Education has remained large­ly unchanged for mil­len­nia. In any class­room, you see a set of stu­dents gath­ered around a teacher who’s writ­ing on the board, or maybe now we’ve added a PowerPoint deck. But, as in many oth­er fields that have been slow to change, the data rev­o­lu­tion is com­ing for education. 

Now, many peo­ple think that this rev­o­lu­tion will come through MOOCs, or mas­sive open online cours­es, in part because com­put­ers can real­ly eas­i­ly col­lect data on all of the sorts of things that stu­dents do. So, where they click and when, what they type into a dis­cus­sion forum, whether they click pause or play on a video.

However, this data is seri­ous­ly impov­er­ished with respect to what stu­dents actu­al­ly think and do in order to learn. And despite the hype MOOCs aren’t the rev­o­lu­tion that we’ve been told they are. On the oth­er hand, real class­room data has enor­mous poten­tial. If only we could get it. And that’s where con­nect­ed sen­sors come in. This is one of the things that my lab is work­ing on.

So, rich data is con­stant­ly being pro­duced in the class­room. Student engage­ment and emo­tions are one of the strongest pre­dic­tors of learn­ing. And facial expres­sions are the data that we use to under­stand this. So, does the stu­dent look frus­trat­ed, con­fused, bored? Are they pleased, excit­ed, or or sur­prised about some­thing? We can also use voice data. So every­thing that the teacher says, every­thing that the stu­dents say, this is very impor­tant data that tells us about learn­ing. So, what types of ques­tions [is] the teacher ask­ing? Are the stu­dents deeply col­lab­o­rat­ing on a top­ic? Are they chal­leng­ing one anoth­er’s ideas in an inter­est­ing way? All of these are real­ly impor­tant pieces of infor­ma­tion about how stu­dents learn. 

Now, nor­mal­ly the teacher is the only one who acts as a ​“sen­sor” in the class­room. So, they’re con­stant­ly col­lect­ing all of this data and fran­ti­cal­ly try­ing to make deci­sions in the moment about what to do. And then maybe at night they can go home and take a minute to reflect by look­ing at the assign­ments that their stu­dents have pro­duced. But until now, it’s been incred­i­bly dif­fi­cult to get a real under­stand­ing of what’s hap­pened moment-to-moment in the class­room. This is where con­nect­ed sen­sors come in. 

So, using things like depth cam­eras or micro­phones arrays, we can col­lect all of these data on the stu­dent and teacher actions that I’ve been telling you about, and then use machine learn­ing tech­niques to under­stand and deep­en our knowl­edge of what works for teach­ing and learn­ing. So, you could imag­ine that such a sys­tem would be like a per­son­al infor­mat­ics sys­tem. You might even be wear­ing a track­er for this type of info on your wrist right now. So, we can right now track the the num­ber of steps that we take, the calo­ries in the food that we eat, or oth­er sorts of infor­ma­tion about our our dai­ly lives.

Now, sim­i­lar­ly to how the sys­tems tell you about your weight loss over time, imag­ine that we can now present the teacher with sum­maries of the most impor­tant things that hap­pened in their class that day. So, we could present this to them, and the teacher could, reflect set their goals for mov­ing their class for­ward, and mon­i­tor their own progress towards these goals. 

Now, for me, even more excit­ing than a dash­board is the abil­i­ty to use these con­nect­ed sen­sors, in real-time, in the class­room, in order to improve teach­ing and learn­ing. So, my lab at Carnegie Mellon is right now auto­mat­i­cal­ly col­lect­ing all of this sort of data in the class­room, and dis­play­ing it to our instruc­tors in periph­er­al displays. 

And of course, this requires some very care­ful design to make sure that it’s a ben­e­fit to learn­ing and not a dis­trac­tion. But as you can see here we’re able to use our own uni­ver­si­ty class­rooms of a liv­ing lab­o­ra­to­ry in order to test the types of tech­niques that might actu­al­ly sup­port learning. 

So here we’re flash­ing the screen read when the teacher’s been talk­ing for too long, and they might want to ask a ques­tion. But then they can go back after class and look at all of the data that’s been col­lect­ed and say, ​“Boy, next class I think I wan­na try to get more stu­dents in the class to par­tic­i­pate. And maybe I can get them to talk ear­li­er on in the class instead of this talk­ing on and on myself.”

Now, I’ve said that MOOCs aren’t the answer, but we can in fact use these same tech­niques to improve learn­ing online as well. So, we can use the cam­era, the micro­phone, that’s already on your com­put­er to detect who’s work­ing. Are they frus­trat­ed or con­fused? And then we can use the same sorts of data to improve the feed­back that we’re giv­ing these students.

Of course there come some risks with this type of data, but isn’t it worth it to share your data on whether you’re the qui­et one in class in order to give the next gen­er­a­tion of teach­ers the oppor­tu­ni­ty to give every­one in the room a chance to speak? And I think that par­ents and even stu­dents them­selves could ben­e­fit from hav­ing this information. 

Now, the White House and many oth­er agen­cies are cur­rent­ly mak­ing rec­om­men­da­tions about just how impor­tant data is going to be for improv­ing edu­ca­tion. However, in the rush to move to online learn­ing, we should­n’t for­get about the rich data that we can get from face-to-face class­rooms and how this can help us improve edu­ca­tion everywhere.

Thank you.