Your smart phone can predict your grades

Interesting but worrying too:

The SmartGPA study uses passive sensing data and self-reports from students’ smartphones to understand individual behavioral differences between high and low performers during a single 10-week term. We propose new methods for better understanding study (e.g., study duration) and social (e.g., partying) behavior of a group of undergraduates. We show that there are a number of important behavioral factors automatically inferred from smartphones that significantly correlate with term and cumulative GPA, including time series analysis of activity, conversational interaction, mobility, class attendance, studying, and partying. We propose a simple model based on linear regression with lasso regularization that can accurately predict cumulative GPA. The predicted GPA strongly correlates with the ground truth from students’ transcripts…Our results open the way for novel interventions to improve academic performance.

That is from a new paper by Rui Wang, Gabriella Harariy, Peilin Hao, Xia Zhou, and Andrew T. Campbell (pdf).  Class attendance, by the way, does not predict grades very well.

For the pointers I thank Eric Barker and Dan Gould.


novel interventions

I can imagine what those would be....

Great idea for an app. Can't be an Asian mother anymore after your child has gone off to college? Buy the Virtual Tiger Mother Application™ and our software will automatically generate reports you can use to guilt your child with.

Hey, Big Data is the only way that we can adequately measure whether designer babies are meeting their targets. Just another small step to a much better world, ensuring that the correct behavioral traits are being followed. Think of it as nature and nurture - the smart phone could even become a sort of Skinner Box device, without the road bumps of the original model.

That is, why they aren't asking questions on exam that could only reasonably be answered by attending class. Of the order of "what did I eat for breakfast?"

Class attendance policies negatively impacted my grades as I could have been studying/sleeping instead of preparing to attend class, attending class, and so on. Our education system is not set up to educate students.

In the paper I found the following statement: "Research that examines academic behaviors has also found that absenteeism and class attendance predict academic performance, such that students who attend class more often performed better than those who missed class [8, 11, 18]. For example, a meta-analysis of studies that examine the relationship between class attendance and performance found attendance to be strongly related to class grades and GPA among college students [12]. However, the StudentLife study [43] found no correlation between class attendance and academic performance."

I was puzzled by the references: does research generally indicate that those who attend class do better than those who do not? Does this small study reach a different conclusion? This seems like a very important question.

Didn't read the paper, but "no correlation"? What does that mean? Were they just looking at bi-variate correlation? That's not very meaningful.

I imagine class attendance does not predict grades well because even though attending class is helpful, people who are really concerned about how well they're understanding the material will tend to attend class more regularly, and those tend to be people who would otherwise have lower grades. It's a selection effect, this is an observational study not an RCT.

Which of course is wrong. We should be promoting a behaviour of attendance, participation, and interaction on a regular and predictable schedule - partly for discipline but mostly for making effective use of this ridiculously expensive institutional property. Education has to be more than merely content understanding and delivery - the instructor should be delivering content in a way that exclusively makes use of the 'in person' delivery. Why have 'class' if a significant section of the class just uses it as remedial or supplemental - why have physical institutions at all? It does not seem cost effective to me to build in that much 'redundancy' into the learning system - especially at the tertiary school level. nonsense.

Caplan, Case Against Education.

N = 30 undergrads. it appears high-performing students self-selected into the study, as the average spring gpas is above about 3.3.

there are 193 possible regressors. with 53 gigabytes of data from continuous monitoring of smartphones, aren't you going to find a lot of spurious correlations?

choosing which to report as "findings" seems somewhat arbitrary. in table 4, every regressor--including class attendance--has a p-value < 0.05

can the lasso methodology really make your research conclusions for you? for the overall gpa, it chooses three regressors. for the spring gpa model, the lasso doesn't choose any of the 193 possible regressors! are we to infer then that no behavioral traits affect gpa?

8/10 lectures:

Professor moves at glacial pace to cover material so even the kid who had no business passing high school can follow. This kid flipped through his phone the whole class and left early.

About 2/3rds of the way through, hot but vapid chick asks a tangentially related question that has little to do with content in class but is sorta kinda loosely related to material, so professor wants to try and be intellectually inclusive about it, but its really a bad question so he's got to do backflips to do it, so it takes up rest of day.

Professor then looks at watch and goes "oh gosh, we didn't have a chance to get to chapter 6 but it'll still be on the test"

Meanwhile I spent a half an hour showering and getting ready, 20 minutes in commute, 15 minutes in line for coffee so I can stand it, and another 20 minutes getting back.

Class is really for the gomers and the second quartile....the only thing I got out of it was reminders about lectures/homeworks...

Hold the presses. Alert the media. Less partying and more studying leads to better grades. This stuff is sort of funny, but sadly actually not. Next up: lasso regularized (seksi!) smart phone data shows that subjects who went to work were less likely to get fired than those who didn't. Keep minting these post-docs! Really!

The interesting question is whether the partying kid who scraped through -or not- did better in society for having 'attended' university - and thus improved society for having gone. I would argue that yes they did - though it is likely $50k+, if not also 3-4 years, poorly spent.

"...likely $50k+, if not also 3-4 years, poorly spent."

You're obviously referring to a public university..."$50k+"...but, given the alternatives available to children 18 to 22 years attending such a school, was it really "3-4 years poorly spent"? It's worth that much just to push them out of the nest and in the future signaling value of the degree...e.g., they will get that public sector job before someone without a degree.

Now a private university, that's another story. Generally the same quality of student, with maybe slightly better options post-high school depending on parents wealth, but the cost is 4x higher. That's "poorly spent".

Another triumph for immigrants, doing the useless academic work Americans don't want to.

"Our results open the way for novel interventions to improve academic performance." Not if by "improve academic performance" you mean "help students get a better education".

Did the smartphones capture student willingness to cry and/or give BJs during office hours? If no, they missed the boat.

Interesting. Thank you for posting.

Well, if you read the study carefully, the study mostly predicts that more studying and less partying leads to better grades (e.g. see page 7) among these N=30 students. All the smartphone is doing is more accurately reporting which one they are doing.

All this really does is give helicopter parents another tool (smartphone surveillance) to use to know when to nag their kids.

So we could save a lot of time and money and just grade students on their smartphone use. The results would probably be just as informative as any information we get from their teachers.

Comments for this post are closed