AI-assisted marking: Giving instructors their time back

Derek Keats

AI recommends, the instructor decides, but this frees up time to engage students more.
AI recommends, the instructor decides, but this frees up time to engage students more.

Can artificial intelligence reduce the burden of marking without taking professional judgement away from educators?

Assessment is an essential part of learning. Students need to know what they understand, where they are struggling and how they can improve. Instructors need to evaluate progress, identify misconceptions and provide useful feedback.
But marking can also be one of the most time-consuming parts of teaching.

Multiple-choice questions can often be marked automatically. Written answers, explanations and applied questions are different. These require interpretation and judgement, and providing meaningful feedback takes time.  This is where AI-assisted marking offers an interesting opportunity.

From automatic marking to intelligent assistance

Chisimba already supports conventional automated assessment through multiple-choice tests and related assessment tools. AI-assisted marking offers the possibility of extending that efficiency to responses that cannot be evaluated using simple predetermined answers. 

Consider a student answering a question about savanna ecology:

Explain how fire, rainfall and herbivory interact to influence vegetation structure in an African savanna.

A meaningful answer requires more than remembering definitions. Students need to demonstrate an understanding of ecological processes and how they interact. An AI system can help an instructor examine whether an answer addresses the important concepts, identify possible omissions and prepare suggested feedback. But evaluating understanding is not always straightforward.

There may be several scientifically valid ways to explain a process. A student may use an unexpected example or demonstrate understanding using terminology different from that in a model answer.

These are precisely the situations where educators need to retain control.

The instructor remains the assessor

Our approach to AI-assisted marking is based on a simple principle:

  • AI can recommend. The instructor decides.

A responsible AI-assisted marking workflow should allow educators to establish assessment criteria, review AI recommendations, correct errors and determine final outcomes. For example, AI might suggest that a student's explanation demonstrates a good understanding of herbivory but overlooks the relationship between rainfall and fire frequency. The instructor can examine that recommendation against the student's actual response and the assessment criteria.

The benefit is that the instructor receives a useful starting point rather than having to construct every piece of feedback from scratch. The responsibility for the judgement remains human.

Better feedback, not just faster marking

Speed is not the only potential benefit. Feedback is most valuable when students receive it while the activity is still fresh in their minds and when it tells them something useful.

AI assistance can help educators prepare more detailed feedback, identify recurring misconceptions and suggest areas that need attention. An instructor might discover that many students understand the relationship between rainfall and grass growth but struggle to explain how grazing and fire interact. This is more useful than simply knowing that the class achieved an average mark of 65%.

Such information can inform subsequent teaching and help identify topics that need further explanation. AI therefore has the potential to support assessment as a learning process, rather than treating it only as a mechanism for allocating marks.

Fairness, transparency and accountability

Using AI in assessment also introduces important responsibilities. AI systems can make mistakes. They can misinterpret answers, overlook valid reasoning or favour particular forms of expression.  This makes it essential to design assessment workflows that support verification and accountability.

A responsible system should make it possible to distinguish AI suggestions from final instructor decisions, apply clear assessment criteria and review outcomes. Students should also understand when AI has contributed to the assessment process. Particular care is needed when assessments influence qualifications, professional progression or other significant decisions. In such situations, meaningful human oversight is not an optional extra. It is fundamental to trustworthy assessment.

Reducing the burden on instructors

Instructors should be spending their time helping students develop knowledge, skills and judgement. They should not have to spend unnecessary hours repeating administrative work that technology can help simplify.

AI-assisted marking offers a way to reduce that burden while preserving the educational value of assessment. Combined with Chisimba's capabilities for online assignments, multiple-choice tests, gradebooks and learning content, it creates opportunities for more integrated assessment workflows.

The aim is to make feedback more practical, marking more manageable and student progress easier to understand.

Assessment should encourage curiosity

There is a broader educational principle behind this work. We want students to become addicted to finding out, not frightened of being tested. Assessment should help students discover what they know, recognise what they still need to learn and develop the confidence to improve.

When technology reduces repetitive marking work, instructors have more time to engage with the learning itself. And when AI is used responsibly, it can support the quality of feedback without replacing the experience and judgement of educators.

Let AI assist with the marking. Keep educators in charge of the learning.

Chisimba — building human-centred AI into an open-source learning platform.

Chisimba is a managed by Kenga Solutions (Pty) Ltd, a company based in Hoedspruit, Limpopo, South Africa with a deep focus on teaching-and-learning about nature and the bush.