Carnegie Mellon pioneers AI project with U.S. Navy
August 20, 2018
With help from the Office of Naval Research, the private university looks to develop AI capabilities for humanitarian aid and disaster relief around the world.
Can radio sensors and video cameras help teachers understand their students? A network of 17 Montessori schools across the U.S. is trying to find out.
As schools around the country grapple with the consequences of infusing technology into nearly every aspect of learning, a network of 17 Montessori "lab classrooms" nationwide is testing gear that would give educators even more access to data that tracks student activities throughout the school day.
Only, in this case, it's not about what students are doing on their devices. It's about where they go and what happens when they get there. The observational technology — which so far includes video cameras and in-shoe radio sensors — records and stores the movement and orientation of students in a classroom, offering educators a chance to review that data and better understand student behavior and engagement.
Three schools in the network, called Wildflower, began piloting the technology in early 2017. Ted Quinn, who joined Wildflower after almost a decade as senior vice president of organizational learning at Teach For America, sees the Montessori schools as an ideal environment to implement educational technology that assists — but does not intrude on — the relationship between teacher and student.
For Quinn, the key question is this: “How can we gather and make meaning of all of the fine-grained information about how students are interacting with educational activity and the educational environment, but do it in a way that respects the educational environment?”
Ready-made for experimentation
The radio sensors and video cameras are the two most ambitious projects in a series of controlled pilots that Wildflower is running, including tests of automated attendance platforms, smart pens for teachers and smartwatches that provide classroom data. In all of those cases, the technology might be cutting-edge, but it generally doesn't interfere with the principles of Montessori-style education, which focuses on individualized, experiential learning.
Classrooms are often void of plastic or metal surfaces and screens — a fact that inherently changes how Montessori schools think about edtech.
"Many of the ways that people are tackling [edtech] by necessity involve children interacting with problems on screens, with headphones on, a kind of online, machine-based instruction experience that can be interesting in its own right, but is not what we were interested in," said Quinn, who oversees the Wildflower schools' various technology initiatives and is a partner in the Wildflower Foundation, a nonprofit organization that provides funding and resources for the network of shopfront schools that bear its name.
"We were interested in asking the question of whether we can tackle the same challenge, but do it in an environment that is much closer to what all of the emerging research on neuroscience and child development would suggest is actually optimal for learning," Quinn said.
Technology is part of Wildflower's history, though. The network — whose individual schools are generally nonprofit — started as a single shopfront classroom in Cambridge, Massachusetts, in January 2014, when former Google personalization head and MIT Media Lab Professor Sep Kamvar didn't like what he saw from traditional elementary schools available for his 3-year-old to attend. Kamvar appreciated the benefits of edtech in traditional elementary classrooms, but not the role it played, often dominating lesson plans.
In current Wildflower foundation CEO Matt Kramer and Quinn, Kamvar found contemporaries who shared his vision of an edtech-enabled, but not led, classroom, and the Wildflower network was born.
Getting the tech on the ground
The radio-sensor technology — which has been piloted in three schools around Cambridge — is embedded in the slippers that students are required under school rules to wear when they enter a classroom. The sensors are similar to what you might find in a Fitbit or Apple Watch, Quinn wrote in a blog post, and are in constant communication with each other — providing data on the strength and distance between each sensor, allowing for a map to be made of the classroom and student locations based on how students move around.
By constantly transmitting location data, Quinn explained, Wildflower teacher leaders are able to go back at the end of a school day and track a student's location around the classroom from the moment he or she enters it. While that might not sound like a ground-breaking achievement, the possibilities are exciting, Quinn said.
"At a very basic level, what it might allow us to do is infer the kind of things that we always want to know as educators — where is the child, what do they know, what do they not know, where should we focus our energy next? And we do that in a way that minimizes testing, because we’re inferring that information just from the student’s authentic educational activity," he said.
Montessori students often have a large say in developing their own lesson plans, and standardized testing is not applicable with the style of learning that teacher leaders facilitate. Instead, teacher leaders often gauge progress on an anecdotal and observational level by taking notes. With the tangible patterns that observational technology provides, however, educators see an opportunity to confirm with data the trends they can spot themselves.
For Mary Rockett, a teacher leader at the original Wildflower classroom in Cambridge, the data collected from the technology pilots provides a tool to help validate what she's observed over 30-plus years as a Montessori teacher.
"I can look at [the data] over a week, a month, a year, the whole three years that I have a child in my class — I don't have a way to do that by hand," Rockett told EdScoop.
Having the ability to review student movement around the classroom, Quinn explained, can allow a busy teacher to observe who a particular student is interacting with and who they are avoiding; which areas of the classroom are getting the most use; and which areas are failing to engage students. For a 25-student, open-floor-planned Montessori classroom, technology that can fill "observation gaps" is a boon for teachers, Quinn said.
"How does this child respond to different ways that we try to support him or her? Auditory or visually? We can learn those patterns as well," Quinn said.
The radio sensor pilot program is currently in its third iteration, after launching in three classrooms in January 2017. After each phase, the sensors and their data collection methods are refined and teacher input is used to improve ease-of-use.
Fully observed, but protected
The second, accompanying phase of Wildflower's technology initiative — the introduction of video cameras in classrooms — recently entered its second iteration, with Quinn optimistic about its potential to improve educators' ability to understand student behavior.
"We haven’t proven this case yet, but what we hope is that we can tell the difference between a child that is sitting with the material but not actively engaging with it, and a child that is in very deep concentration with the material," Quinn said of the video cameras. "We can tell that based on how they’re orienting towards the material, and that kind of distinction is very important for Montessorians."
The slippers and the video cameras work in tandem. The cameras, placed discreetly around the classroom, use machine learning to extract data from the footage. The cameras track the ID numbers that each radio sensor has, mapping each individual student's location. Quinn said that once the basic location and orientation of students is extracted, the rest is thrown out — a security measure, along with Wildflower's use of a virtual private network for the whole system.
Quinn said parents are actively informed and must give affirmative consent for their child to be identified by the observational technology. If no consent is given, the student just shows up as an unidentified data point, meaning their presence is not mapped and their data is not available for the teacher to analyze.
Wildflower parents, Rockett said, have been very receptive to the pilot. They understand the experimental nature of the classroom, and realize the data can be used to better explain student behavior when it comes time for parent-teacher conferences. Hypothetically, they could opt out whenever they want to — a critical aspect of privacy, a security expert told EdScoop for this story — but no one has elected to do that so far, Wildflower officials said.
While Wildflower is diving head-first into these technologies, the future of them remains uncertain.
The observational tools that the three classrooms are piloting could be used in traditional schools, and Wildflower is committed to making its findings open-source for others to access and build on. Even as the technology goes through multiple iterations, though, there's no guarantee any of it is here to stay. That decision falls to the teachers and parents.
"Whether this is successful or not is based on whether teachers think the information we’re producing for them is helpful," Quinn said, "and that's going to be when this gets interesting."
Reach the reporter at email@example.com and follow him on Twitter at @RycJohnston.