TEASIG – AI and item writing for high stakes testing: insights from text and multiple-choice question generation

We now live in an age of myriad opportunities to harness the power of artificial intelligence (AI), especially large language models (LLMs), in a range of educational contexts. From a testing perspective this has included opportunities to develop secure, quality test content (including text passages and comprehension question for reading and listening tasks) using online platforms. In this session, we will discuss how one Awarding Organisation has implemented a process for item writing which exploits the available technology whilst adhering to best testing practice. We discuss how to use AI ethically and efficiently to generate test content, whilst keeping a place for human creativity and ingenuity. We consider the limitations and challenges that AI brings to the fore and consider a process of good practice for content development from the perspectives of both production and review.