Last year, officials at the Montour school district in western Pennsylvania approached band director Cyndi Mancini with an idea: How about using artificial intelligence to teach music?
Mancini was skeptical.
"As soon as I heard AI, I had this panic," she said. "All I thought about were these crazy robots that can think for themselves."
There were no robots. Just a web application that uses AI to build original instrumental tracks from a library of prerecorded samples after a user selects a few parameters.
Equipped with Chromebooks, Mancini's students could program mood and genre, manipulate the tempo or key, mute sections, and switch instrument kits with a couple of clicks. And just like that, an original piece is produced instantly.
The AI program-designed for use by anyone who needs cheap background tunes for media content-enabled Mancini to teach in ways not possible before: Students in an elective course who do not play instruments or read sheet music were now creating their own compositions. For the musically inclined students, Mancini said the software allowed for an even deeper fusion of computer and human-they'd create a track and play over it, combining AI-generated rhythms with live instrumentation.
"For me, music is an emotional experience. I know what I put into my playing and teaching of music. For that emotion to come out of an algorithm, I couldn't wrap my head around it at first. How can a computer replicate that?" she said. "But it can. I'm a convert."
While Montour is embracing AI technology with a full-blown bear hug, most school districts are not-at least not yet. Some are dabbling with applications. Others aren't using AI at all.
And still other educators can't say if their districts are using AI, oftentimes because they're not familiar enough with the technology to recognize it.
Whether that changes with the nationwide distance learning experiment that happened this spring is still to be seen.
This much, however, is clear: School budgets are going to be devastated from the economic onslaught wrought by the virus, and strapped-for-cash districts could delay tech acquisitions other than the devices and hotspots students need to go online as they prioritize necessities. Still lingering are serious questions about privacy, data bias, and just how effective AI solutions are for education.
The 3,000-student Montour district, in the suburbs of Pittsburgh, is using AI inside and outside the classroom.
The district teaches courses focused on artificial intelligence, ranging from ethics to robotics. It partners with universities and technology companies working on the cutting edge of AI. There's even a 4-foot tall autonomous robot, a boxy machine that looks like a filing cabinet on wheels, zooming around the hallways of its elementary school delivering packages.
And on the district's backend IT infrastructure, there are dashboards and programs powered by AI providing educators with real-time data about each student, producing metrics that monitor progress and even forecast future success.
"When we come back to school next year after the coronavirus, we're going to have data on every single kid from their remote learning experience," said Justin Aglio, the director of academic achievement and district innovation at Montour. "Not your traditional A,B,C data, either."
Districts, already inundated with trying to keep up, might also shy away from AI tools in the immediate future while teachers and staff adjust to a new digital ecosystem already pushing the boundaries for many.
"It's not even on our radar right now," said Andrew McDaniel, the principal of Southwood High School in central Indiana, when asked if he's considering incorporating some of the most basic forms of AI, such as Alexa voice devices, into classrooms. "A lot of teachers are looking at what they know works now and sticking to that. They're not going to mess around with much that goes beyond that."
Increasingly, though, voice-activated devices such as Alexa, Siri, and Google Home are being used as teaching assistants in classes. Schools are turning to smart thermostats to save money on energy costs and using AI programs to monitor their computer networks. AI is helping districts identify students who are at risk of dropping out, and math tutors and automated essay-scoring systems that have been used for decades now feature more sophisticated AI software than they did in the past.
Until recently, though, most of those tools have relied on simpler AI algorithms that work on a basis of preset rules and conditions.
But a new age of AI-based ed-tech tools are emerging using machine-learning techniques to discover patterns and identify relationships that are not part of their original programming. These systems consistently learn from data collected every time they're in use and more truly mirror human intelligence.
Ed-tech vendors are pitching advanced statistical AI tools as a way to provide greater personalized learning, tailoring curriculum to a student's strengths and weaknesses. Researchers say it is unlikely advanced AI will transform K-12 education, but it can have a positive impact in areas like adaptive instruction, automated essay scoring and feedback, language learning, and online curriculum-recommendation engines.
Most of the startups pioneering education solutions with this type of AI aren't yet in a position to offer their products on a mass scale in the United States. That's because highly accurate advanced AI systems require access to massive data sets to populate and train the machine-learning algorithm to make reliable predictions. Those algorithms must also have access to high-quality data to avoid reinforcing racial, gender, and other biases.
Bill Salak, the chief technology officer for Brainly, an AI-based content generator and homework assistant that uses machine learning, said his company has traditionally worked directly with students, not districts. Now, however, Brainly is diving into more advanced statistical models for its AI to allow for even deeper personalization, and it is planning to eventually start creating products that could go into the classroom.
Salak said that all AI-based technology vendors face an uphill climb because school districts are consistently underfunded, and if they're going to spend money on a tech tool, it "has to be proven to be effective and contributing to academic goals."
"The education systems prioritize things that will help them meet their goals, and not many outcomes relate to teaching with new tech," he said. "Even if the teacher may see a huge amount of value in something, at the end of the day, that teacher has to have a certain percentage of their kids meeting certain competency standards."
April DeGennaro, a teacher in the gifted program at Peeples Elementary in Fayetteville, Ga., knows firsthand what it's like for district administrators to buy into the idea of using AI-tech tools but not backing up that commitment with funding.
DeGennaro runs a lab where students focus on robotics, and her 4th graders use an AI-based robot called Cozmo. Shaped like a mini bulldozer that can fit in your palm, Cozmo uses facial recognition and a so-called "emotion engine," allowing it to react to different situations with a humanlike personality by showing a range of emotions, from happy or sad to bored and grumpy. Because of COVID-19-related school closures, the AI robots currently aren't being used.
But under normal circumstances, up to four students can use one of the robots at a time with an iPad, coding it to carry out different tasks. At $150 each, DeGennaro said the robots amount to a low-cost investment, but she's had to find her own funding for all seven Cozmo robots in her class.
DeGennaro raised money online, where she got parents to chip in to buy robots. She's also made it clear to those that know her: "For Christmas, for an end-of-the-year gift, or whenever you want to buy Mrs. D a present, buy a robot."
"School districts may like things," DeGennaro said, "but that doesn't mean they're going to fund them."
At the Saddle Mountain Unified School District in Arizona, a new policy allowing high school teachers to use Alexa or Google Home went into effect this year after a group of district officials and teachers walked through several STEM schools in the Phoenix area and saw the devices being used in classrooms.
Joel Wisser, the technology integration specialist for the 2,300-student district, said "teachers walked away impressed," and several decided to incorporate the devices into their daily classroom activities. The district didn't pay for the devices, however. Instead, teachers had to bring their own, and Wisser said he doesn't expect that to change.
One history teacher uses his Alexa as a mini-assistant: reminding him when to return papers to students, answering student and teacher inquiries, providing a Jeopardy-style quiz game, or even playing music set from a time period the class is studying to add ambience to a lesson.
"It's really just a personal assistant, a helper, for him. His eyesight is not great. He has a 46-inch computer monitor and he's not a fast typer," said Wisser. "Being able to talk to a device is much more efficient for him, so he's not spending time at a keyboard typing in the words 'ancient Greek music.' "
Everyone didn't welcome the devices at first. The district's technology director, for one, was "hesitant" because the Alexa was going to be tapped into the district's network, and he wasn't going to have complete control over it, Wisser said.
The voice-activated speakers are also at the center of an ongoing privacy debate since they can record conversations. Wisser said there hadn't been any pushback from parents so far, and class conversations were not recorded.
Christina Gardner-McCune, the director of the Engaging Learning Labs at the University of Florida, said parents, students, and teachers have concerns about what kind of data an Alexa device is collecting in the classroom and "what is it doing on the district's network while there." Even though the recording function on an Alexa can be turned off, Gardner-McCune said some districts don't want anything to do with them.
"A lot of districts are not allowing those devices in the classroom even though they could have some educational purposes," said Gardner-McCune, who is also a steering committee co-chair of the AI for K-12 Initiative, a national working group of teachers and AI experts focused on jump-starting discussion on how to incorporate AI learning into school curricula.
It will take more time and use of AI devices and tech tools in classrooms before districts become increasingly comfortable with them on a larger scale, she said. And more research is needed showing the benefits of advanced AI systems before districts are willing to pony up for them: "For major school districts," said Gardner-McCune, "it's going to come down to how does it affect test scores."
Back in the Montour district, band director and teacher Mancini said her apprehension about the AI music program vanished when she became familiar with the web application and realized "there wasn't going to be a robot in the middle of my room." One of her favorite class exercises using the AI music program involved muting the background music on a movie clip-like the scene where the ship is sinking in "Titanic"-and letting students rework the general vibe by adding their own music.
"Music education has been so traditionally taught one way. We play instruments or sing or learn music theory. This is so far from traditional, and I'm glad I did it because it was so much fun when I got into it," she said. "As teachers, we just need to not be afraid of technology."