The article examines the multifaceted experiences of university teachers as they navigate the integration of generative artificial intelligence (GenAI) into their teaching practices. It begins by setting the context of contemporary challenges in higher education, where universities must prepare students for a rapidly evolving workplace that increasingly demands GenAI literacy. The study adopts a phenomenographic perspective to uncover the qualitatively different ways teachers perceive the value proposition of GenAI, their conceptions of how GenAI can support student learning, and their approaches to teaching with GenAI. Through in-depth interviews with thirty university teachers across various disciplines, the research identifies distinct categories of teacher experiences, ranging from positive perceptions that emphasize GenAI's potential to deepen student learning to more guarded or pessimistic views that see GenAI as a hindrance. The article also explores the associations between these perceptions, conceptions, and approaches, revealing how they shape the overall teaching experience with GenAI. The findings have significant implications for educational leadership, teaching development, and student learning, highlighting the need for a nuanced understanding of GenAI's role in higher education. The study concludes by discussing the potential benefits and challenges of integrating GenAI into the university learning environment, offering insights into how educators can leverage this technology to enhance student outcomes and prepare graduates for the future workplace.
AI Generated
This summary of the content was generated with the help of AI.
Abstract
Generative AI (GenAI) use is increasing across society in many different industries. Despite widespread adoption in workplaces, there is little consensus on the scope of its benefits and challenges at the level of most industries. Universities are being called upon to equip graduates with important knowledge and skills using GenAI for their professional contexts. Higher education, however, faces issues in effectively and sustainability embedding a use of GenAI in the student experience, which requires adjustments to learning and teaching activities, assessment, and learning outcomes and in access to useful GenAI platforms relevant to the various disciplines. As pedagogical models, ethical debates, and technologies continue to develop in this space, university teachers’ experiences of teaching with GenAI have yet to be explored in detail. Adopting a phenomenographic perspective, this study examines university teachers’ conceptions, perceptions, and approaches to using GenAI in teaching. Leveraging semi-structured interviews with 30 teaching academics, variations of teaching using GenAI were identified. Quantitative analysis was also conducted to capture the associations between these variations. By exploring the qualitative differences between these variations, a nuanced and important contribution to the GenAI discussion from the understanding of university teachers is uncovered. The results show that some ways of understanding and teaching with GenAI are more likely to help students develop effective knowledge and skills for the workplace than others. The findings also offer education leaders evidence to help design effective support for teachers using GenAI to innovate in the student experience. Through investigating the university teacher experience of GenAI, this research adds to the growing debate on the GenAI enabled benefits and challenges that are set to shape the higher education sector.
Notes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Introduction
Educating university students for meaningful personal growth and professional careers is a real contemporary challenge for universities because of expectations from individual students and the increasing complexity of the workplace (Chan & Hu, 2023). On the one hand, students tend to see their time at university first as a means of securing a career, not just as an experience of receiving an education. On the other, employers expect graduates to arrive ready at the workplace with contemporary knowledge and skills to engage effectively (Bae et al., 2022). Part of the challenge for universities in meeting these expectations is the changing opportunities for gainful employment as some jobs disappear and other jobs emerge, the skills of which require graduates to develop interdisciplinary views on problem solving, including a meaningful application of new technologies such as generative artificial intelligence (GenAI) across most industries and professions (Prohorov et al., 2024). These graduate workplace capabilities have fundamental implications for university teaching.
In the current climate, most industries are starting to require GenAI literacy as a key graduate attribute of university students (Nartey, 2024), which has a knock-on effect for curriculum design and teaching at university. While the following discusses the disruptions to university teaching and curriculum design brought about by the emergence of GenAI, the issues raised could be related to any fundamental technological disturbance shaping how knowledge is discovered, interpreted, disseminated and applied in professional and educational settings.
Advertisement
The focus of this paper is on qualitatively different teacher experiences of teaching with GenAI. It considers what might constitute an appropriate integration of GenAI into the university student learning experience from the perspective of university teachers. However, what constitutes an appropriate integration of GenAI in university student experiences of learning is a nuanced and contestable space. To put forward the teacher perspective, this paper reports on variations in what teachers perceive to be the value proposition of GenAI, the quality of conceptions that they hold about how GenAI can support student learning, and variations in how they approach teaching with GenAI. It begins with a consideration of the links between GenAI in industry and education, before presenting the framing of the research method to investigate variation in the quality of how teachers report GenAI in their experience of teaching.
Research and development of GenAI in industry and education
Developments in GenAI, particularly Large Language Models (LLM), have been progressively influencing the ways in which industries are adapting GenAI to solve challenges (Tomašev et al., 2020). While the development of industry-specific systems that use GenAI has been gradual, the introduction of LLMs into mainstream workplaces through ‘free’ and subscription services such as ChatGPT, Gemini or Copilot has significantly increased their uptake by businesses and their employees (Wu et al., 2023). Within the workplace, GenAI assistants are used as non-human elements of complex systems to elaborate and enhance the skills of employees. The significance of this type of mainstream introduction of GenAI into the workplace has seen shifts in workforce planning, with many areas starting to require specific GenAI literacy for employment (Javaid et al., 2023). Consequently, employees need to know how to both use and critically evaluate GenAI output (Farrell et al., 2021). As an important employability skill, there is growing expectation from industry that university graduates will have the necessary critical literacy skills that will enable them not only to determine if digital artefacts produced by GenAI are factually correct, but also to evaluate the output for bias or other infelicitous influences, and to judge the extent on which to base decisions with it (Ng et al., 2021). The significance of GenAI innovations for industry development is noticeable from the scale of its integration across multiple industries, leading to nationwide directives, such as those from the Office of the USA President, to guide the safe development and use of GenAI in the workplace (Biden, 2023).
Within the international education sector, the introduction of GenAI is being shaped in part by both governments and non-government organisations (UNESCO, 2023). For example, in the Australian higher education sector, federal government policy is requiring universities to demonstrate a productive use of GenAI in learning and teaching (TEQSA, 2024). The productive uses of GenAI, however, have not been amongst the first emphases by universities for the student experience. Early emphases on GenAI in the higher education sector have been on ethical issues in learning such as plagiarism and academic integrity. For example, Moorhouse et al. (2023) reviewed available university policies on GenAI use for assessment, finding that only limited guidance to its productive use was available, with advice focused primarily on limitations such as prevention and detection. Similarly, university policies have focused on the dangers of the use of GenAI as a study aid due to its tendency to hallucinate, fabricate or embellish information (Rudolph et al., 2023).
These challenges are only part of the picture in higher education, as GenAI is expected to bring about a growing number of benefits for students and teachers (Bozkurt et al., 2023). While GenAI may contribute to efficiency outcomes, there is perhaps greater potential for productive enhancements (Tlili et al., 2023). Now facing daily decisions regarding what GenAI could achieve for their students, teachers must also judge when it might impede student understanding of the subject matter involved. (Reiss, 2021). Teachers are embedding a meaningful use of GenAI in activity and assessment designs (Markauskaite et al., 2022), including helping students to understand the ethics underpinning GenAI tool design and use as part of GenAI literacy development across different disciplinary settings (Lodge et al., 2023). An interesting outcome of this new aspect of pedagogical practice is that designing GenAI-enabled learning and assessment strategies has sharpened teacher reflections on which education practice is uniquely human, as some processes can be enabled by GenAI while others cannot (Chan & Tsi, 2024).
Advertisement
From a pedagogical perspective, the potential and practical benefits of GenAI outweigh its challenges and there is general recognition that GenAI literacy for general and specific contexts needs to be a core part of university student experiences and part of their attributes on graduation (Fleischmann et al., 2024; Nartey, 2024). These observations highlight the need for university leaders to embed the goals of an appropriate integration of GenAI in the university learning environment and how to most effectively align its integration with the learning outcomes of students (Wang et al., 2024). Given that university teachers are responsible for the program and course learning outcomes of degrees, understanding the teacher perspective on the use of GenAI in learning and teaching will help all stakeholders to work towards these goals. Developing a nuanced understanding of the teacher perspective on GenAI for learning and teaching and its implications for the teaching experience is one of the key purposes of this paper.
A phenomenographic perspective on qualitative variations
This study adopted a phenomenographic perspective to understand qualitative variations of teachers’ experiences of teaching with GenAI. Phenomenography aims to understand “qualitatively different ways in which people experience, conceptualize, perceive, and understand various aspects of, and various phenomena in the world around them” (Marton, 1986, p. 31). One of the important underlying assumptions of the phenomenography is that people’s knowledge of a phenomenon can be situated within a limited number of categories, which can be understood as “logically structured in a nested hierarchy of inclusiveness” (Åkerlind, 2008, p. 636). This means that people’s knowledge can be hierarchically ranked from exhibiting a low to high level of sophistication, with less sophisticated knowledge embedded within more sophisticated knowledge (Durden, 2019; Han & Ellis, 2019; Taylor & Booth, 2015). The ontology behinds the phenomenographic perspective is that “an individual cannot experience without something being experienced” (Marton & Pang, 2008, p. 535). The phenomenographic perspective has been widely adopted to understand individuals’ cognitions about learning and teaching, such as perceptions of, conceptions of, and approaches to, learning and teaching.
Previous research into perceptions and conceptions of, and approaches to, learning and teaching
Sustained previous research into the student and teacher experience at university has been undertaken across a number of seminal research programs. One program is by Entwistle and colleagues (for example, Entwistle, 2013; Entwistle et al., 2014), another is by Marton and colleagues (Marton, 2014; Marton et al., 1993) and a third is by Prosser and Trigwell (Prosser & Trigwell, 1999; Trigwell & Prosser, 2020). While there are some differences amongst the programs of research, they each seek to identify qualitatively different experiences of learning and teaching and how relatively deeper experiences of learning by students are logically and structurally related to student-focused approaches to teaching that seek conceptual change, rather than approaches to teaching which focus on delivering information to students. In this study, we adopt the learning and teaching model of Trigwell and Prosser (2020) and look at the teaching experience in relation to GenAI. An analytical model of the experience of learning and teaching is presented in Fig. 1. It is an analytical model as the identification of key variables under each of the constructs allows for a phenomenographic treatment of them by looking at their structure and reference in order to identify qualitative differences. In practice, instances of learning and teaching and the interaction of the variables involved happen simultaneously. By identifying key parts of the experience and looking at them in a relational manner, their key parts (their structure) and the meaning of those parts (their reference) can be investigated.
Fig. 1.
3P model of teaching–learning experience (Prosser & Trigwell, 2017)
In Fig. 1, the teaching–learning experience can be understood to be comprised of two parts: the student experience of learning in the top half of the figure, and the teaching experience in the bottom half of the figure. The model conceptualises teaching and learning experiences as three constructs, Presage, Process and Product with academic context being shared. One of the goals of the underlying research program of this study is to uncover qualitatively different variations of student and teacher experiences of GenAI in learning and teaching processes, and associations amongst the variables across the constructs. For reasons of space, this study considers key aspects of GenAI in the teaching experience only.
A phenomenographic treatment of the variables in the 3P model can be explained by looking at their structure and reference (Marton & Booth, 1997). For example, researchers have investigated student conceptions of learning across a number of different disciplines and have proposed qualitatively different groupings of cohesive and fragmented conceptions. (Chiu et al., 2016; Kapucu, 2014; Limbu & Markauskaite, 2015). In terms of the reference or meaning of cohesive conceptions of learning, these do not separate the learning outcomes of the course and a deep understanding of the subject matter from the concept, they see them as inextricably entwined; and fragmented conceptions view learning as a disconnected list of things which have little relation to each other (Prosser & Trigwell, 1999; Teixeira et al., 2016). Structurally, cohesive conceptions of learning are related to higher order activities such as analysis, synthesis, and application, while fragmented conceptions of learning are related to reproduction, replication, and rote memorisation and learning.
Apart from studies on how students conceive learning, research has also been conducted on teacher conceptions of student learning. For instance, Prosser et al. (1994) interviewed 24 university science teachers about their conceptions of university students’ science learning. and identified two groups of qualitatively different conceptions of learning science. Referentially, three concepts viewed student learning as developing their understanding through experiences which stimulated conceptual development and change (cohesive), while two concepts viewed student learning more about accumulating information and knowledge without the development of in-depth meaning (fragmented). Structurally, the first group conceived of the development of personal meaning to students through their learning while the second group conceived of learning only when external reasons were requiring them to learn.
The same study investigated qualitatively difference in teachers’ approaches to teaching. In this part of the strategy, the investigation sought to understand teachers’ intentions and strategies in their teaching and to uncover if there were any helpful associations to the qualitatively different conceptions of learning which had been identified. Through analysis of in-depth interviews, the university science teachers reported five approaches to teaching science (Prosser et al., 1994):
an information transmission approach with an intention of transmitting information to students.
a teacher-focused approach with an intention for students to acquire concepts.
a student–teacher interaction approach with an intention for students to acquire concepts.
a student-focused approach aimed at developing students’ concepts.
a conceptual-change approach adopting a student-focused strategy aimed at changing students’ conceptions.
These categories of approaches to teaching reflected a relational hierarchy. Teachers adopted the first two approaches tend to make efforts to activate students’ existing conceptions and encourage students to construct their own understandings (student-focused/conceptual change approaches). The other three approaches focused predominantly on how to organise, structure, and present the course content in their teaching (teacher-focused/information transmission approaches). Prosser et al. (1994) further demonstrated that teacher conceptions of student learning and approaches to teaching were logical associated. Those holding fragmented conceptions of student learning tended to adopt teacher-focused/information transmission approaches; whereas there was a tendency for teachers having cohesive conceptions to use student-focused/conceptual change approaches.
Compared with studies on conceptions and approaches in relation to teaching, there is relatively little research on teachers’ perceptions of learning and teaching with the exception of Prosser and Trigwell (1997). This study examined how teachers perceived their teaching environment might be associated with their approaches to teaching. Teachers’ perceptions were measured on the five aspects: control of teaching, appropriate class size, enabling student characteristics, departmental support for teaching, appropriate academic workload. The results suggested that teachers who adopted.
teacher-focused/information transmission approaches to teaching tended to perceive their class sizes to be too large (negative perceptions); whereas those who adopted student-focused/conceptual change approaches were more likely to perceive the appropriateness of class sizes and favourable support from department (positive perceptions).
These relations of teacher experiences in traditional classroom teaching might not be replicated in the contemporary university teaching context, which teachers may constantly experiment and integrate emerging technologies such as GenAI in teaching practice. Hence, it is timely to examine the associations amongst these constructs in teacher experience of teaching with GenAI.
The current study and research questions
In contemporary university teaching, specifically in the current emerging debate about if and how to incorporate GenAI in the learning and teaching processes, understanding variations in how university teachers experience GenAI is likely to provide insight into variations of teachers’ intentions and strategies which both assist and impede a productive use of the technology for student learning. Without understanding what teachers think and do with GenAI, it will be difficult to support their developmental needs so that they, and the sector more broadly, can discover disciplinary appropriate ways of embedding GenAI in the student experience to develop useful capabilities, knowledge, and skills.
In the following, we focus on teachers’ experience of teaching with GenAI to look for any differences in their perceptions, conceptions, and approaches that will help reveal strategies and intentions that help, or hinder, an effective use of GenAI. Our hypothesised relations between university teachers’ conceptions, perceptions, and approaches in teaching with GenAI are displayed in Fig. 2.
Fig. 2
The hypothesised relations between university teachers’ perceptions, conceptions, and approaches in the context of teaching with GenAI
The overarching research question guiding this study is: What are university teachers’ qualitatively different experiences of teaching with GenAI?
This question was further broken down into four research questions:
1)
What are variations of university teachers’ perceptions of the value proposition of GenAI in higher education learning?
2)
What are variations of university teacher conceptions of GenAI for higher education learning?
3)
What are variations of university teachers’ approaches to student verification of GenAI output?
4)
To what extent are perceptions, conceptions, and approaches in teacher experience of GenAI associated?
Method
Participants
Thirty university teachers voluntarily participated in the study. Efforts were made to recruit participants from a diverse range of academic disciplines, as previous research has reported that there are disciplinary specific features of teacher perceptions, conceptions, and approaches variables (Biggs et al., 2022; Lonka et al., 2004; Nelson Laird et al., 2014). Specifically, we used Becher–Biglan typology (Biglan, 1973; Neumann et al., 2002) to recruit participants from the four broad categories of disciplines: namely hard pure disciplines (e.g., mathematics, chemistry, and statistics), hard applied disciplines (e.g., design, pharmacy, and medical sciences), soft pure disciplines (e.g., history, music, and art), and soft applied disciplines (e.g., education, law, and social work). The level of their teaching also varied with a mixture of first, second and third year levels. Table 1 provides a detailed information of the participants’ specific disciplines and levels of teaching.
Table 1
Participants’ disciplines and levels of teaching
Level 1
Level 2
Level 3
Soft pure disciplines
3
1
–
Soft applied disciplines
1
7
1
Hard pure disciplines
5
3
–
Hard applied disciplines
2
1
4
The participants had a variety of experiences of GenAI platforms, from experimenting for educational use to embedding it into program design. They tended to use more than one platform and had engaged with commonly available platforms available at the time of the data collection, including ChatGPT, Microsoft CoPilot, Claude, Gemini, Llama, DALL.E and Sora. The majority of these are text-based platforms, with the last two involving image generation and video generation respectively. While the disciplinary spread and experiences of GenAI platforms varied across the population sample, the total number of cases and frequencies in categories were too small to identify any noteworthy patterns. Seeking predictive patterns amongst participant experiences, platforms and disciplinary boundaries will be investigated in future studies with larger population samples.
Research design and methods
We adopted a mixed-methods research design using the phenomenographic method. Phenomenography offers helpful potential to combine ideographic qualitative analyses (thick descriptions) to identify variation at the level of individuals, with quantitative analyses which seek nomothetic principles which aim to identify transferable principles (Feldon & Tofel-Grehl, 2018). A distinguishing feature of a phenomenographic approach is the development of outcome spaces, which are structured representations of the different ways participants experience phenomena. The categories of description are based on analyses of individual experiences of a particular phenomenon (a type of ideographic analysis), and their groupings into categories enable nomothetic interpretations of the experiences, which aim to identify the range of experiences of a phenomenon and facilitate quantitative treatments of the associations amongst the categories (Feldon & Tofel-Grehl, 2018; Prosser & Trigwell, 1999).
In this study, three outcome spaces are produced looking at variations in: teacher perceptions of the value proposition of GenAI for learning, teacher conceptions of GenAI in learning and teacher approaches to student verification of GenAI output. The statistical significance and strength of associations amongst the categories of qualitatively different descriptions across the three outcome spaces are investigated using cross-tabulation.
Data collection
Before the data collection, ethics approval was sought from the researchers’ institutions. Signed written consent were obtained from all the participants. Semi-structured interviews were employed to elicit university teachers’ experiences of GenAI in teaching and learning processes.
The interviews were open and loosely structured, which allowed the interviewees to have freedom to determine the nature of their responses, and to have ample time to reflect upon and to elaborate as desired.
The three main prompts used to answer the first three research questions were:
“How do you perceive the value proposition of GenAI in higher education learning?”
“What is your understanding of GenAI in higher education learning?”
“To what extent do you require students to verify the accuracy of AI output? How and why?”
Probing questions were used to ask interviewees to expand on their ideas and clarify their thoughts (e.g., “Could you please elaborate…further?” “What do you mean by…?”). The interviews were audio-recorded and transcribed for the analyses.
Data analysis
Data analyses were conducted in two stages: phenomenographic qualitative data analysis and quantitative analysis.
Phenomenographic data analysis
In the first stage, we used the phenomenographic data analysis to generate qualitatively different categories of perceptions, conceptions, and approaches, which offer a rich description of qualitative variations in the teaching experience with GenAI. The generated categories were used to answer the first three research questions. We followed the analytic approach widely adopted in phenomenographic research (Åkerlind, 2018; Marton & Pong, 2005; Yates et al., 2012). In the initial step, the transcripts were read thoroughly and the responses to the questions about perceptions, conceptions, and approaches were separated and collated. Data analysis was carried out separately for each question and described as follows. In the first round of reading, we read the responses repeatedly to determine the qualitative breadth and depth of all the responses. In the second round of reading, we located and highlighted the key features in the perceptions, conceptions, or approaches in each teacher’s response. By focusing on the key features, patterns of responses were more easily identified, which allowed us to identify initial themes that captured the dominant qualitative variations of perceptions, conceptions, and approaches.
We then compared and contrasted similarities and differences in the themes, generating a defined initial set of categories. Through an iterative process, we modified the definitions of categories so that each stood distinctively from other categories. When clarifying each category, we strove to use language which was able to represent the essence of its defining features. In the process of generating and labeling categories, we discussed frequently until agreements were reached on the numbers, names, and definitions of the categories. Using the agreed categories, we each coded the data individually. To establish the trustworthiness of coding, we calculated inter-coder reliability by computing percentage of agreement of assigning data using generated categories by different researchers (Säljö, 1988). Table 2 shows the percentage agreement of two of the researchers with the third researcher before and after consultations, which met the requirement of 80% (Säljö, 1988).
Table 2
The inter-rater agreement
Researcher1
Perceptions of value proposition of GenAI
Conceptions of GenAI for higher education learning
Approaches to student verification of GenAI output
% before consultation
% after consultation
% before consultation
% after consultation
% before consultation
% after consultation
researcher2
90
100
90
100
90
90
researcher3
90
100
70
80
80
90
Quantitative data analysis
We conducted cross-tabulations to examine the associations between perceptions, conceptions, and approaches, which provided the answer for research question 4. We used p-values to examine if the associations between variables were statistically significant. We computed Phi to evaluate the strength of association.
Results
Following the analytical methods described in the previous section, the outcome spaces for teaching experience of GenAI in terms of perceptions, conceptions, and approaches are presented in the following. In discussing the categories, we followed Marton and Booth (1997) by considering the structural and referential aspects.
Results of research question 1–Variations of university teacher perceptions of the value proposition of GenAI in higher education learning
Table 3 presents the results variations of university teacher perceptions of the value proposition of GenAI in higher education learning. Four categories of teacher perceptions were identified.
Table 3
Variations of teacher perceptions of the value proposition of GenAI in higher education learning
Category
Description
Illuminative quotation
A (positive)
GenAI helps to deepen student learning. (n = 5, 17%)
I would like to think that the Gen AI could ultimately lead to a deeper understanding of concepts and theory within the students. So by reflecting on the outputs that the Gen AI is delivering and then having that formative process with academics where we interrogated it with them, they are then able to develop a much deeper appreciation of the material and the context to have a much more and maybe than you know
B (positive)
GenAI plays a conditional role in student learning. (n = 7, 13%)
The potential for if it’s used appropriately and it’s used in a way that teaches the students something, I can see a carefully constructed assessment item like the one that I’m hoping to use in my neurobiology course could help them to learn more because they’re having to evaluate the information that they’re getting back, not just take it at face value. But it would have to be constructed in a way that there was value in that for the students. So some aspect of the assessment would have to reflect that
C (negative)
GenAI may benefit for learning, which depends on teacher’s quality. (n = 15, 50%)
I would say there’s a possibility it [GenAI] could, but it would still have to be the input of the generation of it all. The question around it is the quality of the depth of the educator, I think, in the first place
D (negative)
GenAI impedes learning experiences. (n = 3, 10%)
No one’s learning anything, OK, it’s just a totally useless, futile exercise, all right? And that is a very real risk. So that is, that is the negative value proposition of universities
Category A (n = 5, 17%) perceived GenAI to enable students to develop a deeper understanding through reflection. Perceptions in category B (n = 7, 23%) recognised the evaluative benefits of GenAI for learning but suggested a conditional context for their effective use related to carefully designed activities. Broadly speaking, categories A and B were positive perceptions.
In contrast, perceptions C and D were negative perceptions as they were guarded or pessimistic in their assessment of the role of GenAI in higher education learning. Perceptions in category C (n = 15, 50%) recognised there might be a benefit, but that it was dependent on the quality of the teacher rather than what the students might bring to the process. Category D (n = 3, 10%) was the most negative amongst the four categories, suggesting that GenAI would not help student learning at all.
The reference and structural aspects of the perceptions categories are summarised in Table 4. Referentially, categories A and B recognise the potential value of GenAI in contributing to the development of student understanding, whereas categories C and D do not see such value. Structurally, there is a qualitative shift between the pairs of categories A/B and C/D. Categories A and B reveal a clear awareness of how the careful evaluation of GenAI output in relation to other sources and course learning outcomes can enable student understanding. In contrast, categories C and D do not adopt a student perspective when considering the potential of GenAI to enable student understanding. This may account for why perceptions C and D separate a use of GenAI from any potential to enable learning.
Table 4
Referential and structural aspects of teacher perceptions of value proposition of GenAI in higher education learning
Referential
Structural
Enabling learning (Positive)
Not enabling learning (Negative)
Reflective role for learning benefits
Category A
Evaluative role for learning benefits
Category B
Not much learning value
Category C
Impedes learning
Category D
Results of research question 2 – Variations of university teacher conceptions of GenAI for higher education learning
Table 5 displays variations of teacher conceptions of GenAI for higher education learning. From teachers’ responses, five categories were identified. Category A (n = 5, 17%) recognises that GenAI has the potential to help students develop their analytic thinking through crafting searching and asking meaningful questions. Category B (n = 6, 20%) recognises the potential of GenAI to help develop essential knowledge and skills for a fast-moving future. Category C (n = 5, 17%) considers a role of GenAI in teaching, it but limits its contribution to checking and risk management activities such as rethinking assessment strategies. Teachers holding category D (n = 5, 17%) conceptions are hesitant to use GenAI because of a lack of trust in the technology platform, which fragments GenAI from supporting learning. Category E (n = 9, 29%) separates GenAI from students learning and understanding as it conceives of it as a machine or a type of mechanical reproductive element that simply adds things together.
Table 5
Variations of teacher conceptions of GenAI for higher education learning
Category
Description
Illuminative quotation
A (cohesive)
GenAI can help students develop analytical thinking in context. (n = 5, 17%)
I see it as one of numerous resources that students … can draw on to think about things in new ways, to better understand some of the complexities of, you know, the emerging digital space, because I think one of the things we forget is that, yes, you can use AI to, you know, ask questions of big data or to say test out statistical models or to help you, right? But it can also actually provide you with answers to questions that you might have. Where can I look for X?
B (cohesive)
GenAI can enhance the learning of students for their future. (n = 6, 20%)
But it’s not so much about stopping them from cheating as to how they can actually use it in a way that enhances their learning, even if it’s for assessment. So it’s just a flipping of the way that we think about it, to not do so. I feel like maybe ethically is better than moral is a better word, but we’re not. And we would actually be doing our students a disservice in the same way that we had to embrace it and have to embrace all sorts of technologies to prepare our students for the future. It’s just another step in that process. A very fast moving one
C (fragmented)
GenAI is limited to risk mitigation processes of teaching. (n = 5, 17%)
What I think AI is useful for in higher education is a checker, so, to really audit your teaching and learning processes, because if AI can write your assignment then it’s not rigorous. You know it’s not higher-level thinking. There’s, you know, where we’ve been so used to really assessing those lower order recall skills, even in higher education, but only now we’re kind of caught out. So, it’s only now that people are finding value in connection or evaluating or you know and reflection. So, it makes ….it forces teachers to rethink assessment processes
D (fragmented)
GenAI is not trustworthy for learning and teaching outcomes. (n = 5, 17%)
I see that as also having an impact on what we do in terms of a lot of the work that I do could be done through you know, something like ChatGPT. In terms of preparation, but ultimately, it’s still going to need a human being to course, convene and interact with the students, but it would save me a lot of time in terms of preparation.…….I’ve been hesitant to use it and reluctant for it to learn from me. I’m probably paranoid, but I don’t want it to learn from me
E (fragmented)
GenAI is a tool that provides little help for learning. (n = 9, 29%)
So when I think of generative AI, I really think of them as confection machines that they are simply, they’re sampling machines, you know, they’re pulling different sources, different tags and then confecting them back together into new forms. My experience with ChatGPT is that it’s very, very poor. It gives very, very poor results. I understand there might be better products out there in ChatGPT and with Dolly and Midjourney
Table 6 summarises referential and structural aspects of the conceptions categories. From a referential perspective, categories A and B conceive how GenAI can enhance learning through knowledge and skill development; while categories C, D, and E associate GenAI with risk management in teaching, untrustworthiness, and merely tools, which fragment GenAI from learning. Structurally, there is a qualitative shift between categories A, B and C, D, E. Categories A and B conceive of a link between GenAI and learning and thinking. In contrast, categories C, D, and E separate GenAI from related concepts about learning and thinking. Concepts in these three categories are overshadowed by practical concerns rather than any ideas about learning and understanding.
Table 6
Referential and structural aspects of teacher conceptions of GenAI for higher education learning
Referential
Structural
Associated with learning and thinking (Coherent)
Unlinked with learning and thinking (Fragmented)
GenAI enhances contextualised analytical thinking
Category A
GenAI supports the development of future-orientated skills
Category B
GenAI requires risk mitigation
Category C
GenAI is not trustworthy
Category D
GenAI is only a tool
Category E
Results of research question 3–Variations of university teacher approaches to student verification of GenAI output
Table 7 presents five categories of teacher approaches to teaching with GenAI. Starting with category E (n = 7, 23%), approaches in this category generally dismiss it as unhelpful for learning in the course. In some courses the response is educationally defensible because the outcomes are predominantly about creating hands-on creative products which have little need for GenAI use, while others are more related to unfamiliarity with GenAI platforms and their benefits for learning. Approaches to teaching in category D (n = 6, 20%) limit instruction to informing students that GenAI output is potentially wrong every time it is used. Approaches in this category tend to be reductive in their intent and strategies around using GenAI. Approaches to teaching in category C (n = 3, 10%) focus more on teaching the importance of the values involved in GenAI, recognising that accuracy is a concern of GenAI output, but so too is the issue of appropriately recognising the work of others to which GenAI is referencing. Approaches to teaching in category B (n = 8, 23%) reveal an intent to use the assessment of GenAI output as a way to develop the critical thinking of students, while approaches to teaching in category A (n = 6, 20%) use GenAI output as stimulus to require students to develop their research skills in confirming or correcting GenAI output.
Table 7
Variations of teacher approaches to student verification of GenAI output
Category
Description
Illuminative quotation
A (learning focused)
To facilitate students to challenge reasoning behind GenAI outputs through research (n = 6, 20%)
What it requires students to do is then to weigh evidence, to think about bias, to think about sources… Taking what you get from generative AI and adding it to all of the other information, and for me it just reinforces the importance of breadth and depth and research skills that I wouldn’t rely on one journal article to write something. I would want to read quite a few things to get a sense of where things sit in the middle and what sits on either side
B (learning focused)
To facilitate students’ critical thinking when using GenAI outputs (n = 8, 23%)
I know that we have seen students at the start of chat GPT, who were submitting things and they actually had incorrect references, and the references that were incorrect… I think that that’s the skills that we scaffold up because our first-year students, who are only just being introduced to the concepts of generative AI, probably don’t have that critical thinking skill they need at that point
C (teaching focused)
To teach how to adopt a values-based approach to using GenAI output (n = 3, 10%)
In professional programmes, it’s not just about accuracy. I think it’s also around ethics for use. So I mean what’s, what’s ethical? What does ethical use look like? And if you’re using something that is basically ripped, things from other people’s work, you repackaged it
D (teaching focused)
To teach that GenAI can be wrong (n = 6, 20%)
it’s a learning tool, especially because the (goal for) first year first trimester students is to teach the students that just because AI says it’s so, doesn’t actually mean that it’s correct
E (teaching focused)
To not teach how to use GenAI in relation to the course (n = 7, 23%)
they have to really use 20 min of their time to, to synthesise a lot of information and that there’s nothing that they can plug into to, to ChatGPT or a platform like this to get an answer
Table 8 summarises the referential and structural aspects of the five categories, revealing a qualitative shift between categories A, B and C, D, E. The first two categories use GenAI in teaching from what it might offer for student learning while the last three categories approach a use of GenAI in teaching from the point of view of the teacher and show little awareness of student-centred thinking. In the first two categories, Category B recognised that the assessment of the output of GenAI for accuracy, ethical concerns and the quality of ideas would help students develop their critical thinking. Approaches to teaching in category A had a similar learning focused perspective as B, but went further and recognised the benefits of how the students’ analysis of GenAI output could challenge the underlying reasoning and the quality of the information in relation to other research resources. Categories C, D, E tend to approach a use of GenAI more from the perspective of the teacher than the student. Category C approaches to teaching with GenAI by considering an ethical approach to its use rather than from a perspective of learning, whereas categories D and E approach it predominantly as a tool that can be wrong.
Table 8
The referential and structural aspects of approaches to student verification of GenAI output
Referential
Structural
Aware of student learning (Learning-focused)
Unaware of student learning (Teaching-focused)
Teaching for student critical thinking and reasoning
Category A
Teaching for student critical thinking
Category B
Teaching for a values-based use of GenAI
Category C
Teaching for erroneous GenAI outputs
Category D
Avoid teaching using GenAI
Category E
Results of research question 4–associations amongst teacher perceptions, conceptions, and approaches
Tables 8, 9, 10 present the associations amongst the qualitatively different categories of teacher perceptions, conceptions, and approaches in their teaching experience with GenAI. Table 9 shows a statistically significant and moderate relation amongst the categories of teacher conceptions of GenAI for higher education learning and teacher perceptions of the value proposition of GenAI (φ = 0.590, p = 0.002). Of teachers holding coherent conceptions, a significantly higher proportion tended to report positive perceptions (69.23%) about the value proposition of GenAI for student learning than negative ones (11.76%). Conversely, teachers reporting fragmented conceptions were likely to report a negative perception of the value of GenAI (88.26%) than a positive one (30.77%).
Table 9
Association between conceptions and perceptions
Conceptions of GenAI
Perceptions of GenAI
Total
Positive (A&B)
Negative (C&D)
Coherent (A&B)
9
2
11
Fragmented (C,D&E)
4
15
19
Total
13
17
30
Table 10
Association between approaches and perceptions
Approaches to teaching with GenAI
Perceptions of GenAI
Total
Positive (A&B)
Negative (C&D)
Learning-focused approaches (A&B)
12
11
23
Teaching-focused approaches (C, D&E)
0
7
7
Total
12
18
30
Table 10 shows a statistically significant and moderate relation between approaches to teaching with GenAI and perceptions of GenAI (φ = 0.450, p = 0.024). Teachers whose approaches were learning-focused tended to perceive the benefits of GenAI for developing student understanding provided that the output is verified. In contrast, teaching-focused approaches were related to negative perceptions of value proposition of GenAI.
The relation between approaches to teaching with GenAI and conceptions of GenAI for higher education learning is presented in Table 11, which also shows statistically significant and moderate association (φ = 0.360, p = 0.029). Approaches to teaching that use of GenAI to develop critical thinking and reasoning amongst students tended to conceive of the value of GenAI from the student perspective for its contribution to learning. Conversely, approaches to teaching that were teaching-focused tended to fragment any learning-benefit from GenAI for the student experience.
Table 11
Association between approaches and conceptions
Approaches to teaching with GenAI
Conceptions of GenAI
Total
Coherent (A&B)
Fragmented (C, D&E)
Learning-focused approaches (A&B)
11
12
23
Teaching-focused approaches (C,D&E)
0
7
7
Total
11
19
30
Discussion and conclusion
To contribute to the emerging debate about the value of GenAI knowledge and skills for university students, this intensive study interviewed thirty teachers to explore variations in how they perceived, conceived of, and approached teaching with GenAI. The results are an interesting contribution to the current debate as they offer education leaders, teachers, and students an important perspective of teachers’ experience of teaching with GenAI and the extent to which it is valuable for a university education experience.
Variations of teacher experiences of teaching with GenAI shaped by perceptions, conceptions, and conceptions
The results of this study have noteworthy implications for those who believe that a university education should provide graduates with relevant GenAI knowledge and skills for their careers if they are going to be productive contributors to society and industry (Nartey, 2024; Prohorov et al., 2024). In the current climate, most industries are starting to require GenAI literacy as a key graduate attribute of university students (Javaid et al., 2023). Consequently, employees need to know how to both use and critically evaluate GenAI output (Farrell et al., 2021; Javaid et al., 2023; TEQSA, 2024). These graduate workplace capabilities have fundamental implications for university teaching.
Fig. 3 provides a visual summary of the variations in the teaching experience with GenAI
Fig. 3
A visual representation of qualitatively different experiences of teaching with GenAI
Amongst the teachers interviewed, an unfavourable experience emerged involving negative perceptions of GenAI for learning, involving doubt about its potential to help students develop understanding, to the point where some teachers were unlikely to use it; or if it was used, the approach emphasised risk management issues such as correcting errors and not engaging in plagiarism or other ethics-based infringements. While these last two issues need to be addressed in any student experience, these were threshold of concerns for a second view of GenAI in the learning and teaching experience, one that viewed it beyond just risk-management, and considered it from the perspective of students and its potential contribution to learning.
Another experience of GenAI in teaching from the interviews perceived it as offering evaluative and reflective experiences of learning for students, experiences that helped students to contextualise analytical thinking in order to develop the knowledge and skills they will require for their future careers. This was a learning-focused view of GenAI in the university student experience that emphasised its use to stimulate critical thinking and reasoning. This view valued the importance of embedding GenAI knowledge and skills in the student experience where appropriate so that degree programs continued to be, and were perceived to be, relevant in the current trend to use AI not just in efficient ways, but in ways that added to student understanding and learning.
Within these views of variation of teaching with GenAI, there was not an absolute categorisation of the participants involved, as the contingency tables showed that the associations amongst the qualitatively different categories were more nuanced. In other words, just because a participant who reported a fragmented concept of GenAI approached its use in a teaching-focused way, did not mean that all participants who reported fragmented concepts did the same. The qualitative variation revealed, however, does suggest a tendency amongst the categories of description which is useful for those concerned about supporting teaching development and the quality of future education, including how to most productively approach a use of GenAI in the university student experience. By addressing fragmented concepts, negative perceptions and teaching-focused approaches to become more aware of the student experience and needs of GenAI literacy, the potential for learning benefits from GenAI knowledge and skills for university students is more likely to be embraced by teachers. Interestingly, while the academic context of teacher experiences in teaching with GenAI have distinct features from those in traditional university teaching, the relational patterns amongst conceptions, perceptions, and approaches resemble those observed from teacher experiences in other academic contexts (Prosser & Trigwell, 1997; Prosser et al., 1994; Trigwell & Prosser, 2020). Our study added empirical evidence into the 3P model of teaching–learning experience (Prosser & Trigwell, 2017).
Implications for education leadership, teaching development and student learning
If teachers’ views on GenAI are an important part of the emerging debate because of their centrality in a university education, then it is clear from those concerned about the quality and relevance of degree programs and the university student experience that an appropriate use of GenAI should be part of university student experiences of learning (Bozkurt et al., 2023; Reiss, 2021; Tlili et al., 2023). The views of employers and educators reported at the beginning of this paper suggest that students will need GenAI knowledge and skills to participate effectively in the workplace and this capability is likely to become an important graduate attribute from a university education.
However, we are just at the beginning of understanding what an appropriate use of GenAI in the student experience might look like, requiring responses from teaching team leaders and university leaders. For leaders of teaching teams, it is clear that some considerable work on better understanding the perceived benefits of GenAI for learning will be necessary, including how it can support student learning and approaches to teaching. Where GenAI is regarded as an important part of the student experience and curriculum design, some considerable work is likely to be required to help teaching teams understand relatively more productive ways of thinking about and using GenAI in the student experience (Koeslag-Kreunen et al., 2018), ways that focus more on the learning that can arising from using it than on only focusing on its limitations.
For university leaders, the benefits arising from a use of GenAI in learning and teaching suggest it needs to become part of the learning environment, which will require significant planning and investment (Hoernig et al., 2024). This will include an appropriately governed and risk-managed approach to a systematic use of GenAI for learning and teaching across all faculties (Chiu, 2024). It will also include consideration of policy and infrastructure that enables sensitive university data and information to be ring-fenced from the frontier platforms capture of all data, as well as identity management so that students and teachers can have appropriate access to GenAI. To develop the capabilities of teaching teams, it will involve reconfiguring professional development programs and grants to highlight innovations in GenAI in teaching, along with other incentives such as promotion based on excellence in the use of GenAI for learning outcomes (Trigwell et al., 2012). Only a sustained focused by teaching staff on the contributions of GenAI for learning will reveal how it can contribute to improving the student experience with the right amount of disciplinary variation in its deployment.
While the results of this study offer an important perspective into the debate on GenAI in the university experience, the intensive nature of the investigation necessitated that relatively small population sample. To test the robustness of the conclusions suggested, the qualitatively different positions of the teachers’ perspective should be investigated further in replicated studies. Studies with large quantitative studies will help in this regard and allow the development of more extensive evidence to support the findings, such as the extent to which disciplinary variation might contribute to differences in the way teachers approach teaching using GenAI.
While this study is just a small piece of the puzzle the higher education sector is grappling with, only through similar and complementary studies will essential evidence be uncovered for effective teaching that is required to leverage significant learning benefit for students. No matter how pedagogy and expectations change, teachers continue to be responsible for the underlying program and course learning outcomes of the student experience, both in the way programs and are designed as well in how they are taught. As such, they are key contributors to the current debate on the emerging benefits for GenAI for learning in higher education.
Acknowledgements
Not available.
Declarations
Competing interests
The authors have no competing interests to declare.
Generative AI and AI-assisted technologies in the writing process
During the preparation of this work the authors did not use any generative AI and AI-assisted technologies.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
