The University is collecting data on students to predict student success, but what to do with it is an ethical question that remains unanswered. 

While an increasing amount of universities collect student data, Colorado State University is unique in that recently it began a campus conversation about how to effectively and ethically implement the use of educational data science, known as learning analytics.

Learning analytics involve the collection, measurement analysis and reporting of data about learners, their behaviors and their contexts. Currently, the University collects information on all students including ACT and SAT scores, GPA, and their use of Canvas, the University’s course management website. The data, after being sent to a company with an algorithm to process it, predicts student success. The data is sent back for use by admissions counselors and academic advisors. 

Data companies use different variables and weight assignments to predict student success through the use of black box algorithms. Black box algorithms use inputted data and give a predicted output, but the ways the algorithm weighs and decides the output are unknown.

There are two major learning analytics projects being pursued by the University, said Bayler Shubert, the Associated Students of Colorado State University’s Director of Academics, who has worked with campus partners as the student representative on the topic. The first is a partnership with Education Advisory Board, the program currently used by advisors that collects data on students’ demographics and high school performance. The second is the learning analytics program LoudCloud, which would predict student success for use by faculty. 

LoudCloud is in the pilot-testing phase in order to give faculty an opportunity to see how the program would function, but no real student data is being utilized at this time. 

The Committee on Teaching and Learning, a task force on the Ethics of Learning Analytics, took up the question of how to implement Learning Analytics thoughtfully by means of a set of guiding principles.

“Faculty has been working on this for two years,” Shubert said. “Principles are the first step, and then you move to policy.”

The CoTL task force, chaired by Director of the Center for the Analytics of Learning and Teaching (C-ALT) Dr. James Folkestad, is committed to CSU’s foundational principle of inclusive excellence and supports learning analytics projects that aim to protect diversity and are committed to using analytics for ethical purposes, according to their website. Dr. Matt Hickey, chair of CoTL, charged the task force that developed the principles.

Hickey said predictive analytics when leveraged properly could be helpful, but it is fraught with potential risks. He cited one such challenge as being that educators must identify what population these predictive models are built on.

A particular concern arises when considering the University’s student success initiative, which is fueled by the desire to be a student-ready campus, especially for traditionally underrepresented populations such as first-generation students, students of color and low-income students, Hickey said.

“If the predictive analytics are not built on that diverse of a population, we may be drawing some conclusions about what it means to be successful as a student on the basis of these predictive models that don’t apply to the students we’re most interested in helping,” Hickey said.

Hickey also stressed the importance of these instructional technologies being used to facilitate improved face-to-face teacher-student interactions, rather than replacing them outright.

“We have a responsibility that doesn’t go away just because we grew the campus,” Hickey said. “Leveraging instructional technology to take the place of those face-to-face interactions is an entirely different animal, and one that our committee has pretty significant concerns about.”

Hickey said CSU is ahead of the curve because when they are negotiating University-wide contracts, such as that with EAB, they have been thoughtful as to who owns the student data. However, he identified a risk in the current lack of explicit policy in that individual faculty members could still engage in an IT exchange outside of University-wide contracts.

“That’s part of academic freedom, and we want to make sure that we’re sensitive to that and respect that, but academic freedom entails some academic responsibilities as well,” Hickey said. “My sense is that we have a duty to begin to translate this into some policy decisions.” 

The principles are now being discussed in different areas of campus, including faculty council and ASCSU, in order to recieve feedback from the CSU community. At this time, Hickey’s committee has formally endorsed them, as has ASCSU as part of a resolution passed in consent agenda Feb. 21. Shubert said the next step is to present these principles via a report to the Faculty Council in order to seek their endorsement before moving onto the policy-making process.

According to a report from the CoTL taskforce, the Principles of Learning Analytics are currently under active development and review, and the Committee is actively seeking critical feedback on these principles.

“If these are tools to help us do that better, what’s the problem? My response to that is that there is no problem, as long as we know what better means,” Hickey said. “Better doesn’t mean faster. Better doesn’t mean more throughput of students. Better means better educational opportunity for all the students we present.”

Collegian reporter Natalia Sperry can be reached at news@collegian.com or on Twitter @Natalia_Sperry.