Predictive analytics in higher education has tremendous potential to improve student outcomes. Using student data, predictive analytics can pinpoint students who may need extra attention, then swoop in and help them get to where they need to be.
But as the public grows increasingly critical of how companies and the government use private data, students are understandably expressing concerns about predictive analytics.
In February, a panel of students at EduCon 2.9 said they worry that predictive analytics systems will change the way advisors interact with them—and that being labelled with red flags will pigeonhole them as failures.
See the students' concerns
In a policy paper for New America, Manuela Ekowo and Iris Palmer suggest 18 questions to ask yourself to ensure you're using predictive analytics ethically:
1. What's your game plan? Choose an end goal that you want predictive analytics tools to help you achieve. Otherwise, it's easy to lose focus and difficult to know if the initiative has been a success.
2. Who's involved? Identify the key staff members and stakeholders are and get them involved from the very beginning.
3. What could go wrong? Consider the potential risk factors of your analytics initiative and plan strategies to mitigate them.
4. What metrics are you using? Don't just start collecting data left and right. Determine exactly how you'll measure success.
5. Do you have the right tools? You'll need appropriate technology—and talent capable of using it.
6. What about infrastructure? You're more likely to be successful if you provide robust peripheral support, such as financial resources, good data management, and process changes where necessary.
7. Have you communicated your plan to the broader campus community? Students, faculty, staff, and administrators deserve visibility into the details of your project. Build a communication plan and all stakeholders are on board with the initiative before committing.
8. How will you make the switch? Such a broad initiative can quickly dissolve into turmoil without clear transition steps that everyone understands.
9. Is your data comprehensive and transferrable? You can't run predictive analytics on data you don't have—or on data that can't be transferred into the correct system or format.
10. Do you understand what the data's telling you? It requires an intimate knowledge of the data and your institution's context to understand the nuances of how to interpret predictive analytics results. Provide training to individuals using the data or dashboards, such as faculty and institutional researchers, to ensure that they correctly interpret what they see.
11. Are you respecting students' privacy? Concerns about data privacy and security are on the rise in the wake of high-profile headlines about abuses of data.
12. Are you prepared for a security breach? Get your IT department on board to take preventative measures to protect your institution.
13. Are your algorithms bias-proof? Look for algorithms that go beyond demographic indicators of success.
14. Are you testing your system before using it? Test for accuracy. Test for bias. Test for security. Test and test again.
15. How are you choosing third party vendors? If you're enlisting the help of a third-party vendor to develop your tools, choose wisely. Clearly communicate your concern for ethics at the beginning of your partnership.
16. Do students have access to their own data? If they do, be sure to give them the training necessary to use it to their advantage. Maybe your students want to develop their own learning tools or self-improvement plan. Give them the skills to take control over their own outcomes (Ekowo/Palmer, New America, 3/6).
How one school used predictive analytics to retain an additional 400 students
Next in Today's Briefing
Double the major, double the income? One combo pays off best