Week 4

Teaching AI in K-12 Learners: Lessons, Issues, and Guidance

I found the wide breadth of pedagogical approaches and curricula that have been designed and implemented for AI education for K-12 students very interesting. These methods cover a wide array of ideas, from using simple games, such as in the A is for artificial intelligence paper where the children focused on rules-based learning with rock-paper-scissors game, to complete machine learning applications in other school subjects.

With so much accomplished in a small time frame implies that literature or best-practice reviews become obsolete shortly after publication. The author mentions that the AI4K12 framework has no mention of large language models (LLMs), prior to the emergence of ChatGPT in December 2022. Since their release, LLMs have become the most popular and contentious topic of AI signifying that the field is subject to change at any given moment. 

The challenges for advancing AI education in K12 were particularly insightful, especially regarding the role of other subjects in AI education. Since CS classes are already overloaded with various topics and new material coming out all the time, AI should not solely be the responsibility of CS classes to introduce to student. Even if other subject areas cannot explain the underlying concepts as thoroughly as a computer science class could, there are numerous other ways for them to contribute. They can discuss the implementation of AI in their respective subjects, or how AI is effecting their area of expertise, and the potential biases AI may exhibit in their respective fields.

Exploring children’s attitudes toward an age-tailored AI-powered chatbot

Vahedian Movahed et al. assessed student learning outcomes quantitatively and qualitatively. Seemingly all of the interactions that the children had with the chatbot were recorded for later data processing. With this foresight, the authors were able to derive quantitative information such as the engagement of a student and the engagement of each topic. The average utilization was between one and three questions, with some students asking fifteen-plus questions.

The qualitative assessment consisted of reviewing conversations with the chatbot as well as survey questionnaires and guided interviews.

The quantitative information can further guide development of ask-me-anything (AMA) for better usability and scope. The qualitative information gives a deeper insight to how the children, even by grade level, are understanding or personifying Artificial intelligence and understanding of key themes for children learning. One of the final questions that was asked included a question asking how the children would describe the chatbot to their friends. Through this question the authors were able to gauge the children's understanding of various aspects of artificial intelligence along with their uses. The results ranged from fun and knowledgeable to understanding that the model has limitations.

A concerning factor that was brought up was the children's willingness to trust and confide in the AI as if it were a friend, the authors bring up the need for more rigorous digital privacy information to be brought into the curricula.