Big Professor is watching you

At the University of Arizona, school officials know when students are going to drop out before they do.

The public college in Tucson has been quietly collecting data on its first-year students’ ID card swipes around campus for the last few years. The ID cards are given to every enrolled student and can be used at nearly 700 campus locations including vending machines, libraries, labs, residence halls, the student union center, and movie theaters.

They also have an embedded sensor that can be used to track geographic history whenever the card is swiped. These data are fed into an analytics system that finds “highly accurate indicators” of potential dropouts, according to a press release last week from the university. “By getting [student’s] digital traces, you can explore their patterns of movement, behavior and interactions, and that tells you a great deal about them,” Sudha Ram, a professor of management systems, and director of the program, said in the release. “It’s really not designed to track their social interactions, but you can, because you have a timestamp and location information,” Ram added.

That is from Amy X. Wang at Quartz.


Colleges getting anxious over dropouts...revenue lost

No because every dropout can be replaced with a new student. And I'd wager the dropouts are disproportionately the ones who receive financial aid and reduced tuition.

Not to belittle your logic, big guy, but logic is not your forte. Unless teacher-to-student ratios are fixed, every student counts.

We aren't disagreeing. Every student counts. Every dropout just gets replaced with a new student. The ratio stays the same. If anything they make a little extra, as they don't refund the fees you leave on the table when you drop out.

The student likely wont be replaced that year. That means a full class is no longer full, a dorm bed is empty, campus services get less "dining dollars"... On top of that, rankings in US News and World Report goes down for student retention and schools have to spend extra dollars on the marginal cost of matriculating an extra student...

U of A is a state school, which in Arizona means they are required to accept all students that meet the schools enrollment criterion. there are not a fixed number of places. So every dropout is lost revenue. Although that doesn't necessarily mean that's their motivation.

If it is an out-of-state student, it has a much bigger impact on total revenue than an in-state student.

The composition of the dropout population may differ substantially from that of the population that graduates. This might make the schools concerned over how well they're serving those who drop out, and perhaps create a motivation to decrease that rate.

The school may or may not be hurt financially by a high dropout rate, but the dropout may leave with substantial (non-dischargeable) student-loan debt with little to show for it.

Agreed. There's more reasons than monetary for wanting the kids you educate to stay in school. Schools aren't just, or even primarily, about money. Most are non-profits in fact.

Schools aren’t just, or even primarily, about money.


Most are non-profits

And can always stay that way by increasing the executive directors' pay or in the case of college the presidents' pay...and deans lots of deans.

Some colleges/universities (I'm guessing I could add "public" but perhaps private as well) have target retention and graduation metrics they need to maintain in order to continue to receive funding at a certain level or to get increased funding. Or perhaps they just want the politicians to leave them alone and improving those numbers is one way to do that. So, in a manner of speaking, yes, potentially some revenue could be lost, and it might be much bigger than you would imagine.

In my 3+ years as a department chair, retention and graduation rates have been one of the main topics of discussion by those in higher administration (Provosts, Vice-Chancellors, etc.)

There is the simple charitable interpretation that schools want to provide happy outcomes to the students that they serve and not sad outcomes, like drop outs.

There are lots of cynical observations about academia. This is not one of the good ones.

Can they predict earlier, and just not invite them?

High school surveillance as transcript?

There are lots of funny ways this could develop, and if we won't, it sounds like China might. Big data allows fine grained and centrally defined meritocracy to be applied at scale.

They've spent many years and dollars getting these uninterested students to enroll in the first place. These are the kids who shouldn't have gone to college in the first place. Just cut the remedial courses.

I don't disagree. It probably depends on the college and their requirements, but in general I think higher education is literally oversold.

Still I think the interesting question here is how this will play and replay across society. I once worked in a small company where the office manager asked why I was not at my desk. I explained that I had a bug and needed to walk to the kitchen for a cookie to figure it out. She thought this was funny, as was "two cookie problem." Now, with big companies and big headquarters, who is tracking whom? Utopia for conformists?

Too many scuffles in elementary school and you are put on a criminal justice path. Ender's game at scale.

That's not that different from the past. Too many scuffles in grade school was often a path to criminality then too.

To the extent that the vice principal assigned the nascent thug to be the hall monitor, perhaps. But the interesting potentials for dystopia center on "at scale."

Or would anyone argue for a universal ai/surveillance state as utopia? Utopia for conformists?

I worked at a place where they were looking at how many town-halls and other meetings you were going to. People who are burnt out and/or seriously thinking of leaving stop going to these things. No AI involved, just observing en-masse what you normally would observe with your friends in the same situation.

If anything it's a net improvement that the school can notice these things now, instead of just the people who know the student. They can identify problems and hopefully help out before they get too big. The schools would prefer closer to 100% graduation rates.
re patate vs Kartoffel, peat bogs and shuffling cards

The university that has probably attracted the most attention for Big Brother-ish behavior is not the University of Arizona, but Arizona State, with it's eAdvisor, Student 360, and other initiatives. Most of the articles about ASU do not mention tracking card swipes, but this one does:

"And while not exactly matchmaking, Arizona State takes an interest in students’ social lives, too. Its Facebook app mines profiles to suggest friends. One classmate shares eight things in common with Ms. Allisone, who “likes” education, photography and tattoos. Researchers are even trying to figure out social ties based on anonymized data culled from swipes of ID cards around the Tempe campus."

ASU and other schools report increases in their retention and graduation rates. That's fine but I'm still wary of jumping on the Big Data analytics bandwagon. It doesn't matter how many thousands of variables we track and how many millions of observations we have if the variables lack inherent predictive value. Google thought they'd invented a better faster way to predict influenza epidemics, until they found out the next flu season that they hadn't.

I also wonder about the lack of a theoretical underpinning for the data analysis. The machine learning examples that I've seen have been purely empirical, typically by design: assume nothing, let the data speak. What I have not yet seen is how machine learning handles problems such as endogeneity, or more broadly how they distinguish between correlation and causality. I presume that researchers have been working on this, but I haven't seen a paper that actually uses machine learning to estimate a model that is causal rather than merely correlational. Someone told me that Susan Athey has been working on this, but the working papers of hers that I've seen seem to be more like traditional propensity scoring type models, where simple propensity scores are replaced with calculations that use machine learning algorithms.

Good link, and what I was talking about. We can see this as relatively useful and harmless when the "data" only "helps" you while you are at a college. But if there is one rule about data, it gets bigger and more widely shared.

What are the credit agencies doing at the same time? Do they have a conscientiousness score yet?

Big Professor? This is entirely Big Associate Vice President for Campus Life or someone like that. No one gives professors access to information like this.

from the article:

"Those lists, made several times a year, are shared with college advisers so they can intervene before it’s too late. "

At big schools, advisors are not academics and often poorly advise, at least by the lights of faculty.

The next time I encounter a faculty member who undertakes academic advising with anything but the left hand will be the first. It's yet another function they shouldn't have because they suck at it.

Because Gen Z is smaller than the Millennial Generation, many colleges face declining enrollments. Administrators have two options: cut administrative bloat to balance costs or engage in Big Brother tactics in the name of student retention. College administrators are opting for the second option and, in the process, are pushing faculty to adopt all kinds of "high touch" practices. "The data tell us these students are at risk," the admins declare. "Now hold their hands. Help them fail upward."

All the hyperventilating over campus speech distracts from the annoying, and sometimes sleazy, crap that administrators do to ruin college for both students and instructors.

The third option is to make the courses and the grading easier, essentially extending social promotion to college. There is considerable evidence that grading, at least, is already well down that path.

Not much concern expressed in these comments about tracking one's every movement. Of course, that's consistent with the libertarian-authoritarian movement (once considered polar opposites but not now).

Nope, still pretty much polar opposite. I think you are thinking liberal-authoritarian, which is now just redundant.

What is your opinion on private company employee monitoring? If a data service offers a perv-score based on cable viewing should the company buy it, because that's just prudent?

Not a fan on monitoring. I run by a strict 'is it any of your business' system. Some monitoring is necessary to protect the company, customers, and other employees, but you'd have to prove one of those to get my approval.

Well there is a certain group of people who claim to be libertarians who apparently believe in collective rights to exclude immigrants from certain countries from voluntary employment in the US. Some of whom also think that there is a collective right to uproot an individual who has lived in the US since childhood and forcibly ship them to a place they don't even remember. Some of them also think there is a collective right to prevent other individuals from buying products not made by Americans. I don't know if those people really should be counted as libertarians though. If you define libertarian in a way that excludes these people, then you are correct.

There is one person here who claims to be a libertarian but is an advocate for racial, sex, gender, sexual orientation, and religious discrimination. I guess anything goes!

I can think of one soi-disant 'libertarian' who fancies the creed is a compendium of irritation at things which might inconvenience her.

who apparently believe in collective rights

'Collective rights' is a propaganda buzz term what is known, more colloquially, as 'self-government'.

Man the only poster here who triggers you more than me is Hazel. Get a room already!

I don't see open borders as very libertarian. The right to secure one's border is a property rights issue, and is compounded by the non libertarian welfare state.

Also, I don't see anyone calling for exclusion of immigrants. Illegal</immigrants, sure, but that makes sense, they're illegal.

Yes, and why have all the comments missed the fact that all of this data can be tied back to admissions? There's more than one way to boost the retention rate, and the other way takes an awful lot less effort. Though then the administrators might not have anything to do...

They (certainly should) already know precisely who actually drops out, this data just gives them a somewhat noisy list X months earlier.

I suspect they can predict with pretty high accuracy who is likely to drop out and who is likely to complete the degree, before the students ever set foot on campus.

No doubt. And with "big data," they can do an even better job. When rankings are all that matter, and retention factors into rankings, why go to the effort of keeping the current crop of students in school when you can even better preselect for those who will stay in next year's class? Though that may benefit the potential drop outs who aren't admitted...then again, most would still likely waste a few semesters of tuition on a school of even less (perceived) quality.

"I suspect they can predict with pretty high accuracy who is likely to drop out and who is likely to complete the degree, before the students ever set foot on campus."

Close but not quite. Prior to their setting foot on campus, the errors in prediction are pretty high and the r-squared values are pretty low.

But once they're on campus, their actual college behavior and outcomes do add considerably to the ability to predict their retention probabilities.

But yet another unadvertised weak point of all these Big Data Analytics is this: then what do we do next? If this student has a 95% probability of returning and that student has a 20% probability, then we raise a red flag about that student and intervene somehow. But what sort of intervention is needed?

The article describes some concrete examples of how a university can respond, hopefully helpfully, when it sees those red flags. But how well did the intervention work, and was it the best possible one? That's a whole additional and more complex analysis to do: diagnosis is one thing, deciding on the best treatment is harder.

Again, most of these schools are reporting increases in their retention rates. But I can predict that in the next few years we'll see the same phenomenon that Tyler alluded to in another thread about "the truth about experimental economics": smaller improvements in retention.

I would be remiss if I didn't point out that the leader of the libertarian-authoritarian axis is Cowen's Deutschland friend Peter Thiel, who is the founder of Palantir, which is in the business of spying on people for both government and business.

This seems to be the problem:

Most colleges enroll many students who aren’t prepared for higher education
At more than 200 campuses, more than half of incoming students must take remedial courses

Most of those admissions (of students who require remedial work) are essentially frauds perpetrated on the students by the colleges. They should never have been admitted in the first place.

This just exacerbates the dishonesty of the fraudulent high school certifications many of these kids received.

Time, money, and among the best years of their lives wasted. It does no favors to a kid who can’t read, write, or do math at 9th grade level to send them to “college”.

Arizona's state constitution requires the university to be open to all state residents. They don't have much choice in the matter.

The actual constitutional provision is 'Section 6. The university and all other state educational institutions shall be open to students of both sexes, and the instruction furnished shall be as nearly free as possible. "

That doesn't constrain discretion at all except as regards co-education. (Which, of course, isn't going to prevent some judge from making the preoposterous contention that this provision mandates his preferred admissions policy).

The obvious solution in that case is admit them all, and run real 101 courses that flunk out those who can’t do college work, in the first semester. Faster, cheaper, and better for the colleges and the students. Engineering programs, and others where there is a capacity limit in the upper level courses, have done this for many years.

I teach at community colleges. We take anyone with a high school diploma or a GED. We also offer GED prep courses. Someone needs to offer these services and we're it, so it is not surprising that we teach a lot of remedial classes.

Also, a lot of our students aren't straight out of high school. Our median age is between 24 and 26. How much algebra do you remember 5-10 years after high school?

If you are teaching GED prep, you are a high school (although I realize that for administrative convenience the city, county, or state may elect to assign that task to a community college).

After 10 years, not a lot. But a few weekends with a Schwam’s should take care of that. At most, the first couple of weeks of class. Repeating high school (or jr high for Alg 1) is not college.

I taught a couple of sections of computer science at community college, as a post retirement bucket list thing. You have my sympathy, your task is not easy.

You can debate the appropriate name for what we do, but from a public policy standpoint we are called community colleges.

Based on my experience, teaching algebra to someone years after they last took the course takes a while. I suspect that the composition folks in the English department have the same experience with Comp 101.

While our task is not easy, it is rewarding. We can debate from a public policy standpoint whether or not it is cost effective, useful or whatever. Right now the decision is that someone gets to do it and that's the community colleges.

"It’s really not designed to track their social interactions, but you can"

and soon will.

Really appalling invasion of privacy. The University is a government entity so I hope they get sued soon.

The press releases reported high recall, but did report any measure taking into account false positive rate. I wonder what that is, since the idea seems to be to intervene to prevent dropping out--gotta understand something about efficiency of potential interventions.

The whole business is seriously creepy. The legislature needs to put these data dredgers out of business.

Can't legislate away the future, Art.

Sounds like a job for data privacy laws!

I'd wager a small amount of money that the single most predictive variable in the card swipe data is how often the students go to class. Or some reasonable proxy, like the total number of on-campus swipes.

The textbook companies are datelining the students more and more.

Fascinating. I was aware that textbook publishers were creating ever-increasing amounts of digital content (just as old-fashioned paper textbooks often had an accompanying paper workbook or supplemental text) but I was not aware that the digital content had become so extensive -- and expensive.

I don't teach anymore so I don't have firsthand experience with these digital offerings. I could see how they could improve student learning as Mankiw suggests. But I could also see how they represent another money-making scheme. Or a way for busy -- or lazy -- professors to outsource some of their teaching duties. I taught at small schools, and never used the problems in the textbook or the supplemental texts; I wrote my own questions for my problem sets and exams.

Comments for this post are closed