How to team up with AI: 3 steps for assessment redesign

Does the development of generative AI really change our expectations in higher education? 
Perhaps a little, but not really.  

Student collaboration

We have always expected our students to develop expertise, communication, leadership, and problem-solving abilities by working together, using the latest technology, and combining their unique skills in effective teams. The collective learning process, when students explain concepts to each other and review each other’s work, helps build a refined understanding of any subject. We’ve also always expected students to work honestly and with integrity.

Collaborating with peers provides fresh perspectives, but any submission must be a product of their own understanding and effort, with all sources painstakingly acknowledged. Collusion can hinder individual learning since it allows students to bypass the process of understanding and mastering the material on their own. 

Expectations and options around AI

When it comes to generative AI, our expectations about these processes should stay the same—encourage AI collaboration and preclude AI collusion. However, surrounded by uncertainty and complexity, educators face the difficult task of balancing AI literacy with academic integrity. 

We have three options at the moment: at one extreme, we can use assessments that render the use of generative AI impractical or impossible. For example, in written or oral exams the immediate application of AI tools for generating or accessing information is restricted or irrelevant. But besides being operationally difficult or time-intensive to implement, exclusion of AI is not reflective of real-world expectations and can stifle the development of crucial skills in our graduates. 

At the other extreme, we can pretend things are not changing exponentially, and allow, through inaction, the surreptitious use of generative AI tools. Overreliance on AI produced work is not only undermining integrity and originality, but by supplanting genuine learning it ensures future graduates will be easily replaced in the workplace by the more skilled AI technology. 

The sweet spot

The sweet spot in between is AI collaboration, where we design our delivery and assessments to develop a partnership between students and generative AI tools. We want AI to enhance, but not replace, the creative and cognitive processes involved in academic work. 

How do we reach the desired zone of AI use? In a recent project, we stress-tested a host of assessments from a business school curriculum, to discover what makes them vulnerable or resilient to AI collusion. 

We found that while real-world assessments, artifacts, group work, reflective work, or contextualised tasks reduce collusion with AI, they don’t eliminate it, nor do they make it easier to spot by educators. One-dimensional assessment changes, such as making a standard essay into a reflective one, does not exclude AI use, does not enhance how skilfully AI is used, and does not automatically foster authentic learning. 

A process for assessment redesign

Instead, we propose is a three-step process for assessment redesign:  

1. Understand how a specific assessment task allows for AI collusion by stress-testing it 
2. Map the assessment task against several key parameters, including structure, context, criticality, format, or foundation by taking this quiz  
3. Redesign assessments by moving and combining these parameters to create multidimensional tasks, that encourage AI collaboration (learn more here) 

Don’t just ask for a lengthy case study analysis or a structured video presentation. Instead, ask students to design a short case study on a public company, then work in class in teams to analyse a specific change in the political environment, further develop brief written recommendations as homework, and finally return to class to defend their opinions. AI can help them with idea generation, to research and write the case study, while they add creativity, crucial decision-making, and teamwork. We explore other assessment redesign examples here. 

Authentic assessments

Following these three steps, we can design assessments which compel students to develop expertise across multiple knowledge dimensions and applications. And it is precisely by looking at the interplay of these dimensions that educators can more confidently know AI’s limitations, allocate clear uses for it, and determine the extent of genuine learning for students. 

Designing multidimensional, but authentic assessments, which blend various cognitive processes and skills, will allow educators to team up with AI, to mix students’ unique skills and knowledge with the ever-growing abilities of artificial intelligence. This will help cut down on unnecessary work, so students can focus on developing the skills required to stand out in a job market that will soon embrace AI in every aspect.  

If you want to learn more about our project, use our AI toolkit, and get inspired, have a look at https://aiinhighered.com/assessments. 

With thanks to the MMU Centre for Learning Enhancement and Educational Development for providing project funding. 

This project was collectively produced by academics from the Strategy, Enterprise & Sustainability Department (Dr Carmen-Elena Dorobat and Professor Sarah Underwood) and digital education specialists (Andrew Larner and Jack Sutherst) at Manchester Metropolitan University. The group have a shared background in technology-enhanced pedagogical development, including using large-scale simulations, tablet/mobile use in classrooms and MOOCs.