AI is bringing handwritten assignments back to classrooms

Some teachers in the United States are returning to paper, pencils and assessments completed in class. Their aim is not to remove every technology. They need a reliable way to see what a student can do independently when writing assistants can generate a finished response in seconds.

According to a nationwide survey reported by CalMatters, more than 70% of teachers are concerned that submitted work may not be the student’s own. Nearly three quarters also worry that students are getting less practice in writing, research and reading comprehension.

Why digital tests are harder to supervise

The issue extends beyond copying text from another window. Visual search features built into a browser can analyse a question on screen and display an answer without leaving the test page. From a teacher’s perspective, independent reasoning can be difficult to distinguish from automated help.

AI-text detectors do not settle the matter. They produce false positives, can disadvantage certain writing styles and are not sufficient evidence by themselves. A school policy based only on detection can therefore create as many disputes as it resolves.

Paper as a point of reference

Several California teachers now require tests, some homework and first drafts to be handwritten. This creates a sample that can be compared with digital work and makes the student’s reasoning easier to observe.

Paper is not a universal solution. It makes marking and accessibility harder and does not suit every learner. Avoiding AI altogether also removes opportunities to teach responsible use of a tool that is becoming relevant across many professions.

What the MIT study does and does not show

A MIT Media Lab project compared 54 participants writing essays with a large language model, a search engine or no external tool. EEG measurements showed the most distributed connectivity in the unaided group and the weakest in the language-model group.

The finding needs careful framing. The MIT Media Lab states that the work is a preprint that has not yet undergone peer review. The sample was small, the task was limited to essay writing, and the study does not establish that every use of AI causes lasting cognitive harm.

Clear rules are more useful than a general ban

For each assignment, a teacher can state what is allowed: no assistance, language correction, idea generation, or full use with disclosure and sources. Drafts, version history, a short oral defence and explanations of key choices make the process assessable.

The central question then becomes more useful than simply asking whether AI was used: does the student understand the submission, know how to verify its sources and have the ability to justify the decisions behind it?

Sources

This summary draws on CalMatters and the MIT Media Lab. Photo: Yan Krukau, Pexels.