A lesson for those implementing predictive analytics in schools; community voice can’t be ignored and is often your best asset.
A few hours before the Data for Black Lives Conference official began 50 plus education activists from around the country and across generations discussed the ways big data has impacted public education in the black communities. Below is a synthesis of their analysis. Rather than typing a direct transcript of everyone's statements I've written a series of post discussing the primary themes that came about during my facilitation, following my presentation on Big Data for Education Justice (View Deck).
The structure of each post:
- Observations – A synthesis of notes from a discussion on the current structures that exist in education and how big data has impacted education.
- Recommendation – Based on existing interventions brought up during our discussion and frameworks being implemented by activist in the room.
- Implications – Broader ideas of how we see big data’s role in public education should evolve.
Data types we discussed:
Quick definition Azeem translates to GREAT in Arabic
- Administrative Data – information captured at the school level that has to do with the schools ecosystem. This data includes metrics that measure achievement, demographics, and engagement.
- Resource Data – Any information that relates to funding, budget and how money is allocated.
- Click Stream Data – the capturing of mouse movements and keystrokes, used to learn patterns about human behavior.
Observations about Data-Driven Schools & the use of Risk Scores
Primary Data Type | Administrative Data
We found that data-driven schools and school systems are using information riddled with implicit bias to determine student outcomes. The perceptions of school officials is codified into “objective” data; districts are measured almost exclusively on their performance on exams.
Districts are failing to invest adequate resources in gathering insight from the community and the families they serve and instead invest in consultants and advisors that promise improved schools but continue to push out interventions that lack community feedback.
An example of this is the use of risk scores, which have been found to be racial bias and can force students down a path to incarceration and underachievement.
While a risk score within itself is tool the ways it can be implemented are what makes it oppressive. Most risk scores are designed using legacy (old) databases that have bias “baked in”. They’re also designed in a way that lacks input from the people who will be most impacted by their implementation.
Recommendations:
The communities that will most be impacted by data-driven interventions need to be part of the design and implementation process.
What does this look like, you ask?
There is no one way to structure an effective coalition between grass-root leaders and local government. But below is a great example of what happens when those most impacted are left out.
Marika Pfefferkorn, leader of the Coalition to Stop the Cradle to Prison Algorithm, recently won a victory with the dissolution of problematic data-sharing agreement; the “agreement” and corresponding predictive technology had been created to predict youth “at risk” of future delinquency. This “faulty process” and legal agreement lacked any formal community partnerships or any advanced investments in youth development.

The alliance recently won their campaign to dissolve a joint powers agreement that would have brought risk score based predictive analytics with no community input.
Marika is quoted, “We know that predictive technologies cannot be detached from human bias and error. And while data can be a tool for positive change, it is also clear that there are many risks that we need to unpack in relationship to the JPA and Big Data, Predictive Analytics and Algorithms and their potential to amplify racial and ethnic disparities in the education and the juvenile justice systems.”
You can learn more about strategies used by community leaders by exploring the Toolkit on Organizing to Combat the School-to-Prison Pipeline.
Marika was part of our discussions and championed shared values of the activists in the room:
- Data-driven education policy, like all public policy can not be equitable if there aren’t formal partnerships between impacted communities and government.
- It is also important to use explicit language at the forefront when communicating with parents and advocates.
Marika, myself and our homies from the Data For Black Lives conference. I’m to Marika’s right, five seats down. Marika presenting on their campaign and how they engage working class communities on predictive analytics.
Implication for Officials:
Government efforts function better when agencies partner with the community to gain support and offer feedback. Data-driven intervention should be implemented with community stakeholders having input from design to implementation and the work needs to be iterative. Transformative education success is only possible when impacted communities are seen as community experts and are given access to the data they themselves generate.