Currently, I’m reading Rise of the Robots: Technology and the Threat of a Jobless Future by Martin Ford. As an educator, I’m always on the lookout for insights and information about my field as I read. Specifically, I find myself focusing more and more on how we learn and how we process information. This line of inquiry has led me to books like Ford’s, which discuss how machines learn and process information.
This morning, I got to a section of Ford’s book that involved an explanation of how machine learning works, how deep learning works, by giving the example of how Google Translate developed and continues to develop. Google Translate didn’t develop through gathering all the world’s language experts and having them explain to coders how to proceed with a translation tool. Rather, Google developed recursive learning algorithms that mined the loads and loads of data available on the web. As Google Translate attempts to do its work, it always offers users to provide feedback on the translation. In this way, the algorithms developed for Google Translate are continuously improved upon. The same is being done for computer programs that seek to automatically describe contents of images.
Look at what’s going on here: we have a huge amount of data, an algorithm that seeks to explain the data in a way useful to us, plus input from us. What we want in return is new information that specifically addresses our information need. And all this is happening. This is now.
Although Ford’s book is getting me a little freaked out by just how much machines are going to replace workers of all sorts, I have an idea.
I’m a digital portfolio advocate. I think that digital portfolios offer students the opportunity to more authentically show what they know and what they can do with what they know. My request to teachers has been: implement learning activities to help your students learn concepts, and then have students prove that they learned those concepts through the development of digital portfolio pages. While I believe this is a more natural expression of knowledge, and a more accurate method of assessing what students really know, when students are producing images, videos, audio, written explanations, and infographics to convey their understanding of a topic it’s incredibly challenging for teachers to view the variety of evidence provided by students in order to reach conclusions about a student’s level of understanding. Evaluating learning based on learning portfolio creation and curation is valuable, but difficult. I think this is part of why so many teachers haven’t jumped on the digital portfolio bandwagon, yet.
It’s so challenging to tell students: “We just read this novel together and learned about several concepts related to literature. Create a digital portfolio page to show me what you learned” and then go through what students produce in order to evaluate learning. Giving students tasks like that is very valuable for learning (it gives students opportunities to creatively express their knowledge; it helps students take ownership of their learning and the learning process), but it’s legitimately difficult for educators to evaluate learning this way. No wonder so many educators are continuing to use inauthentic learning assessments like multiple choice tests. It’s easier to throw a multiple choice test at a student and call it an accurate assessment of learning rather than allowing students to really show what they know through presentations and digital portfolio curation. I’m incredibly grateful to the educators who do seek to have students more authentically engage in the learning process through digital portfolio curation.
But getting back to my idea, it seems very much in the realm of possibility for Google to develop a system (similar to what they did for Google Translate), that could mine student produced content for the purpose of evaluating learning. The algorithms of that system could mine student digital portfolios (their embedded writing, images, videos, etc.) for:
- originality (it’s gotta be the student’s own words/images/videos)
- accuracy (comparing learning iterated to current understanding of that concept)
Teachers could use the feedback of this system to help students improve their learning. Students could be free to use their creativity to express their learning. Creativity would be unleashed and valued in educational settings, and the hypothetical Google data mining system could provide feedback on learning. Again, this seems very much in the realm of possibility based on what machines can currently do.
A lot of the talk happening now around mining data is raising important points about ethics and individual privacy. Frankly, our data–how we interact with information online–is being collected and used in creepy ways to make money. I agree that issues around how data is collected and used need to be addressed. Still, I think that one way that tech companies like Google can redeem themselves is by developing tools like the one I imagine above. Use big data to help learning. Oh, and make it free, please. I’m all out of money. I just spent my allowance at the comic book store.
Finally, what about the title of Martin Ford’s book? Wouldn’t this kind of tech innovation threaten teacher jobs? If we had a system that mined student-produced digital content for evidence of learning, wouldn’t that mean we no longer needed teachers? Not quite. With a system like the one I describe above we’d certainly have teachers spending less time grading student work (millions of teachers would weep over not getting to grade student work on nights and weekends), but we would still need teachers to develop plans of action concerning the results of student work–and to create initial learning experiences in the first place. Teachers can then focus on questions like: “What new learning experiences do I need to design to help my students learn this concept? What, specifically, according to the data, do I need to help students with? Which students need what?” The machine isn’t going to teach. It’s going to get better at making connections between what students are producing and the concepts we wanted them to learn.
And Google, as I have no idea how to design such a system, I hereby release any notions of propriety around my imagined digital portfolio data mining and feedback algorithm. I release the idea to you or whatever other tech company is interested in helping us all better understand the universe and each other with a free educational data mining tool. Remember, “Do the right thing.”