Five Ways to Remember Students and Their Privacy in Predictive Analytics
Today, data is the currency of society, it’s everywhere — and highly desired. Corporations, politicians, and even higher education institutions collect every data point imaginable in order to make predictions regarding future behaviors and outcomes.
The fact is that every action we take in society leaves a data footprint. Google tracks how long we spend in stores and reports it back without us ever knowing. Our credit and debit cards create profiles of spending habits. And, as you all have likely learned, Amazon works with other sites to entice you to come back and complete the purchase of a recent search. It’s almost ironic that in an era where big government is viewed by many as intrusive, we seemingly hand over countless pieces of personal information to non-governmental actors.
But in what ways is higher education making use of this data revolution? Colleges and universities across the globe are working to answer questions ranging from: which students will graduate on time? Which students might face retention risks? Who do we need to allocate more financial aid to if we want them to matriculate? What adaptive learning courseware could be designed to successfully personalize learning? And historical data exists to help us tackle all of these questions and more.
When we turn our decision-making into the spreadsheet and model-based process used in predictive analytics, what does it mean for our students — those individuals who represent the various lines on our spreadsheets — though? At EduCon 2.9, a panel of students said they “worry that the data will be used to label them before they have a chance to make their own impressions on a teacher.” Students aren’t inclined to trust the data partially because they aren’t able to view it. Students also note that the algorithms don’t have the power to take external context into consideration when evaluating specific students’ records. They don’t want data being used to write them off as failures or to limit services offered; likewise, they don’t want their academic choices to be restricted or to be viewed as a cog in an algorithm.
Institutions that are using predictives in a way that limit student opportunities or choices likely are failing to harness their true opportunities and power. Integrating big data and predictive analytics does not mean losing the personal touch commonly associated with higher education; instead, it means being able to have more meaningful conversations sooner. It means offering students opportunities that we might have looked past or written off previously because additional personalized solutions can be offered based on our own institution’s data. And, perhaps most importantly, it means helping to best position our students to succeed, our programs to be effective, and our institution to progress. In the best-designed scenarios, it means considering our analytical abilities and insights with what our experience and gut tell us.
So, how do we bring out the best of predictives and big data without losing the personal touch? And how do we use data for the good of helping students succeed?
1. Become Educated About Data Privacy
We will always collect and know more tomorrow than we do today. And because of that, we have an obligation — ethically, and in some instances legally — to respect student privacy on these issues. External vendors should always remember that any identifiable student data should remain the property of the institution, or even better the student. As governments become more interested in helping citizens protect their right to privacy this will only become increasingly important. While the Family Educational Rights and Privacy Act has always dictated data privacy concerns on American campuses, it pales in comparison to the expectations exposed in the European Union’s General Data Protection Regulation. But even with increased scrutiny, we must remember that analytics have become too entrenched in higher education to simply stop relying on it. Instead, we must become mindful of protecting our students’ privacy.
2. Use Data to Improve Holistically
A strong analytics culture aimed at ensuring students are remembered as more than parts of an algorithm is built around a desire to help use data to improve our lives. Within higher education, this means using data to help students succeed — and this does not necessarily mean guaranteeing they are going to retain or graduate, it could also mean helping improve a student’s grade, helping guide a student into a major or degree program, matching a student to an extracurricular activity or even assisting a student with an appropriate campus-affiliated living environment.
3. Use the Availability of Analytics to Ask Great Questions
So, if all of this data exists, what can we do to make sure it’s used in ways to best serve the best interests of our students? It starts by asking the right questions. When the possibilities seem limitless, it can be difficult to figure out exactly what we want to know — let alone how we actually find our answers. But how do we learn how to ask the right questions? It starts with being willing to ask questions about the biggest problems we want to solve. Then it becomes a matter of adding necessary specificity as you think through what you are trying to answer. Asking the right questions is a skill built upon over time.
4. Predictives Should Complement Human Intuition
Predictive analytics and big data can do a lot to help those of us in higher education. But we need to be realistic about its limitations. The data we have access to today isn’t perfect. With targeted messaging and experiences designed specifically for us, we are less likely to test our own boundaries and experience things outside of our comfort zone without making a conscious effort. If we over-personalize learning, we may not be helping prepare students for life after college. Employers, after all, might not be as accommodating or understanding. Beyond this concern, we need to be continually cognizant of the fact that models are just models. They aren’t designed to replace our own intuition or opinions; they exist to make processes and decisions more scalable.
5. Include Students in the Conversation
Today’s students are used to having their data collected, tracked, and used to inform aspects of their life. But it doesn’t mean they like it. During my own on-campus experience working with predictive analytics, I made it a point at every freshman orientation to have time with the incoming class to explain what I collected, what I tracked, what models I ran, and what we did with the results. I invited students to come and see their own results and discuss what they mean and how they compare to other aspects of the campus. And while not every student left feeling satisfied, the promise of transparency and well-rooted intentions helped assure buy-in and cooperation.
As much as I wish it wasn’t true, the models in predictive analytics can be wrong. They’re probabilistic for a reason. And, because of this, it’s essential we remember that an algorithm itself should not be used to determine the outcome for a student. The student believes your campus can help them achieve their goals. So, while we are busy building algorithms and thinking about the latest and greatest methodological developments that can assist us in creating the even better model, let’s not forget the impact those models may have on the student.
