Big Data and the Personalization Challenge: The Murky Middle
What if you knew, through legal means, something about someone where you could intervene to deliver advice to help them perform better or stay out of danger? Should you act on that?
I was recently delivering a lecture at a major university in Texas and one of the participants posed a couple of very thought-provoking scenarios (stay with me as these are sort of long, abstract scenarios, but hey, what better way to provoke an interesting conversation):
What if by virtue of monitoring a student’s social media habits (with the student opting to share their social media activities), the university could 1) flag behaviors that predict student classroom performance problems and 2) could deliver recommendations to the student, faculty and advisors for improving the student’s social, classroom and study habits and increase their probability of graduating on time and with higher grades?
What if by monitoring the student’s study habits and extra-curricular activities (e.g., how often they went to the library and/or the lab and for how long, class attendance, attendance at campus events, extracurricular activities, etc.), the university could generate recommendations to the student, faculty and advisors for improving the student’s social, classroom and study habits and increase their probability of graduating on time and with higher grades?
These scenarios raise all sorts of ethical questions, such as:
- Would not this be in the best interests of the students to perform this analysis in order to help the students achieve their best outcomes?
- Would students want this level of analysis as students come to college to earn a degree and achieve a level of classroom performance that will help them get better jobs and earn more money?
- Is one type of monitoring more intrusive and a potential violation of the student’s privacy (even if they have opted in to share their social media activities)?
- Where does one draw the line between what is helping the student to increase the probability of their college and post-college success, and what is being a “creeper”?
One of the workshop participants called this the “murky middle” where noble and forthright efforts to help students to be more successful may cross the personal privacy line.
Primary Research (My Kids as Guinea Pigs)
I’ve got to be honest. I don’t have an answer. And I’m really torn by the dilemma of wanting the best performance for students (and my kids), but doing it in such in a way that doesn’t violate their personal privacies or come across as being a “creepy.”
So I asked my children what they thought of these scenarios. Each of my children has a unique perspective:
- Alec has already graduated from college and has been in the workforce for a couple of years now.
- Max is currently in grad school and will be living these two scenarios over the next few years.
- Amelia is evaluating colleges and has to determine if she is comfortable with these levels of monitoring.
Here are their answers:
Scenario #1: What if by virtue of monitoring a student’s social media habits (with the student opting to share their social media activities), the university could 1) flag behaviors that predict student classroom performance problems and 2) could deliver recommendations to the student, faculty and advisors for improving the student’s social, classroom and study habits and increase their probability of graduating on time and with higher grades?
Amelia (evaluating colleges): If I gave my social media information to my future college, I would expect them to skim over my social media accounts. I think in the beginning it could be a great benefit, but for an entire year I do not think it would be beneficial. I believe this because everyone wants to be successful in school and life, but a constant watch over the accounts may get stressful and not necessary. It can often create frustration and lead to hidden activities that could result in something more detrimental.
Max (still in college): If there were a human component to it, I would feel like the college is investing in me. However, if it were electronic I would feel violated. For example, if my college had access to all of this information and a counselor gave me a call, I would feel like someone cares. But if I got an email or text message on my phone offering me suggestions I would feel violated and “big brothered”.
Alec (out of college): I don’t like the idea of anyone – especially a school or other authority type institution – examining my social media to identify trends in terms of how I “tick.” There is something decidedly unsettling about any entity, for whatever purpose, altruistic or not, analyzing my social relations and providing advice on how to improve myself. Ditch certain friends? Don’t go to certain parties? These types of recommendations would only lead to backlash to the system as a whole.
Scenario #2: What if by monitoring the student’s study habits and extra-curricular activities (e.g., how often they went to the library and/or the lab and for how long, class attendance, attendance at campus events, extracurricular activities, etc.) to generate the recommendations to the student, faculty and advisors for improving the student’s social, classroom and study habits that will increase their probability of graduating on time and with higher grades?
Amelia (evaluating colleges): Similar to the first scenario, this scenario would only be helpful in the beginning. The main purpose to going to college is to receive an education on my own. Constantly being tracked and told where to go would help you in college, but would not set you up for the real world. In the real world, you are not going to have someone tell you to go study more or go to classes more. If you miss a class or don’t study, you must face the consequence. Overall, growing up and making mistakes is a part of life and to have that stripped at such a young age may lead to a hard reality check.
Max (still in college): At no point in time would I be okay with my college knowing my location. I feel that crosses the line of caring. I wouldn’t give my college access to something I wouldn’t give my parents access to. To be honest dad, I wouldn’t feel comfortable if you knew how many classes I missed and whether or not I was at the library. I am sure the government knows, but sometimes with electronics, ignorance is bliss.
Alec (out of college): This option is more feasible in my opinion. I would be much more comfortable with an educational institution analyzing my educational decisions. Analysis of my study habits, correlations in terms of grades and time spent in the library or class attendance would prove invaluable in making a logical case to present to the student to convince them that their grades are affected by these factors.
So none of my kids would be happy with this level of monitoring and analysis, with some specific observations:
- While Amelia (evaluating colleges) felt that some level of monitoring and analysis might be useful at first, it would eventually impede her personal development. Students are going to college to learn to live on their own; monitoring their behaviors and “holding their hands” doesn’t allow them to grow.
- Alec (already graduated) seems comfortable with the school using some of the classroom data to make performance improvement recommendations, but neither Max (still in school) nor Amelia (evaluating schools) feel comfortable with this.
However, if colleges were to do this level of monitoring and analysis, then the advice or recommendations must have a human component. By having a counselor get involved, you give the data a human face.
I think this last point is critical – give the data a human face. As I’ve discussed many times, the real impact of big data (and data science) is helping the humans in the process by providing recommendations to the humans (teachers, physicians, parole officers, technicians, counselors, coaches, therapists) to make them more effective at the point of customer engagement.
Let me give you one more scenario to chew on: What if by virtue of analyzing the student’s social media and classroom data, the college is able to make the following probabilistic determinations:
- There is 65% chance of the student dropping out or flunking out
- There is a 33% chance that the student is having substance abuse problems
- There is a 5% chance that the student is going to attempt suicide
What should the college do?
Personalized recommendations are a real dilemma in the world of big data – how much monitoring and analysis is “okay” before crossing over that fine line. And that’s the heart of the challenge: it isn’t a fine line; it’s the murky middle.