When Good Work No Longer Proves That Learning Has Happened

For many years, lecturers could reasonably assume that a submitted assignment reflected the student’s effort, understanding and ability.

That assumption is no longer safe.

Generative AI can now help students:

• Interpret an assessment question
• Develop an outline
• Generate arguments
• Write a report
• Create references
• Produce presentation slides
• Develop computer code
• Write a reflection
• Improve the final language

The submitted work may look excellent.

But can the student explain it?

Can the student defend the decisions?

Can the student apply the same knowledge to a different problem?

Can the student recognise when the AI-generated answer is wrong?

This is the real challenge.

Universities may receive better-looking assignments while becoming less certain that students have achieved the Course Learning Outcomes.

Assessment is not only about producing work. Assessment is the checkpoint used to determine whether learning has taken place.

If AI performs the intellectual work expected by the CLO, then the assessment may no longer provide valid evidence of the student’s achievement.

For example, consider a CLO that requires students to:

Analyse a complex organisational problem and recommend an appropriate solution.

If AI performs the analysis, develops the recommendation and writes the justification, what exactly has the student demonstrated?

Editing the AI output is not necessarily the same as achieving the CLO.

This becomes even more serious when conventional final examinations are replaced with unsupervised take-home assessments.

A lecturer may:

1. Take the existing examination questions
2. Upload them into the LMS
3. Give students 24 or 48 hours to respond
4. Rename the activity “Final Assessment”

That is not assessment transformation.

It is simply an unsecured take-home examination.

The assessment loses the controls of a supervised examination, but it does not gain the strengths of a properly designed authentic assessment.

The result may be the worst of both worlds:

• It is not a secure examination.
• It is not an authentic assessment.
• It may not verify individual competency.
• It is highly exposed to AI-generated answers.
• It may encourage collaboration beyond what is permitted.
• It may produce unreliable CLO attainment results.

There is another uncomfortable truth:

Authentic assessment is not automatically AI-resistant.

AI can produce case analyses, lesson plans, marketing strategies, business proposals, research summaries, reflective writing, computer code and project documentation.

Adding a real-world scenario does not automatically make an assessment secure.

Even personalised questions can be uploaded into an AI tool.

        AI detection software will not solve the problem either.

Detection may raise concerns, but it cannot reliably tell us:

• Who completed the work
• How much AI was used
• Whether the use was permitted
• Whether the student understood the submission
• Whether the student personally achieved the CLO

The solution cannot be limited to catching students.

The better question is:

What evidence would convince us that this student has genuinely achieved the CLO, whether AI was used or not?

This does not mean universities should return entirely to conventional examinations.

It means assessment can no longer depend only on the final product.

A polished report is no longer enough.

A complete portfolio is no longer enough.

A professional-looking presentation is no longer enough.

Lecturers may also need to see the student’s:

• Development process
• Decisions and justification
• Response to feedback
• Ability to correct mistakes
• Application to a new situation
• Live demonstration
• Oral explanation
• Individual defence

Without these forms of evidence, universities may gradually produce graduates with impressive assignments but weak underlying capabilities.

The qualification may appear strong on paper while becoming less reliable as evidence of what the graduate can actually do.

In the next article, I will explore how institutions can respond responsibly without banning AI or allowing AI to replace learning.

Related articles:

>>> AI Is Changing How Students Learn. That Is Not Necessarily a Bad Thing.
>>> The Danger of Going All Out - Alternative Final Assessment in Higher Education

#HigherEducation #GenerativeAI #AssessmentIntegrity #AcademicQuality #CourseLearningOutcomes