Abstract:

AI competency is crucial for the new era of AI for college students, no matter what major they are in. Different majors, in industry, academia, or working scenarios, will require certain levels of AI competencies; however, we, educators from different disciplines, may not always be ready to teach our students, especially for the educators in humanities-related domains.

In this paper, we report our experience joining two senior-level classes — one in machine learning and one in English writing — to teach code-free AI literacy from different perspectives. Students completed a semester-long project on disaster communication, each team selecting a different subtopic.

The goal of this joint-class project is to help students build competencies in utilizing open-source large language models (LLMs) to build domain-specific chatbots, test their chatbots, and perform necessary tuning. By requiring students from different backgrounds to form interdisciplinary teams, students are expected to contribute knowledge from their respective courses and majors, providing a productive mix of backgrounds and expertise.

Over one spring semester, two instructors from Computer Science and English scheduled five joint-class lectures covering essential topics for building disaster communication chatbots. Topics included local deployment of LLMs, prompt engineering, retrieval-augmented generation (RAG), vectorization, word embedding, web crawling, data parsing, and chunking.

Following each joint-class lecture, student teams completed assignments to build their own chatbots on their chosen subtopics, such as alternative water sources during natural disasters and refrigerator/freezer food safety during power outages. Teams identified and crawled trustworthy content from state and local government agency websites, prepared the data into machine-readable formats, and trained and tuned their chatbots.

In this paper, we present:
(1) our teaching structure,
(2) the assessment framework used throughout this semester-long joint-class project, and
(3) our observations and lessons learned.
Our training materials and videos will also be made openly available to the research community.

Keywords:

AI literacy, Large language models (LLMs), Retrieval-augmented generation (RAG), Interdisciplinary education.