Facebook privacy: Three lessons on data, apps and taking precautions
A decade ago, taking attendance was a simple paper-and-pen affair.
“You’d mark present, absent or tardy, then send your best student — the one you'd trust not to wander off — to the office, where someone would enter it in on a green screen, and that was it,” said Kurt Madden, Fresno Unified’s chief technology officer.
Those days are long gone at FUSD. The district now takes over 1 million attendance records every week — all clicked in by teachers on an online platform — and merges them with other data on behavior and grades to predict in real-time which students might be having trouble.
Data advocates would like to know even more.
California is one of six states that lacks a way to share this kind of data between all its school districts, universities and colleges, according to a newly released report by EdInsights Center at Sacramento State. Proponents say developing a statewide data system to show student growth over time is key to finding out which students are falling behind as they move from K-12 to higher education and the workforce.
But critics say parents don’t want their children tracked from preschool to adulthood. They characterize early detection systems like Fresno Unified’s as intrusive and initiatives to collect student data at the state level as the work of information-hungry giants like the Bill and Melinda Gates Foundation.
EdInsights Center Executive Director Andrea Venezia said she understands parents’ concerns, especially in the wake of the Cambridge Analytica scandal. Still, having data has led to positive outcomes for students, as well, she says, like more resources being allocated to foster youth who are struggling to keep up with their peers.
“Data itself isn’t inherently good or evil, it’s what’s done with it,” she said.
The California Department of Education keeps a database of K-12 students known as CALPADS that tracks “student demographics, course data, discipline, assessments, staff assignments, and other data for state and federal reporting.”
Members of the public can search some of the aggregate data, like truancy rates and test results at each school, but only “local education agencies” can see individual student data.
But Venezia said CALPADS and databases of higher education students are incomplete and fragmented. New infrastructure makes it feasible to collect, house and analyze more data points statewide while streamlining the way students are identified between K-12 and higher education.
“The primary challenges to creating a new [statewide longitudinal data system] in California are political rather than technical,” the EdInsights study found.
FUSD on the move: 'We can't wait'
Fresno Unified isn’t waiting for the state to make a move, according to director of Data Science and Software Systems David Jansen. The district already has data-sharing agreements with nearly a dozen outside partners and research institutions, including one with Microsoft that’s unique to Fresno, Lindsay, the country of Denmark and Western Australia.
“By the time California establishes a system like this, it’ll be years,” Jansen said. “We’re talking about our students’ lives, success, employability and the vitality of our region. So we can’t wait.”
In addition to directory information, the district collects test scores, information on health, social-emotional learning, housing status and other data points, all of which are shared to some extent.
Outside research groups like CORE Districts look at student performance, campus climate and demographics, comparing Fresno Unified to other districts. Meanwhile, universities and colleges might look at what courses Fresno Unified students took in high school to analyze their academic performance.
Locally, Fresno Unified also has a data-sharing agreement with the Fresno Housing Authority, which has helped the two organizations determine which students don't have access to internet at home, Madden said.
Groups that provide apps or curriculum will collect data in tandem with the district to find out at minimum when students interacted with the product, for how long and what measurable outcomes there were after, according to Madden.
With Microsoft, Fresno Unified implemented a Personalized Learning Initiative that collected data along the way, according to Philip Neufeld, executive director of Information Technology at Fresno Unified.
“From the beginning, we wanted to have a strong research component to this,” Neufeld said. “Tech is just an expensive asset if we’re not using it effectively.”
Like most personalized learning plans, the Microsoft initiative included daily assessments that collected data on whether students had grasped the concept. Those informed what they would learn the next day.
Madden said this trove of data has given the district insight into its 74,000-strong student body. The ability to detect problems early — when students are on the verge of attendance issues rather than when they’ve had a whole semester of them — is one example of what can be done with the data once it’s collected, he said.
"We show those names to principals, and they go, ‘Yes, there’s so-and-so, that makes sense to me,’” Madden said. “But inevitably, there’s the one name that makes them pause and say, ‘Well, why’s that kid on there?’”
The next data points Fresno Unified would like to track are interactions between students and teachers, according to Neufeld.
He says these can illuminate how students are taking in information.
“A student talks to a teacher, then that student talks to another student,” he said. “That can tell us a lot.”
Across town, Clovis Unified does not work with an outside personalized learning or research provider like Microsoft, according to Assistant Superintendent Curriculum, Instruction and Accountability Debbie Parra.
“I get emails every day from someone trying to sell me a personalized learning platform,” Parra said.
Data is collected and analyzed in-house, however, especially personalized learning data. The district also has data-sharing agreements with Clovis Community College and Fresno State in order to track how CUSD grads perform once they get to college — a smaller-scale version of what the EdInsights report proposes.
Parra said she does see the benefit to expanding that system to the state level, especially because it would help standardize the language schools use to describe their students. Being able to monitor students as they move through the system is another plus, Parra said, but what’s done with that information is most important.
“If you can’t use the data to improve something, what’s the point?” Parra said.
Representatives from both Clovis Unified and Fresno Unified preempted questions about student privacy, emphasizing that any data collecting follows the provisions set forth by the Federal Educational Rights and Privacy Act and acknowledging that parents might be perturbed by the term “data collection.”
At Fresno Unified, all data-sharing agreements are written with memoranda of understanding that specify who can and can’t see what kinds of data, and what can be done with it, Madden said. An MOU with Microsoft, for example, specified that the company can’t try to figure out who Fresno Unified’s students are by comparing the information to another database.
Clovis Unified uses an addendum to its agreements that outlines much of the same according to Parra.
For parents who aren’t reassured, there is not an easy way to opt-out completely. Districts are mandated to collect directory information, although parents can request that it not be shared without their express written consent. FERPA protects student data, but makes it available even without consent to certain entities like local authorities, research groups and universities the student may wish to attend.
Parents of FUSD students can also ask that their child’s data is not shared with any outside organizations outlined above. However, it will still be collected, according to Madden, and then kept in-house.
However, parents do have a right to request information on their children from any public agency that collects their records, according to Leonie Haimson, co-chair of the Parent Coalition for Student Privacy. They can also challenge the information, and if those changes can’t be accommodated, add a note that states the records are in dispute.
Haimson said she’s wary of any proposal that would amass individual student data at the state level — especially one backed by the Gates Foundation. The organization is a major donor to a number of education-related organizations, with its grants available to those who “solve common problems using evidence-based interventions and data-driven approaches.”
“It’s been their No. 1 goal to collect student data without much oversight,” Haimson said.
Outside access to data at issue
Haimson said she believes any data analysis should be done instead at the district level using the least amount of data possible. Independent researchers should keep it secure and restrict who has access to it.
“I understand why we want to do more research, but a lot of the existing research is poor. It relies too much on small, statistical correlations,” she said. “Small-scale, randomized experiments yield better information.”
Even sharing a student’s data among teachers has issues, Haimson said, because it can inform a teacher’s opinion of a student. Studies of the so-called Pygmalion effect found that if a teacher has high expectations of a student, that student is likely to perform better. The opposite was also found to be true.
(At Fresno Unified, teachers can typically see behavior records going back a couple of years. A high school teacher wouldn’t be able to see elementary school records, according to Madden.)
Haimson’s view on data seems to be winning out, as California has proven resistant to establishing a statewide system or expanding the use of CALPADS in the past. Two separate bills that would have started the process of establishing a statewide bank stalled in the state Senate and Assembly this year.
A 2016 court case also determined that outside groups can run queries against CALPADS data, but can't access it outright.
Haimson added that even districts may not be aware of the extent their students’ data is being harvested: watching a video on YouTube as part of a homework assignment involves data sharing, too.
“There’s a universe of apps being used in a classroom,” she said. “Everything might be relevant at some point.”