XSEDE Science Successes

« Back

Modeling the perfect classroom

Modeling the Perfect Classroom

XSEDE used to see effects of "student tracking"

Supercomputers have been used to model intricate biological cells and the greater cosmos, but now XSEDE resources and help services are helping school teachers and administrators figure out how to help their students learn.

The Stampede supercomputer, housed at the Texas Advanced Computing Center at The University of Texas at Austin, along with XSEDE's help service called Extended Collaborative Support Service (ECSS), allowed a research team to analyze "school tracking."

Instructors use school tracking—classifying students based on their innate abilities—to make sure they do not attempt to "teach to the middle" while ignoring the highest- and lowest-achieving students. By pairing the best students with other high achievers, the argument goes, those students can accelerate their learning path, while other peer groups also get group-specific learning. The same goes for each group of students.

Plenty of educators, administrators and even civil rights groups have found tracking to be a form of re-segregation because many times minority students are left out of "gifted" or "advanced" classes, especially in cases where the decision to move a student to a different group is made without well-considered and tracked test scores.

A University of Wisconsin researcher recently used XSEDE to help parse through the long-standing arguments for each side. Her work can be seen in a working paper here:http://www.ssc.wisc.edu/~cfu/tracking_paper1.pdf.

"Tracking benefits high-ability students but hurts low-ability students. However, ignoring effort adjustments would significantly overstate the impacts," said Chao Fu, associate professor and Mary Phipps Senior Fellow at Wisconsin.

"We then illustrate the tradeoffs involved when considering policies that affect schools' tracking decisions. Setting proficiency standards to maximize average achievement would lead schools to redistribute their inputs from low-ability students to high-ability students," added Fu.

Fu and co-author of the working paper, Nirav Mehta, an assistant professor at Western University (Ontario, Canada), used XSEDE to take many factors into place, most notably including the "effort adjustments" made by students, their teachers and their parents. Specifically, they studied  three interrelated components that help inform their computer model:
1) acknowledging the changing peer composition when students move through the ability tracks,
2) by tracking both schools' and teachers' efforts, it's possible to see how students would behave if tracking changes are made and
3) by studying tracking decisions, they can see how policy changes would alter school tracking regimes, track-specific inputs and parent effort.

"XSEDE requires heavy computing in a parallel setting and XSEDE is perfect for that," said Fu. "If other researchers in this field, or any other, require heavy computing, XSEDE will help first make the project possible, and then faster."

Helping the helpers
According to Fu, much of the work done by her research team could not have been done without XSEDE's ECSS program, which gives coding and other technical help to research teams to help them best use XSEDE's advanced digital services.

ECSS expert Junqi Yin was more than willing and able to help.

"Professor Fu has several home-grown codes, using mostly either serial or OpenMP to parallelize parts of the computation," said Yin, an expert in the scientific computing group at the National Institute for Computational Sciences (NICS) at the University of Tennessee, Knoxville, and part of the Extended Support for Training Education and Outreach with XSEDE. "In order to efficiently utilize XSEDE resources, like Stampede, and speed up the time to solution, we parallelized the serial codes and optimized parts of OpenMP directives with MPI schemes. The production code now scales up to hundreds of cores on Stampede, which is what Professor Fu wanted in the first place."

"Junqi Jin was extremely helpful, helping us deal with technical and coding problems," said Fu. "We would not have been able to solve those problems on our own."
 
 

Average treatment effects (ATE) of banning tracking: equilibrium vs. fixed effort inputs. When tracking is banned, lower-achieving students receive more inputs from the school, in terms of both peer quality and school effort. In response, parents of these students reduce their own effort. Failing to take into account this reduction drastically overstates the ATE of banning tracking for these students. Chao Fu, University of Wisconsin.