatic. New behaviors come into being. Existing skills improve and transform. All of this is readily observable and can be recorded directly with video and various types of motion trackers. We can literally watch development in action.

3. You have placed increasing emphasis on data science as your career has advanced, particularly with the establishment of Databrary. What is Databrary? Why did you start it?

Video is a uniquely powerful tool for capturing the richness and complexity of behavior. As such, video is ideal for both data and documentation. Databrary is a web-based data library for securely archiving, sharing, and reusing research videos — including raw research videos and video demonstrations of procedures and displays. Sharing and reusing video will make behavioral and social science more transparent and reproducible, and accelerate progress and magnify public investments in research. Researchers can also find video clips to use for teaching and conference presentations.

Databrary developed a rigorous policy framework to allow ethical sharing of data that contains personally identifiable information. A unique access agreement allows researchers to both contribute and use data. Databrary provides persistent identifiers (DOIs) to shared datasets, so researchers gain citation credit for their work. Databrary and Datavyu, a desktop video-coding tool, give researchers the resources they need to understand behavior and to accelerate the pace of scientific discovery. The Databrary community currently includes 1000+ researchers from nearly 400 institutions around the globe and it grows weekly.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.