Having been featured in the Common Sense Media whitepaper, "Generative AI in K–12 Education: Challenges and Opportunities" (p. 31), we've been inspired to write about our approach to AI at Eedi.
The recent rise of AI has sparked both excitement and apprehension in the education sector. At Eedi, we understand these concerns and believe that AI should be used responsibly and ethically, especially when it comes to our children.
Eedi's approach to AI is guided by a set of core principles, reviewed annually:
These principles closely align with the themes highlighted in the Common Sense Media white paper. Common Sense caution against blindly trusting generative AI as a source of truth. It can sometimes generate inaccurate information, such as… distractors that are not actually representative of real students’ misconceptions! Our Head of Education, Craig Barton, recently wrote a wonderful blog post on this topic.
We’re doing work in this space too - read on for an intro to AnSearch, our new tool for generating and refining high-quality diagnostic questions.
At Eedi, we believe in harnessing AI to enhance, not replace, the crucial role of teachers.
Our "human-in-the-loop" approach ensures educators remain at the heart of the learning process, guiding AI at every step. This philosophy aligns with the concept of AI as "sociotechnical", where human judgment is woven into the fabric of design, development, and implementation.
Imagine Craig, an expert teacher, crafting math problems with the help of an AI assistant like ChatGPT, Copilot or Claude. The AI generates a set of questions, but Craig's experienced eye immediately spots room for improvement. He refines the prompt, adding insights about the specific concepts he aims to assess. While the AI produces diverse ideas, Craig applies his deep pedagogical knowledge to sculpt them into perfection. He fine-tunes wording, adjusts numbers for real-world relevance, and introduces clever distractors to catch common misconceptions.
This scenario exemplifies Eedi's vision for AI in education. We're bringing this idea to life through tools like AnSearch, which combines Large Language Models (LLMs) with our proprietary machine learning to assist teachers and content creators in creating high-quality diagnostic questions.
The teacher remains the maestro, orchestrating the AI to produce truly impactful educational content.
Our live tutoring interface offers another glimpse into the symbiotic human-in-the-loop relationship. We built a tool where the AI generates suggested prompts based on specific math problems and student responses, but human tutors decide how to use these prompts, adapting them to each student's needs.
Finally, our misconception tagging tool. It AI to analyse new diagnostic questions we’ve created, identifying semantically similar existing distractors and proposing associated misconception tags. The system prioritises established misconceptions to maintain consistency. Our expert content authors (who are all ex-maths teachers!) then review these AI-generated suggestions to ensure pedagogical soundness. When our content authors introduce new misconceptions, the AI applies a set of human-defined rules for validation. This creates an effective symbiosis: content authors refine AI suggestions, while AI applies author-created guidelines to new human input. The result is a continuously improving system that combines AI's analytical capabilities with human pedagogical expertise.
By intertwining AI's strengths with teachers' unparalleled insights, we're shaping a future where technology amplifies rather than automates teaching. Our goal isn't to replace teachers with AI, but to empower educators to be even more effective in their noble mission of guiding and inspiring young minds.
This approach ensures that as we embrace the power of AI, education remains profoundly human at its core. After all, nurturing curiosity, fostering critical thinking, and igniting a passion for learning demands more than just solving equations – it requires the intuition, empathy, and adaptability that only dedicated teachers can provide.
We understand the concerns around AI predicting student performance. We believe in thoroughly testing and linking predictions with actionable steps that empower students before integrating them into the student-facing platform. This careful approach prevents issues like confirmation bias and fosters a learning environment focused on growth and understanding.
We also recognise the need for robust data privacy measures, especially when it comes to children. Our platform adheres to strict privacy protocols, ensuring data security and responsible use. We believe that transparency and control over data are crucial for building trust and ensuring ethical AI implementation in education.
At Eedi, we are strong advocates for collaboration between industry and academia. We are proud to partner with leading universities like UMass, Vanderbilt, Texas, Stanford in the US and Cambridge University in the UK to conduct research and ensure our approach is grounded in scientific evidence. A big thank you goes to Andrew Lan, who has published countless papers on Eedi data (here, here and here). Andrew is continuously pushing the boundaries on finding better ways to craft high-quality diagnostic questions, linked to misconceptions.
We are particularly interested in deep research partnerships where we collaborate with academic or commercial researchers to validate their research using Eedi's data and prove it through experimentation, further bridging the gap between research and real-world impact. If you are a researcher and have a brilliant idea to boost maths learning gains or student engagement, come speak to us.
At Eedi, we're working to create a world where AI empowers maths teachers and adapts to each student's unique learning journey. We're committed to working alongside educators, researchers, and policymakers to develop tools that address students' genuine learning needs.
Our mission is to empower maths educators with the best tools and insights to help every student reach their full potential. By maintaining a human-centred approach, we're confident in building a future where AI supports, not supplants, the invaluable role of teachers.
