Abstract
This study addresses the limitations of traditional assessment practices and proposes a conceptual model to reframe assessments for authenticity in the context of generative artificial intelligence (AI). Traditional assessment practices often fail to capture diverse knowledge and can be exploited by students' misuse of generative AI tools for unfair academic advantages, which underscores the need for robust assessment mechanisms. This study explores how authentic assessments can be integrated with generative AI tools to mitigate academic dishonesty and enhance the learning experience. Building on existing AI approaches in higher education, this study develops a model integrating generative AI in authentic assessments. This model can serve as a framework for incorporating authenticity in assessment practices while leveraging the capabilities of generative AI. An example illustrating the conceptual model, along with several reimagined authentic assessment types, and mitigation strategies for reframing authentic assessment design, are provided.
TopIntroduction To Authentic Assessment In Generative Ai
Generative AI has recently gained prominence in education and research. In academia, AI tools excel in generating literary texts (Tlili et al., 2023), aiding curriculum development (Cotton et al., 2023), promoting collaborative learning (Dwivedi et al., 2023), and enhancing student engagement (Kuhail et al., 2022). Moreover, generative AI offers personalised assistance (Farrokhnia et al., 2023) and the potential to evolve into a one-on-one personal tutor (Yeadon et al., 2022). Generative AI’s diverse capabilities promise to transform education and research practices; however, students have been using generative AI tools inappropriately to gain an unfair advantage in their academic pursuits (Hung & Chen, 2023). These instances of misuse often involve using such tools to generate essays, answers to assignments, or even responses during exams, which can compromise the integrity of the educational process.
Assessment plays a pivotal role in student experience and institutional administration. However, its implementation within regulatory quality frameworks has led to technological challenges and student issues associated with the integrity of the assessment (Bryant, 2023). Authentic assessments, which emphasise real-world problem-solving and critical thinking (Ashford-Rowe et al., 2014), can help address academic misconduct with generative AI tools by evaluating students' genuine understanding and application of knowledge rather than rote memorisation or plagiarism-prone tasks. An empirical study by Sotiriadou et al (2020) implemented scaffolded authentic assessments, including interactive oral examinations, to enhance academic integrity and reduce academic misconduct. These assessments were found effective in aligning with real-world scenarios, thereby diminishing misconduct tendencies while also improving students' professional identity, communication skills, and employability prospects.
This chapter aims to delineate the attributes of authentic assessment to enhance student learning in an AI-centric environment by distilling the components from the current body of literature, constructing a conceptual framework, and employing this model to maximise the benefits of generative AI in academic contexts. This framework aims to safeguard academic integrity by effectively mitigating the misuse of AI in educational settings. Although the purpose of this chapter is to conceptualise a framework that alleviates misconduct through mitigation, the term 'mitigation' in this context does not imply avoidance of AI use by staff. Rather, it refers to implementing authentic assessment as a strategy to mitigate students' potential misuse of generative AI tools. This nuanced interpretation is essential for accurately conveying the intended message within the developed academic framework.
Key Terms in this Chapter
Higher-order thinking: Advanced cognitive processes like critical thinking, analysis, synthesis, and evaluation that involve deeper levels of understanding and application.
Deep Learning: A subset of machine learning that uses artificial neural networks with multiple layers to analyse and process complex patterns in data.
Academic Integrity: Adherence to ethical principles and standards in academic work, including proper citation, avoiding plagiarism, and respecting intellectual property rights.
Variational Autoencoders (VAE): A type of artificial neural network used in machine learning, particularly for generative tasks, that learns to encode and decode data in a probabilistic manner.
Authentic Assessment: An evaluation method that requires students to apply their knowledge and skills in real-world, meaningful contexts, reflecting actual tasks they might encounter in their professional or personal lives.
Multimodal: In education, this refers to the use of multiple forms of media (e.g., text, images, audio, video) to enhance learning and communication.
Dialogic: Learning approaches that emphasise open, interactive, and collaborative communication, allowing learners to engage in meaningful conversations and discussions.
Generative Artificial Intelligence: AI systems capable of generating new, original content or responses, as opposed to simply processing and providing predefined information.
Generative Adversarial Networks (GANs): A machine learning framework involving two neural networks, a generator and a discriminator, that work together in a competitive manner to create, refine, and evaluate synthetic data.