ChatGPT & Accessibility
Use ChatGPT to Help You Do the Things You're Supposed to Do (But Probably Aren't Doing)
By Katie Healey, PhD, CPACC, AccessibleHigherEd.com
Note: Accessibility can't be automated or outsourced, as you alone know your students and learning outcomes. But technology can help.
AI generative technologies are always changing; I used GPT 3.5 for this page.
Plain Language
Why? Shorter sentences and paragraphs minimize visual clutter on the page and help readers digest content more efficiently. This is especially helpful for people with dyslexia or other learning disabilities. Avoiding or clarifying jargon supports English language learners.
How? Demonstrate this feature for students as an accessibility tool and consider adjusting your own wordy, jargon-heavy writing.
Note the differences between these two explanations of Michel Foucault's biopower.
The second prompt requests an explanation in plain language:
Alt Text
What? Alternative text—or alt text—describes visuals. Embedded in code, alt text is not visible; screen readers voice this image description aloud to blind and low-vision users. Folks with certain cognitive disabilities also benefit from screen readers.
All visuals on learning materials, slides, exams, Canvas sites, webpages, emails, social media, etc. should include alt text or longer image descriptions within the text (or should be marked as decorative if not conveying useful information).
How? Ask Google or ChatGPT. It's slightly different for each program, but you'll usually find an "alt text" option by right clicking or selecting image options. This page also shows how to add alt text to Word, Google Docs/Slides, and Canvas.
ChatGPT's image descriptions are often incorrect, and you should NOT rely on auto-generated descriptions. Your image descriptions should be tailored to your student learning outcomes.
However, it sometimes offers a helpful starting point for writing your own alt text.
Try prompting it to write alt text for an image by pasting the image URL.
You can specify keywords to include (like Foucault above).
Image from Wikipedia's entry on Panopticon.
You can specify the length of description.
Canvas encourages 120 characters or fewer; other programs have different limits.
Note that the original alt text was simply "R.M.S. Titanic," which does not describe the image.
Note the descriptive information contained within the image's URL below.
The alt text is fairly accurate but needs some revision:
And often, the alt text is entirely inaccurate & needs a complete rewrite:
Multimodal Learning
Why? Universal Design for Learning encourages instructors to provide multiple means of representation, engagement, action & expression. Offering multimodal learning opportunities honors learner variability.
How? Try prompting ChatGPT for ideas of activities, songs, podcasts, or films related to a course topic, and demonstrate this for students.
Activity idea
Film
Song
Podcast
Captions/Transcripts
We should always provide captions or transcripts for videos, podcasts, or other media.
How? Use programs like Otter to auto-generate transcripts. Then, copy and paste into ChatGPT and prompt it to clean up all spelling and grammar errors without deleting content.
You can then save as an SRT file and upload.
You'll have to copy and paste the transcript into ChatGPT rather than provide the video link.
I used Otter to capture a portion of A.J. Juliani's The Backwards Podcast episode with Jay McTighe, which does not have a transcript, and pasted it into ChatGPT (see left below).
I then prompted ChatGPT to "correct the spelling and punctuation errors without deleting anything" (see right below).
This broke up the text into paragraphs, improved punctuation for readability, recognized that "grant" was a name and capitalized it, and put quotation marks around statements.
Note that transcripts are trickier with multiple speakers. You'll still need to edit/review transcripts after ChatGPT assistance. Still, this can save you time!