Separating the hype from reality: Insights and recommendations on how generative AI can enrich teaching and learning

We are at a pivotal moment in shaping the future of education. The potential for generative AI to improve teaching efficiency and enhance teacher effectiveness is immense.  It’s a potential that is in the process of being realized in small classrooms and large educational systems alike.

Teachers are arguably the most important school-based determinant of whether and how much students learn. Given this complex and important job, how can generative AI-driven technologies help teachers, and what policy measures should be prioritized to harness the potential of the technology?

How generative AI is improving teaching efficiency and enhancing teacher effectiveness

Though generative AI is not a silver bullet solution, it does have the potential to improve teaching efficiency and enhance teacher effectiveness. Specifically, since our last blogpost, AI is helping to:

• Simplify teachers’ administrative workload. Teachers’ administrative workload carries an opportunity cost that limits the time teachers have to focus on instruction. For instance, a study in the Republic of Korea found that teachers with greater administrative workloads are less likely to provide feedback on students’ assignments. Generative AI tools can help teachers save time in administrative and routine activities, helping to free up time so that they can focus more on their students. Namely, AI systems can help teachers track and process student attendance, which, in turn, has been shown to minimize teachers’ workload and enhance student learning.
• Enhance teachers’ instruction in the classroom. Generative AI has been shown to be an effective complement to teachers’ instruction, helping them implement evidence-based teaching practices that are often difficult to incorporate in practice due to time and effort constraints. Namely, generative AI can support teachers in creating explanations, examples, and practice problems and uncovering students misconceptions. It has also proven to be a useful lesson planning tool. For instance, in India, a generative AI digital assistant is being used to develop lesson plans – including activities, videos and quizzes – in minutes. Relatedly, it can help teachers implement their lesson plans and assess student knowledge by generating low-stakes quizzes and assessments.
• Complement teachers’ instruction outside the classroom through tutoring. With the uptick in climate-related disasters and conflict resulting in students missing more school, generative AI has the potential to extend learning time beyond what’s possible in a typical classroom setting. For instance, in Ghana, an AI conversational math tutor, accessible via WhatsApp, is being used to provide additional tutoring outside students typical math classes, yielding positive outcomes. While it’s not quite the same as a human tutor, there is also preliminary evidence that students enjoy the flexibility and tailored tasks that a generative AI tutor offers, as it allows them to study at their own pace and review information as needed.

Five recommendations for using generative AI to enrich teaching and learning

Although generative AI creates unique opportunities to enrich teaching and learning, policymakers may consider these five recommendations when designing and implementing interventions that utilize generative AI tools:

• Prioritize human cognition: One way to ensure that AI does not hinder human cognition is to focus on developing AI tools that complement and enhance human abilities. Among others, this means moving away from using AI as simply an “answer engine.” Instead, generative AI can be used to boost students’ skills and talents, which can help them learn more effectively.
• Empower teachers: Teachers are the key to the successful implementation of AI in classrooms. They need to be incentivized to learn about AI and provided with the time and support needed to experiment with AI tools in their classroom. Moreover, they also require explicit training, coaching, and ongoing support to effectively integrate AI into their instruction.
• Develop AI literacy: Not only teachers, but decision-makers, principals, and students should develop AI digital competencies that are rooted in competency frameworks. This knowledge includes understanding how to use AI responsibly and ethically. Knowing when to use, but also when not to use these tools is a good starting point.
• Develop ethical guidelines: These guidelines should be constantly assessed and revised given the quickly evolving nature of these technologies. At a minimum, the guidelines should address issues such as privacy, bias, and equity, not as problems to tackle, but as new opportunities for skills development.
• Design AI solutions for a variety of settings: To address the widening digital gap, AI solutions should be designed for students in all settings, regardless of their socioeconomic status or geographic location. Specifically, AI solutions that have been designed for environments with limited connectivity and in a way that is centered on inclusivity should be prioritized. Moreover, tools that are low-cost, easy to use, and do not require a high level of technical expertise are also essential.

Applying EdTech lessons to AI

In addition to these tips, there are EdTech lessons that cannot be ignored. Education systems need to transition into learning mode. Namely, policymakers can start by thinking small (while keeping in mind that small scale tests say little about the heterogeneous effects of different subgroups), partner with non-state actors (when implementations are supported by non-state actors, the results tend to be better), and acknowledge the big picture is not enough (focusing on the average effect can obscure the diverse impacts on students with different characteristic).

Generative AI has the potential to serve as a catalyst for positive changes in education; however, there exists a dual risk of not only missing out on this opportunity but also widening existing digital gaps if the appropriate measures are not taken into consideration.