StepsAI: Step-by-step Generative AI Technology for Mathematics Education

Teaching of mathematics requires imparting students with basic concepts and providing them with questions and solutions to practice on. Similarly, students learn by understanding the concepts and practicing extensively on questions. This teaching and learning approach is common to quantitative subjects across the science, engineering, and business fields (eg. physics, economics). Traditionally, students and teachers have derived questions and solutions from textbooks and other printed and online resources. The advance of generative AI chatbots, particularly in the past half year, has brought advantages and challenges to mathematics education. Leading chatbots, such as Gemini Pro 2.5 and ChatGPT 4o, can solve many secondary school and university level math questions in less than a minute, while allowing easy-to-use input options and providing detailed solutions. For motivated math students, chatbots are like a 24/7 tutor able to provide personalized questions and detailed solutions, and explain mistakes. However, for less motivated students, chatbots are an advanced “calculator” for completing assignments with little thinking or learning. Prof Condon LAU and colleagues recently developed StepsAI technology to exploit the advantages of generative AI for math education while addressing the challenges.StepsAI is currently implemented as an iOS app designed for secondary school math questions and is used by teachers and students at multiple schools in Hong Kong. StepsAI takes a math question as input and passes it to a chatbot for processing. Through an innovative prompting strategy, only the key steps of the solution, no detailed numbers or equations, are outputted to the user. The user, usually a student, is required to attempt each step by providing the numbers and equations as needed. At each step, StepsAI will determine if the user’s input is correct or incorrect, and provide detailed explanation. After eventually completing each step correctly, the student will have completed the question in a way that required extensive thinking and learning. StepsAI leverages the advantages of generative AI by providing personalized questions and solutions and detailed assessment of user inputs. StepsAI also has convenient features such as being able to generate additional similar questions for practice and generating a graded report for teachers. StepsAI is currently used by weaker students in math remedial class, under the supervision of their teacher, for improving their abilities.In this project, StepsAI will be upgraded to be more suitable for university level math problems. University questions and solutions typically contain more special math symbols and complex figures. Thus, we will add math optical character recognition and direct image upload capabilities to StepsAI. StepsAI will be made into a website, with the assistance of CityUHK’s Computing Services Centre, to accommodate users of all operating systems. Further, we will integrate StepsAI into 1st year math and physics courses, such as MA1200, MA1201, MA1300, MA1301, PHY1201, PHY1202, and PHY1203. Students, especially those who encounter difficulty with homework assignments and midterm exams, will be invited to practice on similar questions using StepsAI. We will track their progress over subsequent assignments and exams to measure the effectiveness of StepsAI.