Gen AI and Post-entry Language Assessment (PELA): Where are we now?

Post-entry language assessments (PELAs) have been commonly used in the Australian and<br/>New Zealand university contexts to identify students who may require academic language<br/>and literacies support (for review, see Read, 2015). PELAs have been proven to be effective,<br/>valid instruments in achieving their aims. However, several PELAs, such as Bond English<br/>Language Assessment (BELA; Lydster, 2024), often require students to submit a piece of<br/>academic writing and are self-administered by students online. This raises potential academic<br/>integrity concerns, especially given the increasing development and use of Generative AI<br/>(Gen AI) amongst students. <br/><br/>This poster presents two iterations of PELAs created in response to the use of Gen AI, such as Chat GPT, Copilot, and Perplexity, namely BELA-AI and Bond Essay Skills Task (BEST). BELA-AI, a formative assessment item incorporating Gen AI, was developed for the Bachelor of Medicine program. The 60-minute assessment requires students to write an academic essay, generate an essay and, finally, write a brief critical reflection analysing the two essays. BEST, in contrast, is integrated into the core subject, Critical Thinking &amp; Communication and requires students to write an academic essay with reference to at least one of two prompts on a controversial topic. <br/><br/>BEST Essays are written via Cadmus, a platform embedded into the learning management system of the subject, providing integrity assurance by allowing stakeholders to view integrity analytics and work behaviours of students. Although further research is necessary, initial discussions with stakeholders suggest preliminary evidence that the two PELAs have reduced academic integrity concerns regarding the original BELA, whilst continuing to be the initial “conversation starter” for students with advisors at the Academic Skills Centre.