Using the prism of diversity, equity, and inclusion (DEI), higher education data and analytics professionals are reexamining how and what data is collected to uncover implicit bias embedded in current data practices.
Institutional leaders are being challenged to support DEI goals and strategic plans with data analytics that can document progress on specific outcomes. As revealed in a recent EDUCAUSE report, data and analytics can be used to examine how different stakeholder groups are impacted by current institutional structures. The goal is to design new structures, if needed, to eliminate systemic bias.
Best practices for equitable analytics are evolving. It is now standard practice to disaggregate data to uncover any negative impact of institutional policies. A new sensibility about eliminating and reducing institutional bias for underrepresented demographic groups and moving away from deficit models of students’ potential is required. The authors of the report caution that this work must have support from leadership and that culture shifts are slow in higher education.
Although there is growing recognition for human and financial investment in DEI for data analytics, there are two primary challenges:
- Analysts do not have enough DEI content knowledge to apply their skills to DEI goals
- Or analysts have the technical skills to provide cogent analysis but need the right kind of partnerships to bring DEI knowledge to their work.
Algorithmic bias in analytics
The purpose of data analytics is to draw conclusions from observable data. However, the types of questions asked and the stories told are human decisions that are impacted by implicit biases and different world views that are baked into data processes. According to the report’s authors, “Systemic inequity is propagated because most analytics methods rely on data inputs from systems that are largely known to produce inequitable results.”
There is a growing awareness of the biases embedded in current machine learning algorithms that “learn” through the “existing relationships between inputs and outputs.” As DEI initiatives continue to grow in importance, higher education stakeholders will increasingly question the assumptions that are embedded in analytic tools and practices.
DEI for data and analytics in practice
Some examples of new diversity initiatives for higher education include:
- University of Oregon developed a new machine learning model that predicts student persistence more successfully than previous models. The challenge here is that without thoughtful planning and specialized technology, machine learning can increase the systemic biases of current practices. By using a model that prioritizes equity and trust, the university created powerful tools to intervene proactively rather than reactively if students needed help.
- St. Paul College created “equity by design” dashboards to provide data to support new equity practices. The dashboards illustrate the gaps in student success rates between racial and ethnic groups that drill down to course level. This allows faculty to see any equity gaps in their courses while providing insight for institutional leaders to track progress in closing equity gaps.
- After shifting to a virtual remote learning during the pandemic, Foothill College created a unique study focused on new methods to identify student engagement through a DEI lens to quantify equity gaps in the new virtual learning environment.
The importance of DEI initiatives is critical as the delivery model for higher education continues to evolve. Understanding how technology does or does not exacerbate existing inequity is a precursor to designing equitable learning systems and student support services.