Christina Nuñez, an education technology resource teacher who works with new teachers in the Albuquerque public schools in New Mexico, can’t provide one-on-one help to all of the 300-plus newbies who join the district every year.
So, she found a way to offer what might be the next best thing: A chatbot that can answer many questions for new teachers and their mentors about topics like classroom management, lesson planning, and differentiating instruction.
Nuñez, who created the bot as part of a program for innovative Ķvlog at Google, started by surveying new teachers on their greatest needs. An AI tool—ChipAI—helped her group their responses into seven big categories: curriculum, resources, differentiation, time, management, engagement, and relationships.
Nuñez wanted to offer new teachers and their mentors information that would go beyond what they’d get from a basic internet search. In addition to guidance the district created for all mentor teachers, she directed the bot to draw from resources specific to Albuquerque, including the district’s lesson plan template, curricular materials, and policies.
Developing the bot wasn’t as easy as Nuñez expected. “I thought I could build the bot in one weekend with a good prompt and a pile of documents,” Nuñez wrote in a visual presentation of her work for ISTE+ASCD’s annual conference in San Antonio, Texas, this summer. “Spoiler alert: I couldn’t.”
The AI “hallucinated wildly, misunderstood everything,” she wrote. “I had the data. I had the dream. But no idea what I was doing. The more I asked AI to help, the more it led me in circles.”
Nuñez’s program at Google paired her with a technology mentor, who helped fix the bot’s trouble spots.
Finally, last winter, after months of work, the bot went live to 400 teachers, mostly in her school district, where it is now responding to a range of teachers’ queries.
Nuñez shared this example of a teacher interacting with the bot:
Teacher: “Hey, I have a group of boys who continue to get up during class and when I tell them to sit down, they just laugh.”
Chatbot: “Dealing with students who are not following instructions can be challenging but there are some strategies you can use to address this behavior effectively.”
The bot went on to suggest that the teacher establish clear expectations, use positive reinforcement, and implement restorative practices, with additional information about all those methods.
Chatbot technology is evolving fast and making its way into K-12 schools
Chatbots like Albuquerque’s need to be developed carefully, said Torrey Trust, a professor of learning technology at the University of Massachusetts Amherst.
“There’s really high chance that they make stuff up or have biased results, but if you put the work in behind the scenes to address some of that, I think they can be helpful,” she said.
The chatbot can’t replace a real-life mentor, Nuñez emphasized. But teachers and mentors appear to find it useful. Albuquerque’s bot has been used nearly 5,000 times since its debut last winter, Nuñez said.
She’s also just rolled out a second bot to help field any teacher’s questions about curriculum materials. For instance, teachers could share classroom activities they’ve created on their own or materials they got off a site like Teachers Pay Teachers and ask the bot to gauge whether they meet Albuquerque’s quality standards.
Now, Nuñez is working on a bot for instructional coaches. She’s uploaded some of her favorite coaching resources—including The Art of Coaching, a book about coaching teachers by Elena Aguilar—and is training the bot to offer coaches feedback on their strategies or offer ideas for dealing with difficult situations, such as a teacher who doesn’t take criticism well.
Changes in popular large language models—like the option to create custom GPT’s in ChatGPT or “gems” in Google’s Gemini—have made bots easier to develop now than a year ago, when Nuñez created the tool aimed at new teachers and their mentors, she said.