Scope

Research for this OER was conducted in the fall of 2023. As such, some tools and features were not available to us or were not appropriate to test (for example, Bard was not available in Canada; Gemini did not yet exist). This means that offerings and feature sets are representative of that time. As the landscape on this topic is rapidly evolving, we hope to update this OER periodically.

Large language models, as a category, are covered in depth. Specific tools were used for hands-on testing and validation (ChatGPT, Bing Copilot, AITutorPro/AITeachingAssistantPro ):

• ChatGPT, as the most widely known and category-leading chatbot;
• Bing chat/Copilot, as the most widely distributed free chatbot, the only one currently with RAG, and the one most likely to get wide higher-ed adoption in Ontario due to existing Microsoft IT contractual relationships; and
• AITutorPro/AITeachingAssistantPro (Contact North) as a representative chatbot designed for education in a Canadian setting.

Other chatbots are surveyed in broad terms, but not tested or treated extensively. Other forms of GenAI such as image recognition and generation will be discussed where they are helpful to explain or understand GenAI broadly. All of the LLMs used in our testing and as examples here are under rapid, intensive development. The field of GenAI is constantly evolving, in terms of capabilities of the models, new entrants, and extensibility through/connectivity with other tools. Because of this, we will where possible try to discuss in general terms, focusing on commonalities and characteristics of the technology that are likely to persist, and encouraging mental models and analogs of the technology geared to longevity.

This OER fills a gap in available, approachable ready-to-implement material. There are many articles about Generative AI and LLM-based tools in both the popular and academic press, but most concern only one or two aspects of the tools in depth or address a number of topics very generally. This OER endeavors to provide one-stop shopping for Canadian post-secondary STEM instructors who need to know the benefits and cautions of using LLM-based tools in their classes. The sections on bias, privacy and security, intellectual property and copyright, academic integrity, good pedagogical practices, and assessment design are important lenses through which to examine tools (and strategies to employ them) which are widely promoted but currently poorly understood.