SPS Distance Learning Blog

Disclaimer: This post discusses use cases of generative AI and is not a Distance Learning policy or endorsement of generative AI nor specific products or services. 

Providing An Easier Internet Search Experience

My role as a Senior Interaction Designer within the Distance Learning Department allows me to wear my creative hat while I partner up with other learning designers to create or revise online courses. Creativity has many tasks in the online learning world that are considered mundane, like converting documents and creating style guides. Our department is always looking for new ways to incorporate Generative AI into our work and I found a number of ways to decrease my mundane task list while leaving more time for consulting.

I remember back in 2003, a new search engine emerged that would put Yahoo Search and AskJeeves out of business. Suddenly, information access became easier than searching through page directories. Recently, my experience with popular search engines has reached a critical point where the results can no longer provide me with the accurate answers I seek. In fact, nearly 30% of people redo or refine Google Searches. Nearly every website has been SEO optimized (for good reason) to get their website on the first page of search results. Aside from sponsor links that take up much of the screen on the first search results page, you’re unlikely to find much relevant content unless you add more specific keywords to your search. 

So, I’ve just asked CoPilot, a new Generative AI platform from Microsoft and provided freely to NU students, staff and faculty, with my daily questions and answers and I’m not disappointed with the results. In fact, I find the tool quite helpful in my daily work. While CoPilot doesn’t offer image analysis, ChatGPT does, which I’ll show later in this post.  

I’ll be referring to CoPilot and ChatGPT as tools, because there’s not enough evidence that these AI tools can accurately produce results without human intervention. There are good reasons not to trust these tools as “digital assistants” or co-partner on a project. 

Parsing through obscure descriptions

It’s hard to describe something visual that you don’t know the correct terminology for. During course development check-in meetings, we try to describe complex charts with terms like “arrow thingy” with the “circle thing”. We do a quick image search and get the results we want on how to visualize data to students. Take this image, for example:  

How would you describe this image? If you needed to describe this image, what would you call it? Some sort of conversation from “two different measurements”? This is a question one might ask, if they lack the necessary mathematical or cartography skills. I have input this graphic into ChatGPT, and this was the response:

This is a black-and-white diagram featuring two spheres, one smaller and one larger, sharing the same vertical axis. The following details are present:

1. Larger Sphere:
   1. Appears at the bottom and is represented by a full circle.
   2. The sphere’s surface is defined by a grid of solid lines (longitudinal lines) and dashed lines (latitude lines), creating a three-dimensional perspective.
   3. The southern hemisphere is shown more prominently, with concentric dashed curves meeting at the bottommost pole.
2. Smaller Sphere:
   1. Located at the top and tangent to the larger sphere at their equator.
   2. Has a similar grid pattern, with solid longitudinal lines converging at the topmost point.
   3. Represents a projection or a smaller version of the larger sphere.
3. Horizontal Line:
   1. A straight horizontal line runs through the center, serving as the equatorial plane for both spheres.

This image could represent geometric, mathematical, or cartographic concepts such as stereographic projection, symmetry, or nested spherical coordinates.

This tool can be useful for describing hard to articulate concepts such as complex graphics to faculty or clients. I could also discover connections or references that I might not have otherwise noticed. Also, this description is a good base for writing alt text for screen readers and those who require adaptive technology.

Tasking Help

Dark shades, Bright tints

The other day, I was assigned a task to create a monochromatic palette for this color, #ef553f. It’s a beautiful rusty red color, and I needed both dark shades and bright tints at 10% increments of intensity. Also, I need this data in a spreadsheet. What a task! I’ve asked CoPilot this question: Can you come with the monochromatic scale of this #ef553f color at the intervals of 10% until it reaches 100% black and 100% white. Put those results in a single column for Excel? 

This is what Copilot generated:

To confirm the results, I requested the hex codes to be converted into swatches, and this is my result:

A copy and pasting task became a whole lot easier by doing many of these types of automated tasks with the help of AI tools.

GenAI and Accessible Content; or, I can’t believe someone took a screenshot of this

Looking through course content, you’ll see one of these:

It’s a screenshot with text. In this case, it’s  Northwestern’s June 2025 schedule for undergrad students. Well, screen readers can’t read that for visually impaired students, but ChatGPT can. In my case, I’ve uploaded the image through my account and received an excellent word-for-word copy of the screenshot’s text. This could be applied to textbook scans, old articles compressed as an image file, etc.

Conclusion

As I dive deep into the world of AI, I’ve become more aware of its many limitations. While I showcased a number of instances where I used AI to aid me in automated tasks, I have yet to find an excellent case study for my course development work or media production to use any of these AI tools. While Northwestern established a Data Classification Policy and has approved CoPilot for Level 2 and generally Level 3 data (get more information on level 2 and level 3 data here), other AI tools such as ChatGPT don’t have a strict privacy policy for their users. All of my work when using any type of AI tool is vetted and reviewed by the faculty developer and my teammates in Distance Learning. Despite the pitfalls, my automative tasks list has shortened with the introduction of these tools. 

To learn more about potential uses of generative AI, check out the resources offered by the Searle Center, and/or book a DL Consultation.