Clippy Could Never: AI for Experiential Learning Curriculum Design

The way that we perceive virtual assistants has changed drastically over the last 15 years. Our expectations of their performance, ability, and accuracy have skyrocketed. What once was a funny, annoying voice playing over many of our screens and home speakers, telling us jokes, playing songs for us, and setting our many snooze alarms, has now developed into full blown thought partners, coders, and personal artists. While the ethics of using these machines has both environmental and social concerns, the human desire for progress has once again pushed the limits of techno-futuristic idealism– delivering to consumers what once was perceived as a reality only available to science fiction.  

From simple computer desktop assistants that poorly manage your calendar, to smart fridges that track what groceries are missing from your shelves, the functions of AI have been broadly adapted and applied across the spectrum of everyday appliances and usage. Now, large language learning models have been trained across the entirety of the internet, and other more focused private datasets to provide us assistants for a variety of tasks.  

In academia, we have seen AI deeply discussed, highly contested, and widely applied. Students use it for assignments, papers, and study help– while educators use it to create lesson plans, program technologies, and draft executive communications and summaries.  
In this exploration of usage, our Experiential Learning team at the University of Texas at Arlington Libraries has begun to utilize the applications of AI for curriculum development. When working with faculty, it can be difficult to emphasize the importance of making connections between learning outcomes, assignment structure, formative reflection, and creative activities. We use AI to help create lesson plans that elucidate these connections for faculty. By understanding the links between Kolb’s Experiential Learning Cycle, and presenting syllabus and class information to AI models (In our case, ChatGPT), we have created a smoother system of curriculum development that satisfies the desires of our faculty members.  

The ethics of AI usage has been hotly debated across the internet and throughout a variety of academic disciplines. The greatest concerns lean towards data privacy and usage, bias, and integrity of results and usage. However, some more eco-critically conscious users also lament the environmental impact that utilizing cloud-based models poses.  

The way that any AI model is trained can deeply affect a user’s experience, and potentially perpetuate biases or provide results that are inadvertently discriminatory. In many cases, when using larger public AI models like ChatGPT, it is impossible to know the full makeup of the training data.  
Environmental concerns surround the chip manufacturing, supply chain upkeep, water consumption, and carbon emissions necessary to run, cool, and maintain the large data centers that run these models.  

For these reasons, some users might consider picking specific AI systems that focus on creating more eco-positive methods of maintaining their models, or even running a localized model on one’s own personal computer. Some AI models available through different work platforms (Teams, Microsoft, etc.) also promise to not store/share/use uploaded data for generative training purposes. Using a locally run model can help to fight back against many of these concerns– creating much less electricity usage (just whatever it takes to power your computer), removing your important research or personal information from potentially being used in cloud training, and being much more in control of the training materials provided to the model.  

Before using AI for work related information, be sure to check with your university’s guidelines on ethical usage. 

Be careful when providing the AI with personal information or documentation from your colleagues. If the material isn’t yours, ask before you put it into the AI. If it's publicly available information on the internet, then usually it’s okay to link or upload that to the AI, as it would have already skimmed the internet with that information anyway.  

If the AI makes a mistake, correct it. This will help it learn. Building on previous prompts also helps the AI to understand how to better provide you with results. 

For Curriculum Development:  
Make sure that Learning Goals are always central to your planning. When feeding information to the AI, or editing results that have already come out, ask yourself: are the learning goals reflected in this assignment, reflection question, or project description?  

Create projects that have structured workflows. It’s always exciting to incorporate creative projects and fresh technologies into a course’s curriculum, but make sure that desired outputs can be realistically accomplished. Using structures like the MakerLiteracies can help narrow down and organize what competencies you want your learners to be engaging with.  

Know your material, and research or seek help when you don't know it. We work with many disciplines as curriculum developers, and sometimes that means engaging with material we might not be familiar with. If you need help identifying key artists, biological terminologies, or important authors or composers to create examples for your projects, AI can help identify those things. Or, ask your academic colleagues in the field.  

Overall: Know your knowledge and skills limits,  

Know how well the tools you are using the accomplish the task  

Identify if you need AI for something or if you can do it on your own 

Search Engines still work. Know when it’s better to just research.