Max. Classroom Capacity: New Year, New Tools

Dear readers, 

Happy new year! As we dive into 2025, I wanted to review two cool tech developments that may provide us with tools to help our teaching. 

Custom GPTs

The first tool I want to discuss is ChatGPT’s relatively new feature that allows users to create custom “MyGPTs.” This feature is available only to pay users at this point. Quick review: ChatGPT is a large language artificial intelligence (AI) model that can answer your questions by reviewing its massive training database and (in newer versions) by actively searching the web. Custom GPTs allow you to create a distinct knowledge base and instructions for the GPT, either restricting the GPT to access only the information you provide or allowing it to search the web as well. In other words, the AI will consider the data you upload to it instead of (or in addition to) what it can find on the web and can be instructed to treat that information in specific ways. You can even write “conversation starter” prompts that users can click on if they aren’t sure what questions to ask the GPT. 

Instructors may use this feature to create a compendium of their slides, lecture notes, readings, and other educational materials, and then query the GPT on the contents. I find that unclicking the “web search” option, restricting it to the materials uploaded, to be particularly valuable in providing more certainty that the answers it provides are based on these materials, as opposed to outside materials of unknown quality. This creates many exciting possibilities to help instructors in various ways, some of which I review below. 

Targeted Searches

I know that I’ve talked about ChatGPT a few times (1, 2, 3, 4) in this column, but I couldn’t remember whether I had discussed the MyGPT feature. Rather than wasting a lot of time reading through all of my old Max. Classroom Capacity (MCC) columns, I wasted a lot of time uploading them into a custom GPT! This allowed me to ask the GPT what I’ve written about custom GPTs in past MCC columns. The answer: nothing! 

As an instructor, you might use this search function to identify places where you discussed specific topics or examples across all of your lecture notes. If you have electronic resources that you provide to students (e.g., an e-textbook, readings), creating a custom GPT will allow you to search for places in which specific topics are covered, identify whether you have missed topics, and so on. For example, I wanted to know where the topic of humor in the classroom was covered in MCC. My custom GPT told me that humor was discussed in the October 2017 column in which Tori Howes, SIOP’s Distinguished Teaching Award winner for that year, talked about the importance of humor as a means of connecting to students and creating an enjoyable learning space. 

Summarize and Analyze

A custom GPT can also summarize the uploaded content, develop themes, and perform gap analyses. For example, did you know that over the last 10 years of MCC (my tenure as author or coauthor), the column has had five major themes (at least, according to the custom GPT): pedagogical strategies and classroom techniques; teaching psychological principles and ethics; educational technology; diversity, inclusion, and international perspectives; and interviews with leading educators? Such a summary of lecture notes or readings could help the instructor to structure the class into major sections or develop assessments of learning. 

The gap analysis function yielded interesting results. When asked which major topics in I-O psychology education have not been addressed and should be the topic for the next MCC column, the first idea it generated was “AI-Driven Personalization in I-O Education.” As it happens, I address this topic later on in this column! A similar analysis of course materials could identify topics relevant to the course that are not yet represented in the materials. 

Instructional Materials

Custom GPTs can be used to make slides and other materials. If you have been asked to teach a class with topics outside your expertise, this is useful. Teaching unfamiliar topics is a common experience, especially with instructors early in their careers who are low in the pecking order when it comes to choosing classes to teach. This can also happen when teaching new topics that do not yet appear in textbooks. You can create a custom GPT with source materials for the topic. For example, recently I was involved in a grant-funded project that involved making recommendations concerning homelessness prevention in the City of Los Angeles. As part of this project, I conducted a literature review and created a custom GPT with the articles I found as the source knowledge. This allowed me to ask targeted questions (e.g., “on what pages of which articles is poverty discussed as an antecedent to homelessness?”), and to generate summaries of commonly referenced antecedents, interventions, and populations. These summaries served as a starting point for a slide deck describing our findings. Important caveat: There is no substitute for reading the source materials! But a custom GPT can save you time by quickly generating for you a starting set of bullets or topics that you can refine through more targeted searches. 

Although ChatGPT is best at generating text, it can also generate images via its integrated Dall-E feature (note: there are many other services for AI image generation). I’m not sure how useful this is, but as an example, I asked my MCC GPT to create an image that best represented the MCC column. Here’s what it came up with: 

This image seems to have reached its max classroom capacity for themes, symbols, and random weirdness… 

I have used the Dall-E feature to generate thematic images for specific course units (e.g., a unit on teamwork + CSUN branding) that I then built into a PowerPoint template for my slide deck for that unit. I have asked ChatGPT to generate charts and figures based on data in its knowledge base, but the results have been completely unreliable. Do not trust the data depicted to accurately represent the contents of the knowledge base! 

Course Evaluations

Here’s a fun one: you can upload copies of student course evaluations into a custom GPT, with web search enabled. Then ask the GPT to identify themes in students’ feedback on your teaching and propose developmental activities for you. 

Exam Questions

Another possible use case is to create a custom GPT for your course materials (e.g., slides, readings) and ask the GPT to generate exam questions representing the materials. For this column I had a custom GPT generate multiple-choice format items based on materials I uploaded into a custom GPT. The items were not perfect, yet I found this to be a potentially useful way to quickly generate a set of items that serve as a starting point. Similarly, ChatGPT was able to generate short-answer and essay format questions with corresponding grading rubrics. I was more impressed with the results here, particularly with the detailed grading rubrics that were generated. However, overall, I would be very cautious about uncritically adopting any questions generated by ChatGPT. These are best used to develop a “rough draft.”

Personalized Tutor

Although the exam questions generated by custom GPTs may not live up to test development standards in I-O psychology (a very high bar!), they may be a fantastic learning tool for students! A custom GPT can generate practice exam questions for students based on the instructor’s materials, evaluate students’ answers, and provide feedback, among other possibilities (continue reading)! 

The catch here is that at the current moment, only paid subscribers to ChatGPT can generate or access custom GPTs. However, such AI chatbot tools may be a more common feature of textbooks in the future (some textbook publishers already include functionality like this). Whether through ChatGPT or some other service, I believe that this functionality will be available to instructors soon, so let’s think ahead to that day. 

Let me provide one illustrative example of what this could look like. I use the Capsim business simulation in one of my MBA classes. I created my own customized GPT on the simulation for students (and myself) using the simulation manual and training materials. My dismay at discovering that I couldn’t share this with my students was equal to my excitement when I found out that Capsim had developed a custom GPT accessible through their platform! When asked, the Capsim custom GPT very capably generated practice exam questions, evaluated my answers to these questions, and provided developmental feedback, all with specific details from its custom knowledge base of the simulation manual and training materials. 

Moreover, when asked to create a “gamified approach to help me study,” it created a set of “challenges” in which it generated hypothetical miniscenarios within the simulation, asked me to provide answers and make decisions within these scenarios, awarded me points, and gave me feedback on my answers. I encountered no examples where the info presented was inaccurate or misaligned with its source materials. 

Similarly, I asked the GPT to play a Jeopardy style game to help me study and it quickly generated five categories with questions of different point values. It asked me to choose a specific category and point value (e.g., “Customer buying criteria for 100”), and reminded me to answer in the form of a question. Unlike my late, great fellow Canadian, Alex Trebek, it corrected my mistakes with minimal condescension! How much more fun is this than rereading the textbook?!?! 

At the beginning of each semester do you get a million e-mails from students asking questions about your course, the answers to which all reside in your carefully crafted yet criminally underread syllabus? Wouldn’t it be great if students could instead ask their personalized AI tutor for these answers? Adding your course syllabus to the custom GPT knowledge base would provide students with the answers that they need, sparing you the deluge of emails (or would students still find it easier to e-mail you?). 

The use of personalized AI tutors may make studying more enjoyable, and potentially more effective, compared to traditional methods. 

AI Video Generation

AI-powered video creation services have proliferated recently, and I am excited to explore how they may be used in the classroom. I started exploring the use of various services to generate a video to replace a brief text-based case study from my undergraduate leadership class. I wanted to explore whether a video could be created that brought the case to life and created a more dynamic and engaging experience for students. 

I started with the VEED AI video tool, which is integrated into ChatGPT. This service, like many others, cobbles together stock footage scenes to accompany a user-provided script and creates realistic-sounding narration. The results were hilariously bad! The video transitioned from one disconnected stock footage scene to another, featuring different people in different contexts, including workers in offices, but also a doctor in a surgical mask, and a family watching a soccer game! The key characters in the case were not represented by the same individual in the video across scenes. The reactions described in the script did not correspond to the action in the video. In sum, it was a confusing mess! 

Vyond might provide the functionality best suited to the purpose of creating a video case exercise. Vyond creates animated videos, and therefore provides much greater flexibility in choosing characters, scenery, actions, and narration. I uploaded the text of the case, chose the style, and generated the video in just a few minutes. The resulting 16-second video was generated around only a single line of text from the case rather than the full text. Vyond provides many options to customize and change the video. With a large investment of time, I believe that Vyond could be used to generate an animated video that depicts all the events described in the case. However, it doesn’t seem worth the effort to me. You may feel differently, and if you explore this option further, I would love to hear from you about how it turned out. Also, you can probably think of other use cases for AI video generation for your classroom. I would love to hear about those too. 

As always, dear readers, please e-mail me your questions, concerns, and reactions! Loren.Naidoo@csun.edu