AI tools are being used more and more for teaching and learning. Before implementing AI tools, it is important to think about the accessibility of these tools, as well as how they can be harnessed for improving the accessibility of course materials and activities.
AI tools are continuously being developed, updated and improved. This article includes information which may have changed since its original publication.
In this guide you will find information on the following topics:
- How Accessible are AI Tools?
- Can AI Be Used as a Replacement for the Tool Ally on Canvas?
- AI for Accessible Word Documents and PDFs
- AI for Text-to-Speech
- AI for Video Accessibility
- Alt-text for Images with AI
- Multilingual Support: Translation with AI
- Additional Resources and References
How Accessible are AI Tools?
While AI tools can be used to enhance accessibility, many of the tool interfaces do not meet accessibility standards for use with assistive technologies. Commonly used generative AI tools such as DALL-E and ChatGPT have some barriers to access for those who use assistive technologies. Students may require more time or instruction for access, or may only be able to use some of the platform’s full set of features.
If you would like to use an AI tool in your classroom, but are unsure of how accessible it is, contact LDDI for more information.
Helpful resources:
Can AI Be Used as a Replacement for the Tool Ally on Canvas?
Ally, a Canvas plugin which was used in Faculty of Education courses, is no longer centrally supported by UBC and will not be available for use on Canvas. Currently, there are no AI tools which provide holistic accessibility overviews of websites (although there are other tools which can).
The Canvas accessibility checker is a suitable alternative to Ally, although it lacks Ally’s broad accessibility checking functionality, guided assistance in improving course accessibility, and alternative format provision for students.
Canvas’ accessibility checker can be found during editing on any page, assignment, or discussion in the rich text editor, and will flag missing alt-text, colour contrast issues, unlabeled tables, incorrect heading levels and link issues. However, it will not provide an overview of the accessibility of a course, nor will it flag issues with uploaded documents which may be inaccessible, for example a PDF which has not been OCR scanned.
If you require holistic accessibility checking at a website level, the following tools can be used:
- WAVE by webaim (browser plugin available)
- Lighthouse (Google Developer Tools)
However, these tools may flag accessibility issues native to Canvas which cannot be adjusted at a user level.
AI for Accessible Word Documents and PDFs
Like websites, files such as Microsoft Word documents or PDFs can be accessible or in accessible. The process for making these documents accessible can be complicated, as it requires adding appropriate heading levels, alt-text for images, adequate colour contrast, correct reading order, descriptive links, and for PDFs, ocular character recognition (OCR).
- For word documents, Microsoft has built-in accessibility features powered by AI such as suggested alt-text for images, and other styling features such as title and headings.
- Adobe Acrobat (paid version) includes built-in accessibility features which can vastly improve the accessibility of PDF files, including auto-tagging, OCR scanning, accessibility scans and guides for improvement. Recently, Adobe has introduced Adobe Sensei AI, which can automatically tag PDFs to improve accessibility. Other services such as CommonLook Artificial Intelligence (AI) are also emerging with options for automating accessibility for PDFs. These tools are currently available as demo versions only.
Helpful Resources:
AI for Text-to-Speech
All modern systems (computers and smartphones) have built-in applications for text-to-speech. We recommend making use of these if you would like to have AI-generated audio from the text. In the links below, you will find instructions on how to activate text-to-speech functions in different operating systems:
- How to use ‘Speak’ function on a Mac
- How to use ‘Speak’ function on an iPhone
- How to enable text-to-speech on Microsoft and Windows
If you don’t find a built-in application on your device, you can use the following Google Chrome extensions for text-to-speech:
* Please note that Chrome extensions might store data outside of Canada and they are not FIPPA compliant.
AI for Video Accessibility
AI for Video Transcripts
Chat-based AI platforms have extended capabilities in the realm of video transcription, transforming spoken words in videos into written text. The AI transcription tools have the focus on making meetings accessible to the Deaf/Hard of Hearing. Two of the major features are a high-quality AI-generated audio transcript, as well as an integrated transcription editing platform that connects the recorded audio to words in the transcript.
- Otter.ai (behind a paywall) is an AI transcription tool with a focus on making meetings accessible to the Deaf/Hard of Hearing. While recording a transcript (or uploading an audio/video file) is generally accessible; there is only a visual display that shows the recording is in process. However, when editing a transcript there are significant cognitive overload issues when using a screen reader to edit a transcript.
- Rev (behind a paywall) can also be useful for individuals with hearing impairments or those who prefer written content over audio or video. To be noted, the accessibility of the platform depends on the user interface, and which dictates the ease with which users can navigate and interact with the service.
Helpful resources:
Captioning Videos Using AI
Machine-generated captioning is available through video editing and hosting platforms such as Kaltura, Zoom and Adobe Premiere Pro. These platforms produce captions efficiently, but at varying degrees of accuracy.
- Kaltura is UBC’s video hosting platform and can be used to create machine-generated captions which are typically generated within 30 minutes of the request. These captions tend to have varied accuracy and should be checked manually. Kaltura’s captioning service does not distinguish between speakers, which is an important component of accessible captioning.
- Zoom provides automatic transcription and closed captioning for live meetings in various languages. Automatic transcription can be enabled through user settings, and closed captions can be turned on at an individual user level for use across meetings. Captions and transcripts can be saved on meetings recorded to the cloud and can be downloaded as separate files from the meeting recording.
- Adobe Premiere Pro (available to faculty and staff through UBC Software Distribution) can be used to generate captions. These machine-generated captions can recognize different speakers and can be easily converted into a transcript and exported as an .srt file. The accuracy of these captions is better than other available options but should still be checked manually.
Instructional Video: How to Caption Videos Using Adobe Premiere Pro
Helpful Resources:
- Captions in Kaltura – LDDI
- Enabling or disabling audio transcription for cloud recordings – Zoom
- Enabling automated captions – Zoom
Alt-text for Images with AI
Chat-based AI platforms can be used to generate image descriptions, otherwise known as alternative text or alt-text, and captions for images. However, clear guidance on what must be included for alt-text should be in the prompt provided, such as character length or the focus of the description. Adherence to prompts varies between platforms, so it is recommended that any alt-text or captions generated by AI platforms are manually checked for accuracy.
- Microsoft products such as Word and Teams already include a function called Image Analysis which automatically provide alt-text and captioning suggestions when you add an image to your project.
- Copilot and Bing Chat allow for the upload of an image and can generate both captions and alt-text based on this image. It cannot revise alt-text or descriptions based on a provided image source or webpage. It also struggles to follow strict parameters such as character length.
- ChatGPT Plus (behind a paywall) includes access to the GPT Vision function, where you can upload an image and generate alt-text and captions. It can also revise alt-text and captions if you provide the image source or webpage for additional context. ChatGPT-4 is better at following prompt guidelines than GPT-3.5.
- Other built-for-purpose alt-text generating platforms are being released, such as alttext.ai (behind a paywall), which are used primarily for generating alt-text and captions. However, these platforms still require guidance and manual checking by humans.
Helpful Resources:
Multilingual Support: Translation with AI
Chat-based AI platforms can be used for translating text from one language to another. They can also assist with correcting errors in the use of a second language. They are helpful in providing context and nuance for translations, such as explaining idiomatic language rather than translating literally.
However, these platforms are not necessarily better than commonly used translation tools like Google Translate or Deep.l, or a human translator at long form translation. It may also have limited translating capacity for languages with fewer online resources than widely used languages such as English, Arabic or Spanish.
- Deep.l is a commonly used translation site that has released a AI Writing tool which can be used to help improve written text by L2 users. It can provide corrections as well as style suggestions for business, academic, technical or simple language registers.
- Grammarly (free, plugin available) may be a useful AI tool for L2 users to check their written language for errors, style and tone.
Helpful Resources: