AI & the Emancipation of Assessment

Surely the biggest issue that #chatgpt has thrust in front of the face of education is students capacity to cheat, right?

Looking at this with less 'shockjockism' and from the perspective of teachers, we can talk about the issue more productively by discussing how #ai challenges assessment practices. That is a discussion that every educator should be having right now no matter what sector you are in or discipline you teach. Such a discussion should weigh up pros and cons, ask questions about what we gain and what we lose but more importantly what we value going forward from here.

This article is not going to give that balance to the discussion on #assessment practices in #education. It is, conversely, going to focus on the positive, suggesting that AI can lead to an emancipation of assessment and assessment practices.

What is currently the norm in assessment practices?

I think it is fair to say that currently, education still relies on the tried and tested assessment practices that are mostly made up of:

• Exams
• Coursework
• Essays
• Writing

We label the significance of the assessments that lead to students achieving, passing and being awarded grades/certificates, as summative.

According to NSW education, summative assessment is:

A type of student?assessment?that occurs at the end of a course of study or period of time and indicates a student's achievement level against learning

Much of the rhetoric around summative assessment refers to it as coming at the end of something significant. To be honest, that makes sense. This slide taken from helpfulprofessor represents a common overview:

So, my proposition is that that AI can offer disruption to these norms. It can pressure us into using more diverse practices and even destroy the need to separate formative and summative assessments so that EVERYTHING on a learning journey (you heard it right, EVERYTHING) is factored into a how a student is assessed

In what ways can AI can lead to an emancipation of assessment and assessment practices?

Think about this:

1. A student is involved in discussions in a class. That student shows a degree of understanding of a concept that the teacher is trying to teach. When it comes to the exam or in the coursework (the summative) that students does not show the same level of understanding. Wouldn't it be an improvement if a student could rely on the things said in that discussion rather than the summative?
2. During practical activities that a student is doing in a homework task, they demonstrate a skill that they are expected to develop as part of the discipline being taught. However, they can not perform this skill at the same level during a practical exam. Wouldn't it be an improvement if a student could rely on this homework evidence rather than the exam?

These are just 2 examples and what they highlight is nothing revolutionary. Those educators out there who rely on continuous assessment will be saying that they already do some of this especially number two. The major issue is that such practices are problematic in terms of:

• Scale - Managing 100s of university students on a course and having to continuously assess is a major headache
• Moderation - How do you guarantee that how one assessor views the quality of what they witness is the same as another? Exams are good at this!

But these two problems stem from this:

• Capture, collation and management of data - Particularly for the first example above, there are major logistics in trying to capture everything for everyone even in a small group of learners that then feeds into the assessment.

---------Welcome to the stage #aiassessment------------

#aiassessment provides us with opportunities to efficiently and effectively deal with the capture, collation and management of data. A trained model can, both in practical exercises, in discussions and in just about every way we can think of, be trained to collect, collate and provide the data we need to assess on scale. With the right algorithms and training, it can then take the slack out of moderation.

#GenerativeAI is rapidly expanding and education is a fertile ground for this. I urge you to look at the work of C3L or get in touch with Jarrod Johnson to get a feel of something more than an #aichatbot that they are developing for classroom based assessment.

Concluding remarks

I am going to be bold here and predict that:

• How we assess in the next couple of years (yes, two years) will look very different from how we assess now
• Assessment will not be separate from teaching and learning. It will be synthesised as part of what happens within a learning environment and/or a learning journey
• AI use will be ubiquitous in education and in assessment practices
• We will laugh at the headlines about how 'AI encourages cheating' and 'Let's ban these AI thingies'

#chatgpt is a powerful sledgehammer to the doors of education and assessment but it is merely opening the door to let the 'end-of-level boss' that needs to be destroyed to really disrupt assessment practices.