Artificially intelligent digital agents are being marketed as a way to automate note-taking in the workplace, raising a big question for K-12:
Are classrooms next?
Take, for example, EVA, a 鈥渄igital voice assistant鈥 created by Silicon Valley startup Voicea. The AI agent can automatically read users鈥 calendars, dial itself into their meetings, and use natural-language processing algorithms to create real-time transcripts of what鈥檚 said. As a meeting progresses, EVA can also respond to voice commands (鈥淓VA, add that to my to-do list鈥) and 鈥渢rigger鈥 words (鈥渢hat鈥檚 a good point鈥) to highlight what鈥檚 most important.
In an interview, Voicea CEO Omar Tawakol described the technology as a way to help the masses employ the same listening and learning skills as top executives.
鈥淩eally good CEOs are 100 percent focused on their conversations, not looking at a screen,鈥 Tawakol said. 鈥淥bviously, the same thing is true in classrooms. You don鈥檛 want people on their phones or opening up their laptops pretending to take notes.鈥
But for the time being, at least, even Voicea is keeping its distance from the education market.
There are practical concerns: Automated transcription in busy high school classrooms or large lecture halls is far more difficult than in a controlled conference room or online meeting platform.
In addition, an emerging body of research on computer-based note-taking offers reason to be skeptical that AI note-takers would actually help students learn.
And before adopting new tech, educators should drill down to specific questions on what problems they鈥檙e hoping to solve, for which students, under what circumstances, said Robert F. Murphy, a senior policy researcher at the RAND Corporation who has written about artificial intelligence in K-12.
鈥淚 applaud attempts to think about applications of technology that can help students take more effective notes and highlight the key takeaways,鈥 Murphy said. 鈥淚 just don鈥檛 know if this particular application is going to provide that.鈥
Taking Notes With Tech
Before considering what鈥檚 technically possible, it鈥檚 worth asking whether outsourcing note-taking to artificially intelligent digital agents is even a good idea.
Big picture, taking notes can help students learn in two different ways, said Linlin Luo, a postdoctoral researcher at the University of Regensburg in Germany.
First, the mere act of taking the notes鈥攍istening to what鈥檚 being said, deciding what to record and highlight, and connecting that information to your prior knowledge鈥攊s valuable in and of itself. Researchers call this the 鈥減rocess function鈥 of note-taking, Luo said.
Second, having a set of notes that can be reviewed later is also useful. This is note-taking鈥檚 鈥減roduct function.鈥
The proliferation of digital devices in the classroom is already changing both, according to Luo.
Because students can generally type faster than they can write, for example, taking notes with a laptop tends to result in a more complete set of information. Laptop note-taking also reduces the cognitive load associated with multitasking, making it possible for students to process what they are hearing in the moment. And a study by researchers at California State University, Northridge suggests taking notes via laptop is particularly valuable when assessments are also delivered via computer.
At the same time, however, taking notes via laptop has its disadvantages. Students using such technology take fewer 鈥渧isual鈥 notes, such as charts, drawings, and maps, Luo said. Their notes also tend to contain more verbatim transcription, with less attention paid to expressing important or confusing ideas in students鈥 own words. And hand-written notes often seem to serve as a better long-term 鈥減roduct鈥 for students to review later, especially when it comes to understanding concepts and ideas.
The ideal scenario, then, is likely a rich set of notes, that includes a mix of text and other forms of information, taken in a 鈥済enerative鈥 way that allows for deeper mental processing and connections to prior knowledge. And such notes should be something that students actually review later.
Could artificial intelligence help make that vision a more consistent reality?
Tools such as EVA could certainly help students take more complete notes, allowing them to pay closer attention to what is being said during class, Luo said.
But more information isn鈥檛 always better. Both the process and the product functions of note-taking could suffer if note-taking is outsourced to AI, she said.
鈥淚f something else is taking a complete set of notes for students, I don鈥檛 know if that will help them integrate new information with prior knowledge,鈥 Luo said.
鈥淗ow many things do we record, and think we鈥檙e going to go over it, but we never actually do?鈥
Limited Classroom Applications
From a practical standpoint, other AI-powered technologies already in use by K-12 schools鈥攊ncluding adaptive software, automated essay-scoring tools, and voice-activated 鈥渟mart鈥 speaker systems鈥攁lso offer lessons for what to expect from digital note-takers.
So far, the K-12 applications of such tools are mostly rudimentary, said Murphy of RAND. Few use machine learning and other advanced AI techniques that allow digital tools to discover patterns and identify relationships that are not part of their original programming. And while some educators are excited about the instructional potential of higher-powered AI consumer products, so far tools such as Amazon鈥檚 Alexa speaker systems are mostly being used for basic purposes like spelling practice and managing classroom transitions.
There are also significant privacy worries, Murphy said. And many schools鈥 broadband infrastructure is not currently equipped to support widespread adoption of such tools.
In plotting Voicea鈥檚 business strategy for its EVA note-taking assistant, Tawakol has been cognizant of such issues.
Compared with the workplace, 鈥渢he problem with class notes is that audio capture is not always good,鈥 he said, making the accuracy of AI notes worse than what students can often record on their own. It鈥檚 a big reason why Voicea is not currently going afer the education market.
Furthermore, try to imagine: Would teachers and students need to consent to have their every classroom discussion recorded and processed by digital agents? What might a real classroom look like if every student is busy telling their digital assistant what to notate during a group discussion?
Add all the questions up, and the best use cases for artificially intelligent note-taking systems in the near-term might be relatively limited, said Murphy of RAND.
鈥淚f you鈥檙e absent or have a learning need that makes it difficult for you to keep pace, I could see value,鈥 he said.
And longer term, the companies behind artificially intelligent note-taking systems will have to offer more than just automated transcription if they want to break into the education market, Tawakol said.
That could mean making it easier to highlight what鈥檚 important, or to quickly navigate notes, or to share key bits with teachers or other students.
And hypothetically, he said, digital assistants could also be pitched as a way to alleviate students鈥 concern that their own notes won鈥檛 capture everything. That could free them up to use other tools鈥攅ven pencil and paper鈥攖o capture in-depth thoughts and reactions to the material that most impacts them emotionally and intellectually.
鈥淭he innovation has to be in the user interface,鈥 Tawakol said. 鈥淵ou could get the best of both worlds.鈥