Recent reporting by Education Week’s Benjamin Herold gives a glimpse of the direction that school datafication is travelling … and justification for asking serious questions along the way.
A report in Education Week by the EdTech journalist Benjamin Herold details the recent work of school officials in Florida in response to the 2018 mass shooting in Parkland. In particular, this has involved plans for the creation of a mass database designed to collate data on students that might anticipate potential perpetrators of school shootings.
The database (described by Herold as ‘sprawling’) is designed to combine records on millions of students with likely indicators. These included records of being bullied, behavioural and disciplinary incidents at school, diagnosis and treatment records for substance abusers, having been found guilty of committing a crime, or simply being mentioned in unverified reports of suspicious activity made to law enforcement.
Some of these datasets cover millions of students. For example, records of having received an involuntary psychiatric examination implicate over 2.5 million individuals. Similarly, records of foster care placements encompass over 9 million individuals. Alongside these official records, it is mooted that some form of monitoring of social media postings will also be cross-referenced with this data – with plans to contract this out to specialist surveillance companies such as Abacode and Social Sentinel.
In short, this is a vast and multi-faceted database. However, while these plans are now well-advanced, the project has come to the news media attention due to an emerging back-lash. In particular, Herold reports in June 2019 that further work has been delayed due to legal wrangling over the extent of information that can be legally shared, alongside the inevitable bureaucratic complications that surround any large municipal project.
This case highlights growing interest in collating various sources of student-facing data, alongside the administrative appeal of drawing new inferences from mass datasets. Many Florida politicians and parents understandably see the state’s plans as a valid use of student data in the name of ‘school safety’. This is an emotive area, with few effective responses in a country that is seemingly unwilling to introduce effective gun control. In such circumstances, increased digital surveillance offers a compelling alternative for policymakers and school officials keen to be seen to be ‘doing something’.
Indeed, the much lauded state legislation passed last year in the wake of the Parkland school shooting mandated that state Education and Law Enforcement departments worked together to develop a “centralized integrated data repository” that provides access to “timely, complete, and accurate information” from various state agencies and social media. As Herold reports, this subsequently prompted officials to frantically sift through existing data sources that they considered might have some relevance to profiling potential school shooters – hence the ‘sprawling’ nature of the resulting data repository proposals.
So, regardless of whether these plans come to fruition or not, the Florida case gives us a taste of the potentially huge scope of the re-appropriation, re-circulation and re-combination of school data. It also points to the need for caution before generating any single data point on a student or teacher that is personally identifiable, and therefore able to be connected to other personally identifiable records.
For example, there may well have been very valid reasons for school agencies to initially begin to collect data on students who have been the victim of harassment on the basis of their race and/or sexuality. The original architects of this data would not have foreseen its later inclusion in the Florida school shooter database – yet this is precisely what happened.
The Florida database is an extreme example, but highlights to need for vigilance over the ‘mission creep’ of school data. It also highlights the need for democratic accountability of how school data is initially generated, and then later shared and reappropriated. This is not to dismiss the benefits that can undeniably result from the use of massive combined ‘big data’ sets in education, but draws our attention a number of questions and concerns. For example:
- What is the conceptual underpinning of the profiling process? Why are these particular datasets chosen as valid indicators of possible involvement in future school shootings? To what extent is this profiling driven by the existence of available data? What alternative forms of information might be more suitable than relying on existing data?
- What are the consequences of an individual student being identified by this database as a likely ‘threat’? To what extent will school leaders be informed of individual ‘risks’ within their schools, and how are they expected to respond? Do students (or their parents) have a right to know about their profile on this database and what it says about them? If so, is there are right to respond/challenge the calculations being made? In what cases do individuals have a right to opt-out of this profiling?
- How might the Florida school safety database be reappropriated in the future? What other forms of profiling might it be used for (beyond its current remit of identifying potential school shootings)? What other databases might it be connected to and combined with?