3 key strategies for schools, districts considering predictive analytics
Experts say starting small with defined goals and understanding context are a few ways to stage interventions and help students and teachers succeed.
It's not uncommon, in some circles, for the use of predictive data analytics to conjure up visions of a dystopian future where that information is used to selectively identify people who might potentially cause problems and proactively sequester them from society. Those concerns have most prominently been directed toward the idea of predictive policing, especially as law enforcement agencies use genetic data gathered by companies like 23andMe to solve cases, or in social work, where additional context beyond data gathered can mean the difference between a child being removed from a home.
But those fears have also crept into education alongside the use of analytics in academic and behavioral intervention strategies, as well as partnerships like the one between Kansas City's public schools and the Ballmer Group, which aims to improve data sharing between schools' student information systems and case management software in use by local nonprofits. On the latter, Future of Privacy Forum education privacy lawyer Sara Collins told Education Week she would urge schools to refrain from installing the software until they better understand "who the information is being shared with, how much control the school will retain over the data, and what is being done to protect the data."
To some extent, concerns like these have contributed to many of the best practices in place for schools and districts.
"Data use in education is not necessarily any different than data use anywhere else. No one should make any decision off of a single data point. That's not the purpose of data," Data Quality Campaign Executive Vice President Paige Kowalski told Education Dive, noting that multiple pieces of information are necessary to inform a conversation, and with principles like governance, privacy, training and communication taken into consideration.
Start small with specific goals
The biggest mistake a school or district can make when getting into the use of predictive analytics is trying to examine too many things at once. It's best to start small with defined goals and scale up from there, according to Robert Craven, director of educational services for Tustin Unified School District in California.
"I think picking something around achievement and embedding it within professional development goes a long way toward getting a district to kind of the first steps into analytics," Craven said. "That's where you see a lot of power initially and then it becomes a good spot, because people see we’re not out for that 'gotcha moment,' [and] we're not out for Big Brother."
Among the best places to start: using data gathered from test results or other checkpoints to identify students who may be struggling, looking at additional context in each situation, and following them over the longer term to avoid just dropping in after the next exam or checkpoint. This approach, Craven said, requires really working at building a relationship with struggling students throughout the year, becoming aware of who they are and what the data says about both their needs and their strengths.
Context is everything
At St. John's Prep in Danvers, Massachusetts, Assistant Principal for Teaching and Learning Kerry Gallagher says administrators, counselors and learning coaches will often run reports on students with low grades, examining whether students have them across multiple classes or if multiple students have them across one class. But she notes this isn't the only evidence used to determine an intervention.
"If you pull a report and a student who traditionally has done really well suddenly has four Ds and an F, you would want to reach out to the school counselor, the teachers and the parents and ask, 'What's going on with the student?Is there more information about his situation that we need that we can help him recover from this?'" Gallagher said. "I wouldn't say the analytics alone would ever trigger a formal intervention. It would just bring attention to a situation that needs more attention."
One scenario she cited might involve a social crisis, for example, in which a student is cut from football, and football is their life, or perhaps there's a family situation involved. "[Students] make mistakes, and so their data demonstrates their mistakes," she said. "But you have to look at the context around that mistake in order to determine what the child needs in terms of support from the school."
Craven echoed that sentiment for student and teacher interventions alike. The data can allow teacher coaches to have more informed, but informal, conversations with educators within professional learning communities where they're embedded and have built trust within a group. "It’s nonthreatening then," Craven said. "You don't have the principal there. It’s not evaluative. And you can hopefully make some headway."
But there are also still benefits for principals, who can use those data points to pull in an individual teacher and ask, for example, why students in their class are going in the opposite direction compared to peers who might be accelerating according to district data, Craven said. "We don't want it to be the only point, and we don't want it to become something that is utilized all the time to evaluate teachers. But it does help you get some of those conversations."
Likewise, if you're working in a scenario where a student is getting all Bs but has an F in one class, a school counselor "becomes invaluable" in identifying context contributing to the low grade, Craven said, adding that it could be anything from a student having trouble with a specific educator's teaching style to their being placed in a math class they're not yet ready for.
"Especially at huge high schools where you have 3,000 kids and seven counselors, you can’t follow every kid all the time, so that becomes a great way to be able to know who needs some help or at least a conversation," he said.
Don't frame interventions as punitive
When data and contextual information suggest the need for intervention, it's important that educators don't frame it in too punitive of a manner. Doing so can lead students to develop stigmas that they're not good enough to succeed.
"This is no different than when you go to a doctor and they run all those different kinds of tests on you," Kowalski said, noting that data-based interventions aren't framed punitively in other areas of life. "What [doctors] usually come back with for you are a series of short-term interventions they want you to think about, and they give you some sort of long-term idea of where you're headed on your current trajectory. And what they're saying to you is, absent some sort of intervention now, whether it's a medication or a different kind of diet or an exercise, here's what things will look like for you as you age."
Those interventions are never intended to punish you, she said, but to support you in your goals so you can have a longer, happier, healthier life. "That's how we need to think about using information for our students. How do we use information in an appropriate and ethical way with parents and teachers at the table to help ensure that our young people are successful?"
Gallagher echoed that sentiment. "We're not in the business of punishing students," she said. "We actually like the children. That's why we're here."
Follow Roger Riddell on Twitter