One of the most important applications of Artificial Intelligence in education relates to innovation inschool assessments. The challenge lies in promoting change in a context where students literally have technology in the palm of their hands in the classroom, while doing homework, or even during a test. 

It is becoming increasingly clear that assessment methods should have changed since the first browser (also known as an Internet browser) was created. Recent and growing discussions on the topic among teachers and the school community underscore that it is necessary to change the way students are assessed.

According to a concept developed by Justin Reich, director of the MIT Teaching Systems Lab, Artificial Intelligence is an “arrival technology” —that is, it is a tool that leaves no room for preparation; it simply appears and impacts the entire environment and how it functions.

Generative AI—originally through ChatGPT, for example—made its way into schools without any prior planning. This is very similar to what happened when cell phones began appearing in schools around 2010: it was only in recent years that a public debate began to establish restrictions on the use of these devices in schools. There was no preparation, and now we’re dealing with the consequences.

In addition, here’s an important point to consider: Would it make sense to ban it outright? One thing is certain: even a ban isn’t enough to bring about changes in the way we interact with one another—changes that will impact school life.

With AI, the situation is even more complex, given that we’re dealing with software. And the first casualties are assessments and homework assignments. In a world where, with just one photo, it’s possible to get a detailed answer in seconds, how can we be sure that a given conclusion was actually written or thought up by the student?

The simplistic solution that might come to mind is to use Artificial Intelligence itself to combat AI. But in the vast majority of cases, that option doesn't work. Detectors aren't reliable, and this is a futile and dangerous game.

So what should we do, given that assessments are essential to the learning process? How can we tell if the educational objectives have been met?

One possible answer may lie in the scale developed by four international education researchers: Mike Perkins (British University Vietnam), Leon Furze (Deakin University), Jasper Roe (Durham University), and Jason MacVaugh (British University Vietnam). The tool categorizes the use of AI in school assessments into five levels, with the goal of enabling teachers to select the appropriate level of generative AI use in assessment activities based on the learning outcomes they wish to achieve.

See also:

• Beyond the Hype: How Technology Can Make a Difference in Education
• What role do school leaders play in the ecosystem in which they operate?

The first level focuses on learning objectives that require assessing knowledge students must master on their own. Given the fundamental assumption that all students have access to AI, it is important in this case to conduct a synchronous, supervised assessment without the use of technology, ensuring that students rely solely on their existing understanding and skills.

At Level 2, students are allowed to use AI as long as it does not interfere with their ultimate learning goal. However, to avoid completely outsourcing their cognitive processes, students must demonstrate the entire process they followed to arrive at the answers.

At the next level, AI can be used as a collaborator that “critiques” and “offers feedback” on the student’s work. Through this partnership, the student should apply a critical perspective to their interaction with the AI. This helps them develop an analytical mindset and even incorporate the tool into their workflows in a productive way.

At the penultimate level, there are no restrictions on the use of artificial intelligence. Students direct it toward their goals, but that does not mean they should be uncritical. They must still provide a detailed explanation of the entire rationale behind their decisions.

Finally, at Level 5, there is an explicit objective in which the use of AI is mandatory. In other words, the tool ceases to be merely a support tool and instead incorporates a layer of capacity-building to use it effectively in order to achieve concrete objectives. It is worth noting that, at all levels, the common thread is intentionality: that is, what lies behind the assessment and the learning experience. Students need to learn how to work with AI, but with a variety of objectives in mind.

The use of technology at different times may or may not make sense, and the scale presented here can help achieve these objectives. However, the task of defining best practices for the use of AI does not fall solely to teachers. There is an urgent need for a discussion at the institutional level—involving public officials and the school community—on how to create a support system for adapting the schools’ assessment ecosystem. 

In this rapidly changing landscape, where technology is advancing faster than educational practices can keep up with, rethinking school assessments has become not only necessary but urgent.

Artificial Intelligence is not a threat to learning, but an invitation to reinvent it—provided it is implemented with purpose and responsibility. It is up to the educational community, together with policymakers, to develop viable and ethical approaches so that assessment remains a tool for meaningful learning and not merely a reflection of the limitations of the past.

About the author:

Share on social media: