Learning Artificial Intelligence Through Youth-Led Conversational App Development Experiences

• In Event: Teaching Artificial Intelligence in K–12: Innovative Learning Designs in an Evolving Landscape

Abstract

OBJECTIVE
Spoken AI-powered conversational apps are increasingly ubiquitous and allow for capabilities such as controlling music or checking the weather (Van brummelen et al., 2021). Our team developed an age-appropriate interface for middle schoolers so they could create their own spoken conversational apps. The goal of our camp is for campers to learn CS and AI concepts by actively creating new conversational apps through authentic inquiry. Research questions investigate the extent to which campers achieve significant knowledge gains, increase interest (Hidi & Reninger, 2016) in AI, and a progression of user-centered design practices when developing their conversational agents.

CURRICULUM AND INTERFACE DEVELOPMENT
Curriculum development focused on the five AI4K12 (ai4k12.org) big ideas and progressions to create learning objectives and accompanying lesson plans using culturally responsive pedagogy (CRP; Gay, 2002) and Universal Design for Learning (UDL; Rose & Meyer, 2002) to ensure that students engage in personally relevant, authentic instruction. Additionally, we created AMBY(AI Made By You), a youth-centered interface (see Figure 3.1) that leverages Dialogflow, a natural language processing platform used to build conversational applications. There was careful alignment between AMBY and the lesson plans. For example, lesson plans focused on broad AI experiences, a connection between AI, conversational AI, and use of AMBY. These lessons embedded the Use, Modify, Create approach (Lee et al., 2011) to scaffold students' creation of chatbots within the AMBY interface.

METHODS & DATA
Prior to camp, camp facilitators engaged in professional development focused on the curriculum, AMBY, and developmental, cognitive, and pedagogical needs of the campers. Constructs measured include campers’ interest formation, ability beliefs, and content knowledge. In Summer 2021, 14 students participated in the summer camp (2 girls and 12 boys, 11 Black, 3 White students). Session 1 of summer 2022 included 19 students, but only 13 completed both the pretest and posttest, including 6 girls and 7 boys (1 Asian, 9 Black, 2 Latinx, 1 Native American, and 2 White). Two students identified as more than one race/ethnicity (Black/Asian and Black/Native American/Alaskan Native).

FINDINGS
Summer 2021 data revealed camp facilitators promoted learning, engagement, and development of the students’ conversational apps. Although students created chatbots using Dialogflow, the interface was challenging for middle schoolers. Thus, the AMBY interface was developed with clearly articulated goals, including speech recognition chief among them to allow campers to focus more efficiently on project goals. In Summer 2022, students created a range of conversational apps including ShopBot, which suggests clothing brands and JerryBerry, which teaches facts about Black history and Black influencers. On pre/post assessments, significant findings were found in ability beliefs (p < .001), identity formation (p = .012), and intentions to persist in computing endeavors (p = .03).

SIGNIFICANCE/CONCLUSION
Beyond its impact on these participants, the project contributes research findings and best practices for engaging middle school students in learning experiences around AI and spoken conversational technologies. Meaningful engagement offered by personally relevant projects has the potential to transform how middle schoolers view STEM and computing careers.

Authors

• Maya Israel, University of Florida
  Presenting Author

• Joanne Barrett, University of Florida
  Non-Presenting Author

• Tom McKlin, The Findings Group, LLC
  Non-Presenting Author

• Yukyeong Song, University of Florida
  Non-Presenting Author

• Xiaoyi Tian, University of Florida
  Non-Presenting Author