About

Summary

The New Jersey Families Study (NJFS) is a video ethnographic examination of how families support their children's early learning. We accumulated 11,500 hours of video with audio footage from 21 families with young children in Mercer County, NJ that agreed to a two-week naturalistic observation of their daily lives, behaviors, and activities while interacting with their children at home. Each family had a two- or three-year-old child. Otherwise, the final sample is highly diverse in terms of race and ethnicity, social class, family structure, and place of residence. As many as eight high-definition video cameras with microphones were located strategically in up to four rooms in participants' homes—rooms where most parent-child interactions occur. Cameras with motion sensors were activated simultaneously and continuously throughout the day and evening for two weeks. Survey and interview data collected during six additional points of contact with families supplement the video data. Our goal now is to open these multimodal data to a worldwide research community, while honoring promises of confidentiality made to participating families. By making NJFS data more widely available, we are able to expand the range and depth of new scholarship on the nature, frequency, and quality of parent-child interactions. In so doing, we aim to make two contributions: first, to the understanding of early childhood development and education and, second, to the emerging science around big data.

 

Contributions

Our overarching goal is to design, build, and sustain an early childhood database for the global research community using NJFS data. Meeting this goal will advance two important objectives. First, it will contribute to an improved understanding of early childhood development and education. Our study breaks new ground. It is the first time anyone has attempted an in-home naturalistic observation of this breadth, intensity, or duration. The unprecedented texture and granularity of the video data can lead to new ways of understanding. We illuminate how families support their children's early learning by shining a bright light on parent-child interactions, especially those that prior literature has suggested are influential in the formation of children's cognitive and social-emotional skills—the amount of talking and reading parents do with children; children's sleep patterns and practices; their diets and nutrition; their exposure to electronic screen time; structures and routines at home; and the way that stress outside the home affects parenting practices. Creating a community resource to engage a larger research audience will open up the data along each of these dimensions to more diverse perspectives and interpretations.

Second, we aim to contribute to the emerging science around big data. Our unique project joins forward-looking data with a host of cutting-edge data science techniques. The latest advances and important future discoveries in computer vision, speech recognition, artificial intelligence, and machine learning will allow us to handle a staggering amount of data, to create a user-friendly database, and to roll it out in a timely way. Our study could serve as a model for future research projects that aim to open up huge amounts of digital data (video, audio, etc.) to a larger research community. It could lay the groundwork for the development of totally new types of secure data-sharing platforms. Funding would enable these rich data to be housed in a robust environment that would facilitate future research using methods that have yet to be developed.

 

Where We Are Today

A more accurate measure of the size of our NJFS video archive shows that we have twice as much data as previously estimated—11,500 hours of in-home video recordings instead of 5,700 hours. It would take almost 16 months of non-stop viewing to see all this footage. With support from the Burke Foundation, we are building a searchable database to house 504,000 individual video clips. Metadata derived from room views, date and time stamps, and household demographic and socioeconomic characteristics are being associated with each video clip. Filters based on these metadata will facilitate easy access to the videos. Preliminary work has also begun in three related areas: (1) identifying participants in the clips; (2) using image detection to identify and blur instances of child nudity; and (3) evaluating automatic speech recognition software using a sample of NJFS audio clips as an initial step toward creating a digital transcript of the audio on all NJFS video clips.

The New Jersey Families Study promises to revolutionize our understanding of the roles of parent-child interactions, family dynamics, and the home environment in helping young children get off to a strong and healthy start in life. We are seeking additional external support not only to accelerate the curation of this unprecedently rich array of innovative data but also to offer it as an early childhood database to the global research community.