Have you ever tried to put algebra or more complex math online and realized that something that looks perfectly clear on screen can be a mess for a screen reader? Or wondered what actually happens when a student using NVDA or JAWS reaches an equation in your course?
If so, this post is for you.
More and more STEM courses are offered online, and there are good reasons for that: they can reach a lot of students and, in some cases, do so at a lower cost (Chirikov et al., 2020; Xu & Xu, 2019). But moving a course online also exposes problems that are easy to miss when we design mainly for what we can see on a screen.
The social model of disability & the failures of modern media
Gravel et al. argued in 2015 that it is “our learning environments, first and foremost, that are disabled” (as cited in Nieminen & Pesonen, 2020, p. 5). That idea stuck with me. So, to get a better sense of what accessibility looks like in practice, I went back to Jung et al. (2022) and their work on how visualizations are described for people who are blind or have low vision.
What they found was not especially reassuring. Their sample included visualizations from major news organizations as well as academic collections, and many of the visualizations they examined did not include useful text alternatives.
It is only a snapshot, but it points to a bigger problem. Accessibility still gets treated far too often as something to fix at the end, after the chart, equation or learning activity has already been built.
Accessibility online
Math is a particularly good example of why this matters. An equation is not just a row of characters. Meaning comes from position and structure: what is above or below something else, what is inside a fraction, what is a subscript, what is grouped together, and so on.
A PDF can show all of that beautifully to a sighted reader and still be difficult for a screen reader. The issue is not whether the equation looks right. It is whether the file contains enough underlying structure for assistive technology to understand the relationships in the equation.
Microsoft Word’s modern Equation Editor does a better job than many people realize. Current versions of screen readers such as NVDA can read and navigate supported modern Word equations. That does not mean every combination works flawlessly, though. The version of Word, the equation format, the screen reader and the user’s settings can all make a difference, so testing still matters.
Word also has Read Aloud. It is useful, but it is a text-to-speech feature, not a full screen reader. I would not use it as proof that an equation is accessible. NVDA, JAWS and Windows Narrator provide much more of the functionality a learner using assistive technology would actually rely on.
JAWS is another widely used screen reader, but unlike NVDA it is commercial software. Some universities maintain institutional licenses, while others may not. Cost can therefore remain a practical barrier for some learners.
Ideally, then, a course should not depend on one specific piece of assistive technology behaving perfectly. The more robust approach is to build the math properly and test it in more than one environment.
Enter LaTeX — but with an important caveat
LaTeX is widely used for technical and scientific documents, especially when the math gets complicated. It is free, and once you get used to the syntax, it can be a very efficient way to write equations.
For example, an equation created in Word can be viewed in a linear form, and supported versions of Word can work with linear mathematical syntax such as UnicodeMath and LaTeX-style input.
Here is the important part, though: LaTeX itself is not what makes math accessible on the web. A browser does not simply look at LaTeX code and know how to present it to a screen reader. Something has to render that code into structured mathematical content, often MathML or another accessible representation.
That is why the workflow can vary quite a bit from one LMS to another. In Moodle, for example, one institution may use an equation editor while another may rely on LaTeX delimiters and a math filter. So the safest advice is not “paste LaTeX into HTML.” It is: find out how your own platform handles math, and use the method it actually supports.
When that structure is there, a compatible screen reader can do much more than read a flat string from left to right. In many cases, the learner can move through parts of the expression and hear the relationships between them.
There is a benefit for sighted learners too. Properly rendered equations usually hold up better when zoomed, enlarged or reflowed than equations saved as images.
So I would not reduce the recommendation to “use LaTeX.” The real goal is structured math that survives the trip from authoring tool to LMS to browser to screen reader.
There is still some work involved. Someone on the course team may need to learn the syntax, understand how the LMS renders it, and troubleshoot the occasional equation that does not behave as expected. And yes, knowing enough of the math to recognize when something has gone wrong is helpful.
The importance of text alternatives
Equations are only part of the problem. Charts, graphs, diagrams and other visualizations can carry just as much meaning, sometimes more. If the only useful information is inside the image, a learner using a screen reader may simply miss that part of the lesson.
Jung et al. (2022) examined what people who are blind or have low vision need from visualization descriptions. W3C guidance makes a similar distinction between short alt text and the fuller descriptions that complex images sometimes need.
For a complex visualization, I would think about a few things:
- Say what kind of visualization it is: line chart, bar chart, scatterplot, diagram, and so on.
- Give the main takeaway early. What should the learner notice?
- Include axes, variables, scales and units when they matter to the interpretation.
- Describe the important trends, comparisons, outliers or relationships instead of narrating every visual detail.
- Make the underlying values or data available when learners may need to inspect them.
- If the chart is complex, do not try to cram everything into the alt-text field. A longer description or an accessible data table on the page is usually more useful.
- Keep the language direct, and if you can, test how it actually sounds with assistive technology.
The point is not to describe every pixel. The point is to give someone access to the same useful information the visual is meant to communicate.
What the research tells us
Jung et al.’s study is useful, but it is also a snapshot in time. Websites change. Publishing systems change. Accessibility practices change. I would use the study as evidence that there was a real, documented problem rather than as a permanent verdict on any one publication.
The same goes for informal spot checks. Looking at a few images on a website can be interesting, but it does not tell us whether the whole site is accessible. If you include that kind of observation, it is better to call it exactly what it is: an informal check.
If you want to dig further into the topic, Jung et al. (2022) and the W3C Web Accessibility Initiative resources in the references are good places to start.
Designing accessible online STEM courses
The web was not built from the start with blind users in mind. We have much better standards and tools now, but that history still shows up in the way content is created. For course teams, the practical lesson is simple: accessibility works better when it is part of production from the beginning, not something added during a final QA pass.
For STEM courses, there is no single magic tool. Word’s Equation Editor can be part of the solution. So can MathML, an accessible math renderer, NVDA, JAWS and other assistive technologies. The right workflow depends on the platform. You also do not automatically need to turn every Word equation into LaTeX for NVDA; current NVDA versions can read supported modern Word equations directly. What matters is the final experience in the actual LMS, browser and screen-reader setup a learner will use.
Learning some LaTeX is still worth considering. It can make complicated equations much easier to author and maintain, especially in systems that already use it as an input format. I just would not describe LaTeX itself as the accessibility solution. It is one authoring tool in the chain.
Even with better tools, advanced STEM courses can be hard to review properly. One challenge is finding people who understand both the subject matter and the assistive technology well enough to catch subtle problems.
That can become circular: inaccessible courses reduce participation, and limited participation can make it harder to build a larger pool of experienced reviewers. The answer is not to assume that blind STEM experts barely exist. It is to involve disabled learners and professionals earlier, and more often, in design and testing.
There is another problem that is less technical: the way we talk about math. When an instructor reads an equation aloud, some information is often left unsaid because a sighted learner can look at the expression at the same time.
Take Hâ‚‚O. We normally say “H two O.” We do not stop to explain that the 2 is a subscript because, in most contexts, that would sound painfully obvious. But once formulas get more complicated, the difference between a subscript, superscript, fraction or grouped expression can completely change the meaning.
As the math gets harder, there is less room to rely on what the learner can infer visually. The spoken version or screen-reader output has to preserve enough of the structure to make the expression understandable.
More work is still needed to establish consistent practices for speaking, transcribing and presenting complex mathematical notation. This can help instructors and learning experience designers create more accessible online environments.
There is a catch, though. More detail is not always better. If an equation takes half a minute to read aloud, the beginning may be gone from working memory by the time the learner reaches the end. Being able to navigate through an equation in smaller pieces can be far more useful than hearing one extremely long description.
Transcripts run into the same problem. If every mathematical relationship is written out in prose, a long derivation can become almost unreadable. A transcript works better when it sits alongside properly structured mathematical notation rather than trying to replace it.
In other words, there is no single format that solves everything. Good accessible design usually means giving learners more than one way into the content: structured equations, clear narration, descriptions, data tables and, where possible, the ability to explore the math interactively.
***
I hope this article gives you a useful starting point. Accessible STEM content can be fiddly, and some of the details get technical quickly, but the goal is straightforward: a learner should not lose access to the meaning of the course simply because that meaning happens to be expressed visually.
Step-by-step checklist
- Use a structured equation tool or supported mathematical syntax instead of turning equations into images whenever possible.
- Find out how your LMS actually renders mathematical notation and whether it exposes accessible mathematical structure such as MathML.
- Test a representative sample of equations in the browsers and screen readers your institution supports.
- Use short text alternatives for straightforward images, and longer descriptions or data tables for complex visualizations.
- Do not rely on Read Aloud alone as an accessibility test; include testing with a full screen reader.
- For complex STEM content, involve people who use assistive technology and understand the subject matter whenever possible.
References
Chirikov, I., Semenova, T., Maloshnok, N., Bettinger, E., & Kizilcec, R. F. (2020). Online education platforms scale college STEM instruction with equivalent learning outcomes at lower cost. Science Advances, 6(15). https://doi.org/10.1126/sciadv.aay5324
Jung, C., Mehta, S., Kulkarni, A., Zhao, Y., & Kim, Y.-S. (2022). Communicating visualizations without visuals: Investigation of visualization alternative text for people with visual impairments. IEEE Transactions on Visualization and Computer Graphics, 28(1), 1095–1105. https://doi.org/10.1109/TVCG.2021.3114846
Nieminen, J. H., & Pesonen, H. V. (2020). Taking universal design back to its roots: Perspectives on accessibility and identity in undergraduate mathematics. Education Sciences, 10(1), 12. https://doi.org/10.3390/educsci10010012
W3C Web Accessibility Initiative (WAI). Images Tutorial and Complex Images guidance. https://www.w3.org/WAI/tutorials/images/
Xu, D., & Xu, Y. (2019). The promises and limits of online higher education: Understanding how distance education affects access, cost, and quality. American Enterprise Institute.