Abstract:

As traditional modes of continuous assessment (CA) become increasingly vulnerable to completion by generative artificial intelligence (genAI), this paper asks how we might retain some of the perceived benefits of CA while minimising validity threats. The discussion is informed by qualitative investigations of contemporary Higher Education learners’ assessment preferences conducted at an Irish university during the 2023-24 academic year as part of a broader research project on students’ perceptions of academic integrity. The study deployed mixed methods and comprised a survey and a series of seven focus groups. The data considered here stem from individual answers to two open-ended survey questions and co-constructed narratives on assessment from the focus groups.

The main thematic strands in our analysis reveal a general preference among participants for the assessment of process over product; for narrative guidelines on when, how, and to what extent GenAI can be used in the completion of an assessment; and for ‘practical’ (or authentic) assessment. In this, they broadly reflect the most prominent strands in the literature of assessment in the genAI era. However, we aim to elucidate students’ rationales for these preferences with a view to opening paths to future research on shared visions of valid assessment.

Perhaps the most notable strength of CA identified by our participants was its promotion of fairness in assessment. This concept of fairness was multidimensional, and related to creating multiple chances for students to accrue marks and to demonstrate a range of different skills; to minimising the negative impact of a ‘bad day’; to controllable and staged preparation; and to motivation to study regularly. These learning-oriented benefits are perhaps best captured by the case study of projects which – among all of the ‘traditional’ CA types discussed – are deemed most advantageous. In co-constructed narratives, in particular, students highlight their promotion of ‘authentic’ understanding and learning and their role as preparation for the workplace.

By contrast, however, individual accounts of the advantages and disadvantages of CA in the era of genAI were considerably less positive, perhaps revealing experience-based negative perceptions of particular instances of CA. Interestingly, these individual responses tend to posit continuous and terminal assessment as binary opposites, with CA being vulnerable to genAI-facilitated misconduct and terminal assessment being less vulnerable – or perhaps invulnerable – to it.

The contrast between socially mediated positive readings of the role of CA and individually constructed critiques of it animate our conclusions, which explore rationales and implementation strategies for research-based assessment which is not easily completed by genAI. Indeed, the authentic understanding, durable learning, and workplace preparation identified by our participants as its main strength are arguably more important components of assessment than they have ever been. However, we note that participants also noted scope for improvement of CA processes. We close with a brief discussion of how these might be balanced with the CA characteristics most valued by learner participants.

Keywords:

Continuous Assessment, Generative AI in Education, Academic Integrity, Assessment Design, Student Perceptions.