Uncategorized

Learning from Our Sensor Technology

By November 21, 2017 No Comments

In a previous blog post, we described our plans to pilot emerging technology to support our teachers in enhancing their existing Montessori classroom observation and record-keeping. If you’re unfamiliar, observation is a key component of Montessori education. Teachers observe students as they go about their work, take careful notes, maintain meticulous records about student engagement and progress, and regularly reflect on all of the above in order to build their understanding of each student’s unique development path and plan their actions in the classroom.

Our aim in this pilot was to better understand how we could support teachers’ extensive classroom observation efforts. We are starting with simple tracking of interactions, but we’re working toward being able to gather a wide range of information that can inform teacher practice and capture critical information about what students know and are able to do in a way that connects to the real work they do. The ultimate goal is to enable data-driven practice in a fully authentic, child-centered environment.

In the pilot, we put small, low-power sensors – similar to what you would find in a Fitbit or Apple Watch – in each child’s shoes and in classroom materials, so we could detect the moments in the classroom when children, teachers, materials and classroom areas were near each other. We used this proximity to give us a rough idea of when educational interactions were happening. We then built an online interface where teachers could review the synthesized data.

The pilot was short (only a few months), but we learned a ton. First, we learned a lot about the specific technical challenges of our system. None of these were unexpected, but it was invaluable to be able to gather detailed data about these challenges in a real-life classroom setting.

We learned (as expected) that it’s a burden on the teachers to have to gather up all of the sensors at the end of each day in order to transfer the data and to change the batteries in the sensors at the end of each week. For technology that intends to save teachers time, adding tasks to their to-do lists is a negative.

We also learned more about all the ways in which the sensors can be unreliable and, more significantly, all the situations in which proximity is not a good proxy for educational interactions. For example, just because a child is sitting next to a lesson tray doesn’t mean they’re using it.

More broadly and interestingly, we came away with a much richer sense of the kinds of information that are most valuable and useful to Montessori teachers. As we talked with the teachers and looked at the data together, it became very clear that a simple record of what the students were doing was a great start. In any Montessori classroom (typically about 25 students), it’s hard to keep track of what each child is doing. We confirmed having accurate data recorded in real-time could fill a much-needed observation gap for teachers.

However, to make the data really useful, we need to augment this record with evidence of student interest and engagement. That is the most important evidence of student progress, and it guides the teacher as they determine what lesson they should introduce to the child next. These include things like the degree of independence of the work choices, patterns of repetition in those choices, and indicators of deep concentration vs. distraction, etc.

We also learned that, while the prospect of purely passive and automatic data collection remains intriguing to our teachers, there’s also much more we can do to make it easier for them to capture and make use of their own observations. In particular, there is a lot that teachers can observe even in passing that is simply too time-consuming to write down and transfer into a record-keeping system.

All these learnings have sparked new areas of exploration. By the end of the pilot, we managed to fully automate the upload process, so that data was flowing continuously to the cloud while the sensors remained in place. We also identified a number of ways to increase battery life, and we plan to implement these in the new version of the hardware.

More significantly, in order to increase the accuracy of the interaction data and to begin to capture subtle indicators of student engagement and interest like where they are focusing their eyes and how they are using their hands, we are exploring the addition of a layer of computer vision to the system. Initial tests are very promising and we can see that the system is starting to capture some of these subtle indicators.

In order to make it easier for teachers to capture and make use of their own observations, we are starting to work on new mobile and smartwatch interfaces designed around the kinds of things that teachers want to observe regularly (e.g., a quick scan of the classroom every 15 minutes to see who has work out and who is concentrating). For teachers who don’t want screens in their classroom, we are investing in smartpen technology that will allow teachers to mark up paper forms and the resulting data flows in real time to our servers.

Finally, we are continuing to sketch and experiment with different ways of visualizing all of this data, getting advice from our teachers at every stage as we learn our way together. We’ll keep you posted as we continue this exploration of new ways of collecting and using data in the Montessori classroom.

Ted Quinn

Author Ted Quinn

Ted explores innovation that has the potential to support and strengthen the authentic Montessori environment of our classrooms

More posts by Ted Quinn