Generative AI Tools in the Classroom: Beyond Talks

Artificial Intelligence (AI) has been profound in the last five years. Generative AI (GenAI) has been one of the most utilized functions within the AI realm. Large Language Model (LLM) such as ChatGPT now has a feature called text-to-image that is able to generate images only from text, known as a prompt. The text-to-image models facilitate various downstream tasks, including style transfer, personalization, semantic editing, image restoration, image enhancement [1]. This rapid transformation has also made it necessary for conventional education approaches to be reshaped [2]. Generative AI tools now sit inside the ordinary classroom and creative production in the university/college. The collaboration between human and machine intelligence, enables students to achieve more using these GenAI tools, such as Meta AI, Grok (X), and Dall-E among others. Adobe 2025 Creativity with AI in Education Report notes integrating creativity with AI is not just a supplement for class activities, it is essential instructional practice and learned skill for empowering the next generation to succeed [3]. Furthermore, Professors acknowledged AI’s value in bridging the digital divide and enhancing education despite the numerous challenges and deficiencies [4]. 91% educators believe that creative AI fosters deeper learning by enabling students to examine topics from various perspectives and integrating it into different disciplines [3]. Art and education has been scrutinized a lot since the rise of artificial intelligence. Unlike other disciplines where the use of AI is for analytical purposes, Art and Education adopts AI for the creative purpose. The use of GenAI is more than mere assisting students, but it can create and produce the artwork itself. A composition of image, text, illustrations, color; they can be easily created by using a good command (prompt). As in students’ perspective, students may miss out on deeper learning experiences and lack the confidence that comes from navigating to find their creative voice [5]. However, introducing GenAI as a potentially valuable technology tool that links students to current industry demands [5]. This shows how the industry has shifted to find a creative designer who is not only able to have a keen ‘eye of design’ but also understand how to use GenAI within their creative workflow. The gap identified is that educators want to incorporate GenAI tools within the curriculum while students are reluctant to use GenAI for creative purposes. Reconciling these two contrasting viewpoints can be a further discussion within the education realm. This pilot study aims to answer the gap using a pilot classroom study setting to examine how art and design students describe their experience using GenAI tools, such as Meta AI, Grok, and Dall-E across a structured image-generation task and drawing on students' own reflections.

This pilot classroom study setting adopts a qualitative-dominant, mixed-methodology design. Students’ open ended reflections were used as a primary data source. It will be analyzed using inductive thematic analysis. The inductive thematic analysis approach means the themes identified are strongly linked to the data [6]. The inductive analysis is therefore a process of coding the data without trying to fit it into a pre-existing coding frame, or the researcher’s analytic preconceptions [6]. The steps involve: data familiarization; keyword identification; code selection; theme development; conceptualization through the interpretation of keywords, codes, and themes; and, finally, the development of a conceptual model [7]. This research design is similar to the pattern established in comparable studies of AI in arts education titled “Artificial intelligence (AI) -integrated educational applications and college students’ creativity and academic emotions: students and teachers’ perceptions and attitudes,”[8] and “An Innovative Approach in Arts Education: Student Experiences of Abstract Art Practices Supported by Generative Artificial Intelligence” [2].

Seven undergraduate students from Indonesia enrolled in the visual communication design. All of the participants were aged 20-22 years old, consisting of 2 male and 5 females. The pilot classroom study setting research was part of a classroom activity therefore consistent with authentic-classroom-activity design also used in related research on generative AI in language and design education [9].

The activity is titled “Playing with AI Image Generation: Exploring the Possibilities of Promptings.” Students were instructed to: 1. Generate a minimum of 3 prompts using free tools such as Meta AI, Grok (X), and Dall-E with the option of using other GenAI tools such as MidJourney or Stable Diffusion; 2. Record their process using provided template; 3. Evaluate the image generated using rubric provided; the rubric scores were collected but fall outside this paper’s focus on student’s experience. 4. Submit a personal reflection on the process. During the activity, all seven students completed the task using Meta AI, Grok (X), and Dall-E; One student additionally substituted Leonardo AI in the place of suggested tools. Each student working from a same-shared researcher provided prompts. Six prompts split evenly between two prompts categories: 1. Specific/representational prompts (n = 3): scene-based prompts specifying concrete elements, setting, and stylistic constraints (e.g., time of day, color palette, thematic elements). 2. Abstract prompts (n = 3): concept-driven prompts describing mood, symbolism, or non-representational visual ideas (e.g., "a swirl of energy representing the passage of time"). Giving the same prompts allows researchers to compare the result while also maintaining flexibility and creativity for the students to use words while prompting.

The thematic analysis of the students' reflections summarised below with quotations from the students. All seven students reported a form of surprise in their responses. Six students described it as a positive surprise at unexpectedly high qualities the AI can produce. In contrast, only one student showed how poorly one AI tool performed. This contrasting surprise, highlighted by students, shows how AI tools can bring high expectations or low expectations in their execution. “(the thing that surprise me the most is)...how AI can adapt to what is in my imagination, and also can provide emotions in an image,” participant ID: 01CA. “I am quite surprised how AI can generate images so fast and give us some choices,” participant ID: 02MN. “AI has significantly improved since the last time I tried it, which was around mid-2024. Back then, AI mostly focused on generating realistic photos, but now some tools have shown significant progress in producing a more stylized visual outputs.” wrote participant ID: 03LN “The general fact that it was able to generate an image from a sentence (as well as animate it) and the quality of image that it was able to generate,” wrote participant 04ML. Contrastingly, only one student mentioned a different perspective regarding the use of AI as a tool. “How badly grok's performance was,” wrote participant ID: 06NV without mentioning what makes the AI tool bad compared to other AI tools.

Six of the seven students identified AI value early within the creative process. The students found AI usefulness in ideation, brainstorming process, and reference gathering rather than in production or final artworks/outputs. Students described AI as a tool to explore initial ideas, gather references effectively, generate concept images or otherwise-unfindable stock images. Here are some excerpts from the students. “AI can facilitate the exploration of initial ideas and also make my work time more efficient. AI will be very useful for me as a designer, it can provide new information and ideas, shorten my working time and help me when I experience artblock.” wrote participant ID: 01CA. “(AI will help the most) in gathering information and references, because we do not have to gather some data from another media to another media. AI can simplify our research by gathering information for design.” wrote participant 02MN. This statement is echoed by other students (participant ID 03LN) who wrote, ““However, from a technical standpoint, it has great potential as an instant visualization tool for the ideation phase as it clearly could help save time, money, and energy during the early stages of production.” Another student thinks that AI will help her the most and save time during texturing and creating background. “(AI will help the most).. Generating images for texturing, backgrounds. It saves time.” write participant ID 07YT. However she also stated in her closing statement, “Generative AI is cool and helpful for artists, especially Dall-E that might be able to help with layouting, brainstorming, etc :3” Layouting and brainstorming are mainly located within the early stage of the creative process.

Three students reported that vague prompts will produce poor image results. However, an explicit and detailed prompt will produce better image results. Prompting was seen as a necessary skill to produce better image generation. This prompting-skill set was needed in order to make the image desired by the students (designer). Changes in prompts can lead to significantly different outputs [10]. “ AI still needs a very detailed prompt to get our wanted result.” participant ID 02MN, clearly stated that a detailed prompt is needed to create a result that the designer wants. “I will make it pay closer attention to the details of the input to increase its accuracy.” write participant ID 05HT. The word input in her statement refers to the way she plans to do the prompting. Participant 01CA also stated a similar principle regarding detailed prompt, “when making a prompt I also have to make in detail what I want so that there is no misunderstanding or wrong results.” A detail prompt will make the AI tool have better and detailed images. Prompts mediate the conversation between humans and computers, therefore prompt engineering has received extensive attention [11]. In conclusion, prompting is seen as a necessary skill by the students to produce better results.

Perhaps, this is the most important theme within this classroom study setting research. Two students proactively framed AI as incapable of replacing a designer. This skepticism happens because AI can not convey the genuine feeling in an image hence positioning it as a tool of help rather than taking over the creative process. “AI still has some limitations but it is enough to help us. The differentiation between designer and AI is that AI can not express the feelings of their art. So it can be summarized that AI is not for taking over a designer's job, but it is about helping a designer's job.” stated by participant ID 02MN. Here, the student placed AI as assistant rather than replacing designers. “.. I don't think it would directly improve my work as AI tends to miss some of the fundamentals of art.” participant ID 04ML. What the student said confirms how she will use AI in the early stages of the creative process and yet to replace designers as AI tools is still missing the fundamental principle of art and design.

Two students reported ethical concern in regards to AI image generation especially regarding ownership, artistic labour, and the lack of clear regulations. They are referencing the same public controversy around the AI-generated “Ghibli-style” image trend. One student (participant ID 06NV), only referenced the same controversy but expressed bluntly without elaborating details. “I hate Ghibli AI images.” “Until now, I have never used AI in my work,.. due to both lack of quality and ethical reasons,” wrote participant ID 03LN. Furthermore, she explained, “ I believe using it currently violates key ethical principles since it also does not have a clear regulation yet.” “The creation of the AI tool that can generate images from a prompt may be commendable, but the use of AI generative images for professional work needs to be questioned.” wrote participant ID04ML. Both of the students raise an ethical unease about the use of AI in their works.

The rapid development of emerging technologies such as GenAI is not merely about completing tasks but also active agents that involve within the creative process itself [12]. The framework of Technology Acceptance Model (TAM) has been long used to understand technology usage intention with perceived usefulness and perceived ease of use as its center core determinant [12]. In this study, the perceived usefulness of GenAI tools, seen by the students on theme 1. Most of the students, six from seven students; agree with the usefulness of GenAI tools located in the beginning of the creative process. The interactive nature of GenAI, which works like a chatbot, has helped students to use prompting in a detailed manner. This kind of interactive process promotes active participation and hands-on learning for the students, hence improving student engagement [10]. A positive surprise reported by six students regarding the quality of the image generated by GenAI tools is aligned with research by Hwang, et.al (2024). The presence of high ‘happiness’ and ‘surprise’ scores underscores the excitement and satisfaction students experienced when achieving their intended visual outcomes [9]. The novelty of AI-generated images mirrors historical shifts in education, such as the introduction of multimedia tools, which leveraged interactivity to trigger curiosity [10]. Authorship and ethical unease still underlying student’s view towards GenAi tools. The Ghibli-style trend is the highlight mentioned by the students. The students' concern resonated with the presence of ‘sadness’ and ‘fear’ highlighted by Hwang et al. about AI’s growing role in content creation. The fear of overreliance on AI and the potential erosion of human artistry [9]. These findings support treating art and design student’s response to GenAI as multidimensional rather than total acceptance or resisting the idea of using GenAI as a tool. Broader dimensions of AI tools used, prompting skills, AI usefulness within the creative process, and ethical unease were shown within this study.

This study examined how art and design students used GenAi tools through a pilot classroom study setting. There are several findings worth noting. First, students consistently placed AI value as an early tool used within the design process. The usefulness of AI lies within ideation, brainstorming process, and reference gathering rather than in production or final outputs. Second, the students' acceptance of AI tools do not fall into one single acceptance nor rejection. There are multidimensional issues regarding their acceptance. As a pilot study, this study is limited by its small sample and location. Therefore, this study can not be generalised for the whole sample of art and design students. As a pilot project, further research can be enhanced based on this study. Future work could extend more towards bigger samples, using more recent AI tools, exploring teachers/lecturers' perspective, and using different methodologies to investigate the phenomenon.