Creating secure assessments to meet the challenges posed by AI

Higher education institutions are increasingly embracing the benefits of generative AI, shifting away from the initial fear when ChatGPT took the world by storm in November 2022. When it comes to effective and secure assessments, however, there is still much to be determined. This was the focus of a Times Higher Education webinar on exam security, held in partnership with Turnitin. The webinar panel was made up of academics and assessment experts from medical and health science institutions around the world. They explored strategies for building secure assessments in the age of AI and the hurdles institutions face in this process. 

The popularity of AI tools has disrupted assessments and exam security in most disciplines. “When the ChatGPT bomb exploded I felt protected because most of our healthcare and medical exams are multiple choice or single best answer, unlike other disciplines where there are essays that students could cut and paste,” said Terese Bird, educational designer at Leicester Medical School. 

However, Bird later changed her stance, realising that medical professionals might be required to use AI in their future careers and considering the emphasis on essay writing at the postgraduate level. This prompted Bird and her colleagues to consider the role of AI in assessment. “We now use it in formative assessments,” she said. “We give students a question, ask them to use ChatGPT to research the answer, and then critique it in teams.” 

Due to the digitalisation of teaching and assessment, institutions must establish guardrails against new challenges to exam security, through measures such as online proctoring and offline digital assessments. Eric Connelly, regional vice-president for North America at Turnitin, said the challenge lies more in completely disconnecting from the digital world while allowing students to use a digital device. 

Panellists agreed that using AI in assessments should prompt a discussion with students rather than it being a unilateral decision. Beginning with transparency is crucial, said Daniel Trujillo, director of assessment at Harrison College of Pharmacy at Auburn University. “We’re working on a policy around AI use for students, asking questions such as, ‘did you use it in brainstorming or fact-finding?’” said Trujillo. “But we also owe it to them as academics to be upfront if we’re using it.” 

Sabeeh Mashhadani, director of the Assessment Office at the Arab Board of Health Specializations, said his focus was on trying to move away from high-stakes summative assessments. “Formative assessment is a great area to use AI because it can offer immediate feedback and adaptive assessment for each individual. For example, determining the next test set for a particular student,” he explained.   

Matt Koss, academic instructional designer and academic technology coordinator at Michigan State University College of Nursing, argued that AI can be transformative in supporting time-pressed instructors who have to balance teaching with their professional workload. “Lots of our instructors are trained as medical professionals, not instructors. AI can help them to build exam questions and provide a good starting point,” he said. 

Solutions offered by technology partners such as Turnitin can help institutions improve existing assessment models and navigate challenges surrounding academic integrity. Turnitin’s ExamSoft helps institutions design secure, effective and engaging digital assessments to evaluate students’ knowledge and skills and ensure the quality of academic outcomes, especially in high-stakes disciplines. 

The panel: 

• Terese Bird, educational designer, Leicester Medical School
• Eric Connelly, regional vice-president for North America, Turnitin
• Matt Koss, academic instructional designer and academic technology coordinator, Michigan State University College of Nursing 
• Sabeeh Mashhadani, director of the Assessment Office, Arab Board of Health Specializations 
• Sreethu Sajeev, branded content deputy editor, Times Higher Education (chair)
• Daniel Trujillo, director of assessment, Harrison College of Pharmacy, Auburn University

Resources:

• The UK Quality Assurance Agency for Higher Education's advice and resources on generative AI
• Jisc resource page on AI
• UNESCO's guidance for generative AI in education and research
• UK's AI Standards Hub
• Download “Crafting your GenAI and AI policy: a guide for instructors” by Tricia Bertram Gallant from UC San Diego at the top of the article

Explore more insights from events held in partnership with Turnitin here.

Find out more about ExamSoft by Turnitin.