By Dr. Daniel Jackson
Associate Professor of Mathematics, University of Maine at Farmington
I should begin with an important clarification.
I am not the Dr. Daniel Jackson from Stargate SG-1.
Although I admit it is suspicious that both of us spend our careers deciphering mysterious artifacts left behind by advanced civilizations. The television version had alien glyphs. I have student submissions generated by ChatGPT.
Recently, the Los Angeles Times published a collection of letters responding to concerns about AI and academic cheating. Several readers proposed a straightforward solution: bring back blue books, handwritten exams, in-person proctoring, and perhaps, if we’re feeling especially nostalgic, cursive penmanship (Los Angeles Times, 2026).
As a mathematics professor, I am sympathetic.
Every generation of educators eventually arrives at the same conclusion:
The future would be much easier if students would simply stop living in it.
The logic goes something like this:
Students are using AI.
Therefore, we should return to assessment methods from before AI existed.
Following this reasoning, the invention of automobiles should have triggered a renaissance in horseback riding. The calculator should have revived the abacus industry. Google should have inspired a nationwide return to card catalogs. Smartphones should have brought back switchboard operators.
And yet history stubbornly refuses to cooperate.
The Educational Panic Hall of Fame
The current AI panic is remarkably familiar.
When writing became widespread, Socrates worried that people would stop exercising their memories. Writing, he argued, would create the appearance of wisdom without genuine understanding.
Then came the printing press.
Books suddenly became available to ordinary people. Predictably, authorities worried that students would acquire information too easily and lose respect for traditional gatekeepers.
Centuries later, calculators arrived.
I still meet mathematicians who fondly recall predictions that calculators would destroy mathematics education. Instead, calculators mostly destroyed long division worksheets.
The internet was next.
Search engines, Wikipedia, online databases, and digital libraries were all supposed to usher in an age of intellectual decline. Yet universities somehow survived despite students having access to more information than the Library of Alexandria ever dreamed possible.
Now AI has arrived, and once again we are hearing a familiar refrain:
Students can access knowledge too easily.
The irony is that many of these concerns are being voiced from devices that contain calculators, encyclopedias, libraries, maps, translation tools, statistical software, and instant communication with nearly every human being on Earth.
Apparently all those technologies were acceptable.
This one crossed a line.
The Real Problem Is Not Cheating
What’s fascinating is that some of the most important recent research is arriving at a very different conclusion.
A major Cornell University study analyzing responses from more than 95,000 students found widespread use of generative AI and concluded that the central problem is not merely misconduct but assessment validity. The researchers argued that assessment reform is now “necessary and urgent” because traditional assignments may no longer provide reliable evidence of learning or competence (Cornell Chronicle, 2026).
Read that again.
Not “we need better surveillance.”
Not “we need stronger plagiarism detectors.”
Not “we need more webcams.”
We need better assessments.
This distinction matters.
Educational researchers increasingly argue that discussions about cheating are often obscuring a more important question:
Are our assessments actually measuring learning?
As a mathematician, I find this question much more interesting.
Suppose I assign fifty algebra problems.
A student completes them.
A symbolic algebra system completes them.
An AI completes them.
A graphing calculator completes parts of them.
What exactly was I trying to measure?
If the answer is “the ability to produce answers,” then technology has been threatening that assessment model since at least the 1970s.
If the answer is “the ability to reason mathematically,” then perhaps I should be assessing reasoning more directly.
AI did not create this problem.
It merely exposed it.
The Curious Case of the Blue Book Revival
Many universities are responding to AI by reviving handwritten, in-person examinations.
To be clear, there is a place for proctored exams.
They can be useful.
Sometimes.
But let us not pretend that the blue book is an educational silver bullet.
Imagine a hospital announcing:
We have become concerned that physicians rely too heavily on modern diagnostic tools. Therefore, all medical licensing examinations will now be conducted using only handwritten observations and nineteenth-century equipment.
Patients would flee the building.
Professional competence is demonstrated in the environment in which professionals actually work.
Engineers use software.
Scientists use computational tools.
Accountants use spreadsheets.
Programmers use AI assistants.
Researchers use search engines.
Writers use editors.
Mathematicians use calculators, computer algebra systems, statistical software, numerical modeling tools, and increasingly AI.
Why would we design assessments that systematically exclude the tools graduates will actually use?
Imagine training airline pilots by insisting they never touch an autopilot system because “real pilots should navigate by hand.”
That may sound rigorous.
It also sounds like a great way to produce graduates unprepared for their profession.
What UNESCO Thinks Is Worth Measuring
One of the most thoughtful recent discussions comes from UNESCO’s work on assessment in the age of AI.
Their argument is refreshingly simple.
When machines can generate essays, solve routine problems, retrieve information instantly, and assist with complex tasks, education should focus less on rote reproduction and more on higher-order thinking, creativity, judgment, ethical reasoning, and problem solving (UNESCO, 2025).
In mathematics, this means moving beyond:
- Can you execute a procedure?
- Can you reproduce an example?
- Can you remember a formula?
Toward questions like:
- Is this solution reasonable?
- What assumptions are hidden here?
- How would you verify the result?
- What happens when the model fails?
- Which method would you choose and why?
- How would you explain this to another person?
Those are much harder questions.
For humans.
And for AI.
The Great AI Detector Fantasy
Of course, some institutions have attempted a technological solution.
If students use AI, perhaps software can detect AI.
Unfortunately, reality once again refuses to cooperate.
MIT Sloan Teaching & Learning Technologies notes that AI detectors have significant reliability problems and can generate false accusations against students. OpenAI itself discontinued its own AI detection tool because it could not achieve sufficient accuracy (MIT Sloan, 2025).
As mathematicians would say, the false positive rate matters.
A lot.
Imagine a smoke detector that occasionally accused your refrigerator of being on fire.
You would not trust it.
Yet some institutions have been tempted to trust AI detectors with consequences far more serious than a nuisance alarm.
The result is a peculiar technological arms race:
Students use AI.
Faculty use AI to detect AI.
Students use AI to evade AI detection.
Faculty use AI to improve AI detection.
At some point, one wonders whether the humans involved might be able to contribute something useful.
Authentic Assessment Is Not a New Idea
Ironically, many of the solutions being proposed are not radical at all.
They are simply forms of authentic assessment that educators have discussed for decades.
Consider the following assignment:
Use any resources available—including AI—to analyze a real-world problem. Document your process. Explain your decisions. Evaluate the reliability of your sources. Critique the AI’s suggestions. Defend your conclusions.
That assignment measures something meaningful.
Students must exercise judgment.
They must evaluate evidence.
They must identify errors.
They must synthesize information.
They must communicate clearly.
Most importantly, they must do exactly what educated professionals actually do.
Research is increasingly finding that AI-inclusive assessments can provide richer evidence of learning when students are required to explain their reasoning, critique AI outputs, and justify their decisions rather than merely produce answers.
In other words:
The important question is no longer:
Can students generate an answer without assistance?
The important question is:
Can students think effectively while using the tools available to them?
Those are very different questions.
The Future Belongs to Tool Users
The students graduating from UMF today will enter workplaces saturated with AI.
Nobody will reward them for refusing to use available tools.
Nobody will say:
Congratulations on solving that problem without technology. We don’t care that it took three days longer and produced a worse result.
The competitive advantage will belong to people who can combine human judgment with technological capability.
The skill is not avoiding AI.
The skill is knowing when to trust it, when to challenge it, when to ignore it, and when to use it effectively.
That is a profoundly human capability.
And it is exactly the sort of thing higher education should assess.
So yes, let’s keep some blue books around.
They make excellent historical artifacts.
Right next to the slide rules, overhead projectors, card catalogs, and newspaper editorials predicting that calculators would destroy civilization.
Meanwhile, the rest of us should get back to the much harder task of designing assessments that reflect the world our students actually inhabit.
After all, authentic assessment has never been about preventing students from using tools.
It has always been about determining whether they can think.
Even when the tools get better.
References
Cornell Chronicle. (2026, May 21). Widespread AI misuse means higher ed must rethink assessment.
https://news.cornell.edu/stories/2026/05/widespread-ai-misuse-means-higher-ed-must-rethink-assessment
Los Angeles Times. (2026, June 16). Letters to the Editor: To combat AI cheating, colleges should go back to basics for exams.
https://www.latimes.com/opinion/letters-to-the-editor/story/2026-06-16/colleges-ai-cheating-exams
MIT Sloan Teaching & Learning Technologies. (2025). AI Detectors Don’t Work. Here’s What to Do Instead.
https://mitsloanedtech.mit.edu/ai/teach/ai-detectors-dont-work
UNESCO. (2025). What’s Worth Measuring? The Future of Assessment in the AI Age.
https://www.unesco.org/en/articles/whats-worth-measuring-future-assessment-ai-age
No comments:
Post a Comment