From Pasco to St. Paul, how to target kids for success, not crime | Column
Big data can be used for good, not bad. Here’s how, writes a Harvard specialist.

A recent Tampa Bay Times investigation reported that the Pasco Sheriff’s Office “keeps a secret list of kids it thinks could ‘fall into a life of crime.’” Triangulating data from multiple public agencies, like histories of abuse from child welfare and academic failure from the school system, the office generated a prediction of more than 400 students who may become criminals. One can only imagine how a list like this could be used.

Miriam Greenberg
Miriam Greenberg [ Provided ]

This is an egregious violation of ethics for which all public departments engaged in data sharing must be held to account. While it is possible that agency leaders were unaware of the data sharing arrangement, ignorance does not absolve leaders from culpability. Instead, this should be a call to all school system leaders to understand how their data is being shared and put guardrails in place to protect student information.

The rapid pace of data innovation in education has not been matched by a parallel growth in the understanding of the ethical implications of our work. Data analytics has been long conceived to be value-neutral, a technocrat’s endeavor. However, working with education data is ultimately a human endeavor, beset with choices that will have real implications for students, educators, and families. There is an increasing need to interrogate our ethical responsibilities regarding how education data is accessed, used, shared and managed without abandoning the use of data altogether.

Arrangements like the one in Pasco may be more prevalent than we think and will go unnoticed unless there are strong community advocates. For example, in 2018, St. Paul Public Schools, the city of St. Paul and Ramsey County, entered into a unanimously approved agreement to share child welfare, education and juvenile justice system data in order to use predictive models to flag at-risk children. Members of the community caught wind of the agreement and ultimately convinced public officials to dissolve the agreement and involve families in future decisions about how their students’ data would be used.

While there is a great deal of fear around big data and the use of predictive analytics, predictive algorithms aren’t the problem. One can use predictive analytics to turn predictions on their head and help students succeed. For example, in 2007, Chicago Public Schools responded to research from the Chicago Consortium of School Research demonstrating that students who end their ninth-grade year “on track” are nearly four times more likely to graduate from high school than those “off-track.” The district promoted the use of data to monitor students’ risk rates in their ninth-grade year so that teachers could intervene before students fell too far behind. As a result, Chicago schools radically improved graduation rates, which, in 2020, are at a record high. Using predictive models to figure out who is at risk and equitably allocating resources to support them is a good use of data.

The problem isn’t data or prediction or algorithmic bias; algorithms are designed by humans and can be misused by humans. We must make better decisions for how we collect, store, share, and analyze data, guided by the call to do no harm.

Related: Pasco’s sheriff uses grades and abuse histories to label schoolchildren potential criminals.

At the Strategic Data Project, an initiative of the Center for Education Policy Research at Harvard University, we believe in the transformative power of data to change education. We’ve spent the last decade finding and training emerging data fellows to use education data to improve student outcomes. Working with hundreds of data strategists and school systems across the country, we’ve learned that challenges like these are prevalent and require us to refocus our efforts.

Public sector leaders and the data strategists who work for them must protect the privacy of students, engage with families about how data will be used before creating agreements between agencies, and most importantly, ensure that data is only being used to help students succeed rather than usher them into the criminal justice system.

From Pasco to St. Paul, these incidents are about more than data security. They are about the children behind the numbers whom we have a duty to protect.

Miriam Greenberg is director of the Strategic Data Project, an initiative of the Center for Education Policy Research at Harvard University.