Katerina Holmes
The education system has often undergone overhauls in an attempt to improve the standards of student learning, allowing them to have access to better repositories of information and to help them channel their true potential. In this near-constant process, AI in education has also made its mark, allowing educators to simplify certain concepts or administrative responsibilities. However, more recent developments in the world of AI have been more disruptive, given that students have begun slowly circumventing core learning processes and tasks with the help of AI applications. ChatGPT being one of them, numerous tools backed by AI are now in the market, making it complex for educators to sift through the numerous AI-generated submissions coming their way. While the risks posed to the time-tested learning methods are obvious, an important yet overlooked aspect of these technologies is their impact on student assessment.
It is a known fact that excessive reliance on AI tools might end up impacting student learning; however, understanding the assessment of students in a period where AI is getting a serious push from the scientific community is equally important. AI’s percolation into education might not be a recent phenomenon, although its wider range of capabilities is now making way for its formal inclusion into the education space—something teachers are not quite up to speed with. Conversely, students being quick learners and early adopters of new technologies seem to be much more at ease deploying AI tools in their coursework. However, understanding AI will soon impact the way teachers assess students is just as essential. From AI scores to AI in teaching, a few important observations follow.
Evolving Student Assessment: Speculating How AI Scoring Tools Might Play a Role
The complete automation of student assessment might not be in the best interests of learning.
Image Credit: Photo by Andy Barbour
The inclusion of AI and AI-supported applications has grown in the education sector over time. However, the advent of technically advanced AIs with complex linguistic backing and language models is gaining momentum among students. Students have been using language model AIs such as ChatGPT to generate essays, papers, and answers to complex questions in their assignments, essentially finding an easy pass to their tasks at school. While inherently unethical, regulations on the use of the application are bound to stymie the widespread deployment of ChatGPT. Regardless, it must be stated that language model AIs and constant improvements are bound to remain and eventually become integrated even with education. Recent events have made it apparent that student usage of these applications have numerous implications, but what impact will it have if used to design assessment models by teachers and other academics? This question will need a multipronged approach that considers both the drawbacks and advantages of artificial intelligence models such as ChatGPT.
Considering that language model AIs and AI tools are often pre-trained and rely on set parameters alongside samples of similar styles of writing and expression, the assessments of students that do not rely on conventional methods of writing or choose to express themselves differently might not be judged fairly by an AI scoring model. Bias might also impact the way AIs evaluate student responses, given that the data AI tools are trained on might be merely reflective of a lack of opportunity rather than a lack of capability. This indicates that if the samples inherently contain data that cannot be entirely relied upon, neither can the outcomes. While this is a challenge, developers and scientists do believe that this can be mitigated by training these language model AIs on a wider set of samples to ensure the network has more references to rely on when carrying out assessments. Despite apparent challenges, thinkers and certain sections of academia still believe that AI can be beneficial to the overall assessment process. Freeing up the manual load on the teacher to set up individual learning pathways, AIs seem to show a degree of promise when it comes to simplifying tasks for teachers, allowing them more time to pay attention to the academic performance of their students.
Future Possibilities for AI in Teaching
A combination of technology and human touch are ideal for sensitive professions such as teaching.
Image Credit: Photo by Julia M Cameron
The recent COVID-19 pandemic exposed traditional learning establishments and systems to several lacunae that needed to be addressed in an environment where online learning was essential. The resulting technology vacuum also presented educators with the potential to try out AI-supported tools to bridge the gap between students and teachers, while also taking off the added stress from the latter’s shoulders. Online student assessment was an especially tricky aspect of remote learning that presented numerous challenges for schools and universities looking to conduct examinations from a distance. AI has the potential to be tailored to monitor remote learners and perform automated assessments to allow teachers to remain focused on creating a set pathway for their students. Similarly, AI can also be used to generate learning analytics and reports to help students understand their balance of strengths and weaknesses when it comes to their coursework. Combining language model technologies with human supervision can also allow teachers to generate fool-proof assessment paradigms that are suited to students’ unique capabilities based on their learning capacities.
Will AI Scores Replace Conventional Assessments?
AI can be used to support conventional assessment techniques to improve students’ academic performance.
Image Credit: Photo by Jessica Lewis Creative
Despite the requirement of more pointed research to help actualize these systems in practice, it must be noted that AI and AI scoring cannot completely replace human teachers and student assessments. AI is often given to a mechanical approach relying only on its training data. Without nuance, a sensitive task like grading students cannot be outsourced completely to a machine-based algorithm. Given that the teacher-student relationship can entail several unique experiences that help both the teachers and students, AI can only be used as an additional aid to support and improvise on the existing methods of assessments. AI-generated assessment models and analytical tools can also allow teachers to craft more pointed strategies to train students on core concepts with more pointed efficiency. As technology and pedagogy both develop at a quick pace, the educational sector requires a greater deal of investment and research to put into practice the frequent and numerous advancements of modern science.