This is the second part of our study, Artificial Intelligence in Education: Can AI bring the full potential of personalized learning to education?
In this part, both we take a look at considerations when applying AI to education.
The realm of educational artificial intelligence (AI) applications is vast and evolving rapidly. To inform policy decisions and navigate the complexities of AI in education, it is crucial to understand the distinctive properties that shape these AI applications. This blog post aims to shed light on some key considerations outlined in a recent study.
Many AI applications in education focus on automating didactic activities, such as selecting teaching materials, reviewing tests, and providing constructive feedback. They focus on whether a student is knows facts, figures, and are they able to master a core academic skillset. However, education extends beyond mere knowledge transfer; it includes teaching social competencies, and social emotional learning. AI can play a role in facilitating collaborative activities, like group assignments, by, for example, selecting students based on individual strengths and weaknesses on a variety of skills and topics. AI can also play a role in creating an environment of exploration, encouragement, and motivation, which is typically isn't covered by applications that optimize purely for academic prowess.
Educational AI applications can either focus on teaching specific subjects or embrace a cross-curricular approach, considering a student's entire performance across various subjects. While primary education often involves cross-curricular teaching, higher education faces challenges in providing a holistic view of students' performance. AI can bridge this gap by monitoring performance and identifying trends across subjects.
Most education system traditionally determine subjects and difficulty levels by the year and level of the student. AI, through adaptive learning systems, can challenge this norm by allowing students to excel in specific subjects at different levels, ages, and paces of learning. This flexibility can potentially redefine the mastery process, allowing students to demonstrate proficiency in individual subjects at a pace that is personalized to each individual.
Educational AI applications exhibit varying levels of autonomy. Some AI tools provide insights into student performance for teachers to interpret and use at their discretion. In contrast, adaptive learning systems possess a higher degree of autonomy, determining when and how students study specific course materials. As technology advances, more autonomous systems may emerge, providing direct recommendations to students without teacher intervention. But we need to be careful that this autonomy doesn't lead to learning bias.
AI applications can support both traditional and new educational styles. From identifying absenteeism to optimizing timetables, AI aligns with current educational methods. Blended learning, combining traditional education with digital tools, takes a step further. AI, with its sensors, virtual reality, and gamification, can enhance student engagement and motivation. Additionally, AI holds the potential to improve distance learning by leveraging data collected from Massive Online Open Courses (MOOCs). For example, is absenteeism still a problem if the student is highly engaged in learning at home?
AI applications often employ advanced predictive models that are not always 100% accurate. A trade-off exists between the accuracy and interpretability of these models. While more complex algorithms may yield higher effectiveness, the ability to understand and interpret their decisions becomes challenging. Tools like Shapley Additive Explanation (SHAP) aim to provide transparency by revealing how changes in input affect model outputs. However, one of the major concerns, is that there are challenges in understanding how AI models work, especially in situations with low levels of transparency in the model itself.
There is a great opportunity in AI, that isn't available in traditional education. With AI, we can integrate data from various sources; This includes teaching content, pedagogies / teaching methodologies, individual assessments and results from global sources on a student, school, district, and country level. Standardization initiatives at the national level aim to create uniformity, but the AI field lacks such standardization, resulting in fragmentation. Modularity and portability are critical for ensuring interoperability among different AI systems in the educational landscape.
As educational AI continues to shape the future of learning, we need to start understanding these properties and what is actually going on under the hood of the AI engines. Policymakers, educators, and technologists must collaborate to navigate the complexities, ensuring the responsible and effective integration of AI into education. Embracing these key considerations will pave the way for a harmonious coexistence of technology and education, fostering an innovative and adaptive learning environment.
Link to the original paper (source)
