Kyron Learning
By our AI Review Team
.
Last updated October 26, 2023
Kyron's AI-powered tutor helps fourth graders learn math through responsibly designed interactive video lessons
What is it?
Kyron Learning provides video lessons with expert human teachers and interactions powered by conversational AI. This means that instead of replacing a human teacher with a chatbot, the conversational AI enables students to engage with the human teachers in a more interactive and conversational way. Kyron has engaged with a group of expert teachers who pre-record a set of instructional videos that combine lesson plans, their knowledge of typical student responses, and explanations of how they address common mistakes and misunderstandings. In a Kyron Learning session, these pre-recorded videos of human teachers take the student through a math lesson, asking questions to gauge a student's understanding of a particular math topic in a simulated one-on-one interaction. Students can then respond by text or speech. Kyron uses conversational AI technology to interpret student responses in order to choose the next appropriate video response that will progress the student along their learning path.
Kyron Learning was released in April 2023 to pilot customers and schools. Kyron is initially focused on providing standards-aligned math curriculum support targeted at fourth grade students.
To use Kyron, students need a computer and a headset. Currently, Kyron is piloting in a subset of schools, and pricing information is not yet available.
How it works
Kyron Learning uses conversational AI, a type of artificial intelligence (AI) that can simulate human conversation. It is made possible by natural language processing (NLP), a field of AI that allows computers to understand and process human language. Common applications of conversational AI include chatbots, virtual assistants, and customer support. While generative AI chatbots also exist (e.g., ChatGPT, Bard), the distinction here is that conversational AI bots are not generating new and original content. Instead, they analyze natural language inputs and provide contextually relevant responses that have often been predetermined.
Kyron’s main input is children’s speech, and for this it uses a speech recognition model. Speech recognition, also referred to as speech-to-text and automatic speech recognition (ASR), is a type of artificial intelligence (AI) that recognizes spoken language and translates it into text. This text can be used on its own and is also readable by other systems that use text as an input. Speech recognition works by analyzing audio, breaking it down into individual sounds, digitizing those sounds into a computer-readable format, and using an algorithm to predict the most suitable words, which are then transcribed into text. Speech recognition is not the same thing as voice recognition, which is a biometric technology used to identify an individual's voice.
Kyron combines conversational AI with something called dialog modeling. This allows Kyron to have control over the interactions between what students say and the videos that Kyron shows to them in response. One way to think about this is that effective dialog modeling should mimic a really great human call center agent. Just as you want a human agent to have the right information to handle expected conversation scenarios, they also need to understand less expected questions and be able to respond with helpful information. In the case of Kyron Learning, this means that the team has worked to match the intent of what students say with the right prerecorded teacher video in response.
Highlights
- We applaud Kyron Learning for its clear teaching principles.
- The company's responsible AI practices, particularly when it comes to the choice of conversational AI over generative AI, and the focus on inclusion and diverse representation among the platform's teachers have contributed to the product's high rating.
- The company's Participatory Disclosures provided a strong basis for our evaluation against the Common Sense AI Principles.
- While many of the latest interactive AI tools rely on generative AI methods that can be difficult to fully control and trust, the content that students see on Kyron is entirely human-generated and determined, which essentially eliminates the possibility of outputting harmful content to the student.
- Kyron is conducting ongoing effectiveness studies that have shown positive results. According to Kyron, 75% of students give Kyron the highest rating, 93% of teachers would recommend it to other teachers, and very early measurements saw an 18% jump in the number of students who understood the content after the lesson vs. before the lesson.
- Kyron ensures consistency and equality by providing identical responses to students for the same questions or errors.
- The company's Participatory Disclosures provided a strong basis for our evaluation against the Common Sense AI Principles.
- Expert teachers are fully integrated in Kyron's end-to-end development.
- The tool's limited availability during its pilot phase and careful rollout facilitates safe, continuous monitoring and learning.
- Given the restriction of Kyron to specific, prerecorded math lessons, it is unlikely that learners will become addicted to the tool or encounter harmful content.
Harms and Ethical Risks
- You say "toh-MAY-toh," I say "tuh-MAH-toh." Even though Kyron Learning is currently available in English and Spanish only, people speak those languages in many different ways that are always changing. This variation is shaped by many different regional, social, and contextual elements. This is especially true for what is called dialectical variation, in which grammar, vocabulary, and sound choices lead to many different ways to say the same words. Children's speech can vary even more. They have shorter vocal tracts, inaccurate articulation, they make syntax errors, and their pronunciation and grammar is still developing. Should Kyron’s speech recognition models struggle to recognize the wide variety of how children speak English and Spanish, this could not only hamper accuracy and make the technology more reliable, it could also amplify unfair biases.
- Parents, caregivers, and teachers should closely monitor how well Kyron works for individual students to avoid unnecessary anxiety for students who may not be well served by the technology.
Limitations
- Keeping up with language variation is challenging. Even with a massive amount of speech data, it's a huge challenge to align any speech recognition model across the almost endless variation that exists in spoken language. Kyron uses Google's natural language processing (NLP) technology, which includes Google's extensive speech recognition capabilities. While this covers a lot of ground, it is unclear if this will work as well for children's speech as it does for adults. Kyron currently addresses this by working with a diverse set of schools across the U.S. This has helped the company to understand the different ways in which answers are expressed by children and update its models accordingly. As Kyron's user base grows, it may become more difficult for the product's speech recognition to keep up with the additional variations in language.
- While Kyron Learning hopes to expand to other subjects and grade levels, it is currently available only for fourth grade math.
- Kyron leverages the expertise of its teachers to anticipate the various ways that students might approach questions, both correctly and incorrectly. While the tool's performance in pilot testing provides insights into its effectiveness, the expert teachers' videos are unlikely to capture the full spectrum of potential student interactions.
Common Sense AI Principles Assessment
Our assessment of how well this product aligns with each AI Principle .
Editor's note: Common Sense Growth, Common Sense Media's affiliate, is a minority investor in Kyron Learning. Our ratings are written by independent experts and aren't influenced by our affiliates, developers, media partners, or funders.
Additional Resources
EdTech Ratings
Best games, apps and sites for personalized learning
Free Lessons
AI Literacy for Grades 6–12 | Lessons