Introduction
Mathematical reasoning is one of the hardest things to measure at scale. On paper, students show their working naturally. They sketch, cross things out, try different approaches, and build toward an answer step by step. But when testing moves online, much of that process disappears.
Digital math assessment was supposed to change this. The promise was that capturing responses digitally would give us richer data about how students think, beyond whether they got the right answer. But in many systems, the tools available to students for entering math have actually narrowed what they can express, pushing programs toward simpler question formats that are easier to deliver but harder to learn from.
This article looks at why that gap persists, what it costs for assessment programs, and how assessment platforms and modern math input tools, such as MathType by Wiris, are helping close it
When the Interface Gets in the Way
Ask a student to solve a multi-step equation on paper, and they’ll pick up a pencil and start writing. Ask them to do the same thing on screen, and the experience changes depending on the platform—they might need to browse a symbol palette or type unfamiliar LaTeX syntax.
This kind of friction is easy to underestimate from a design perspective , but for students, it changes the task. Research from the Phoenix project in 2025 found that writing mathematical notation digitally requires effort that pulls cognitive resources away from conceptual understanding. In other words, instead of focusing on the math, they’re figuring out the tool. A 2023 study reinforced the real-world impact of this, finding that the mode of delivery itself—tablet versus paper—can measurably affect young children’s math test scores.
However, the problem isn’t that digital input is inherently worse than paper. It’s that many tools haven’t yet been optimized for how people naturally work through math. The interface sits between the student and the assessment, diverting attention from what is being tested.
What That Friction Does to Assessment Quality
When an input method is difficult to use, it doesn’t just affect the student experience—it can also affect the quality of the assessment itself. This is because the assessment starts measuring something it wasn’t designed to measure. In measurement terms, this is called construct-irrelevant variance—a concept established by educational researchers Haladyna and Downing in an influential 2004 paper. Put simply, the test reflects how well students can navigate the interface, not just how well they understand the math.
This matters at the program level. If some students score lower because they struggled with the equation editor rather than the math, those scores don’t mean what they’re supposed to. The assessment loses validity, and the data it produces becomes less useful for decision-making.
At scale, this kind of noise compounds. It can distort comparisons between schools and undermine confidence in national results. For programs that use assessment data to inform policy or allocate resources, the stakes are high.
So, for assessment leaders evaluating their current tools, the question is straightforward: Is the input method sufficiently transparent that it doesn’t affect results? If students need training on the tool before they can show what they know, that’s a sign the interface is introducing noise into the data.
The Multiple-Choice Trade-Off
One common way programs try to reduce that noise is by relying more heavily on multiple-choice and other fixed-response formats. It’s worth being direct about why so many digital math assessments lean on them: These formats work. They score reliably, scale well, and produce consistent data across large populations. For many programs, they’re the practical choice.
The trade-off shows up when those formats become the only way students can respond. Multiple-choice questions can show whether a student can recognize the correct answer or eliminate incorrect ones, but they reveal much less about how that student arrived there. For example, they don’t tell you whether a student can construct a solution, choose a strategy, or work through a multi-step problem. And when a math curriculum emphasizes reasoning and problem-solving, but the assessment only captures selection, there’s a mismatch between what you’re teaching and what you’re measuring.
This tension is illustrated by comparisons between the National Assessment of Educational Progress (NAEP) and the Program for International Student Assessment (PISA). PISA places greater weight on constructed-response and multi-step reasoning items, while NAEP relies more on multiple-choice. Both are rigorous assessment programs, but they measure different aspects of mathematical performance—in part due to differences in assessment format.
Downing’s 2002 work on construct underrepresentation described exactly this risk: When assessments rely too heavily on narrow item types, they fail to capture enough of the skills and knowledge they are intended to measure, limiting the validity of the conclusions drawn about student learning.
However, the point isn’t that multiple choice is inherently bad. Used alongside other item types, it can provide reliable and valuable evidence of student learning. The challenge is ensuring that the assessment as a whole captures the full range of knowledge and skills that the program intends to measure.
How Students Can Express Math More Naturally
Technology-enhanced items (TEIs) are one reason this picture is starting to change. Allowing students to interact with math visually and dynamically makes it easier to capture reasoning and process alongside final answers. And the shift is happening at the input level too.
For example, handwriting recognition has reached the point where students can write equations with a stylus or finger, and AI converts their input into clean, structured notation in real time. Visual equation editors let students build expressions by clicking or tapping, without needing to learn syntax. The cognitive load drops, and the focus moves back to the problem.
For younger students, this is a significant shift. Instead of learning how an equation editor works before they can take a test, they write math the way they would on paper. For older students working with advanced notation—integrals, matrices, chemical formulas—the same principle applies. The tool should support the complexity of the expression without forcing the student into a rigid workflow.
Mathype by Wiris helps address many of these challenges. Students can write mathematical expressions naturally without a mouse, touchpad, stylus or keyboard, with handwritten input automatically converted into structural digital notation. An intuitive visual editor removes the need to learn complex syntax, while support for more than 500 symbols enables everything from K-12 arithmetic to advanced mathematics, chemistry and STEM notation.
Integrated directly into TAO’s assessment environment, MathType extends the assessment workflow rather than introducing a separate tool or process. Students can enter complex mathematical expressions more naturally, while assessment teams can deliver, score, and analyse responses within the same platform. Together, TAO and Mathtype help reduce the gap between how students think through mathematical problems and how digital assessments capture that evidence.
Overall, though, the broader shift these tools represent matters more than any single product: When math input feels natural, the assessment gets closer to measuring actual mathematical thinking.
“Our goal with MathType is to allow students to write math as naturally on a screen as they would with paper and pencil; we eliminate the cognitive friction of the interface. This ensures that the student can focus on mathematical reasoning rather than on how to input it. By making it easy for students to show their step-by-step reasoning, MathType unlocks deeper insights, allowing for more comprehensive and actionable feedback”, explains Clara Abelló Gàllego, Chief Product Officer at Wiris
When Usability Shapes Who Can Be Assessed Fairly
Making math input feel more natural isn’t just about usability—it’s also a fairness issue. If a student can’t use the interface because of a visual impairment, a motor limitation, or unfamiliarity with the input method, the assessment isn’t measuring their math ability. It’s measuring their ability to operate the tool.
This is a validity concern that extends beyond input methods. Design choices such as screen reader compatibility or the size of user controls affect whether students can participate meaningfully in the assessment. For example, research from the StereoMath research project found that existing equation editors imposed a high cognitive load, particularly on students with disabilities, while more natural input approaches reduced that burden across the board.
Modern math input tools are increasingly designed to address these challenges. For instance, full keyboard access means students who can’t rely on a mouse or touchscreen can still complete every interaction. Meanwhile, MathML-based alt text generation makes equations readable by screen readers, turning visual notation into something that can be heard and understood. Together, these developments reflect a shift toward treating usability and accessibility as core design principles rather than afterthoughts.
How Platforms Capture and Use Richer Evidence
Improving the student experience is only one half of the challenge, however. The other is what happens on the backend—how platforms capture, store, and interpret the evidence generated by richer input methods.
This is where open standards matter. For example, standards such as MathML and Question and Test Interoperability (QTI) help ensure that mathematical content, scoring rules, and assessment metadata can move between different platforms without needing to be recreated or reconfigured.
Process-oriented assessment is also becoming more practical. A 2025 study analyzing logs from France’s national grade 9 math assessments found that digital process data—input sequences, revision patterns, and time-on-task—could reveal solution strategies and misconceptions that final answers alone would miss.
Platforms built on modular, standards-based architectures are better positioned for this kind of work. TAO, for instance, uses the QTI and Learning and Test Interoperability (LTI) standards natively and integrates math editing directly into the assessment workflow. As a result, input tools, scoring logic, and assessment delivery operate within a unified system without the need for custom integrations.
“Assessment leaders increasingly recognize that the quality of the evidence they collect depends on the quality of the student experience. Our partnership with Wiris reflects a shared commitment to making digital STEM assessment more intuitive, accessible, and effective. By combining advanced math and science input capabilities with TAO’s standards-based platform, we’re helping organizations create assessment experiences that better capture student thinking and support the future of digital learning and evaluation.” — Miguel Prieto, VP of Corporate Strategy, Open Assessment Technologies
What To Look for When Evaluating Math Input Tools
The best math input tools reduce friction without sacrificing accessibility, interoperability, or the quality of the evidence collected. To evaluate them effectively, instead of comparing feature lists, test how students actually use the system under real assessment conditions:
- Ask students to complete sample problems using handwriting, keyboard, and touch input to see where friction appears.
- Run a timed session with students who haven’t seen the tool before and note where they get stuck.
- Have someone navigate the full assessment using only a keyboard and a screen reader to test accessibility end-to-end.
- Check whether student responses can be exported in structured formats (like MathML or LaTeX) that your scoring and analytics systems can read.
- Try importing and exporting a sample item package to confirm QTI compliance works in practice, not just on paper.
- Test with your most complex item types—matrices, chemical equations, multi-step proofs—to see if the tool handles them cleanly.
- Ask the vendor to show you what process data looks like—input sequences, revision history, time-on-task—and check whether your team can actually use it.
Strengthen Your Math Assessment Capabilities With TAO
Digital math assessment hasn’t struggled because the technology doesn’t exist, but because the input layer hasn’t kept pace with what modern curricula require of students. The gap between what we want to measure and what digital tools have historically allowed has been real.
That gap is now narrowing, however. Handwriting recognition, visual editors, and accessible input design are making it easier for students to show their thinking. Open standards and process data are giving programs the infrastructure to capture and use that evidence at scale. The programs that treat math input as a strategic design decision—rather than an add-on—are the ones building systems that can grow with their needs.
If you’re ready to explore how natural math input can work within your assessment program, TAO’s integration with MathType by Wiris brings handwriting recognition, visual editing, and accessible STEM assessment directly into a standards-based, interoperable platform.
Schedule a demo to see how TAO and MathType by Wiris enable students to enter mathematical and scientific notation naturally while giving assessment teams access to richer, more valid evidence of learning.