• Andreas Harboe Salskov Ager
  • Martina Konecná
4. term, Information Science, Master (Master Programme)
Due to the recent deployment of large-scale AI models, various new generative AI models have been released in the fall of 2022 and spring of 2023 with the ability to learn from content and generate content. As there is no standard definition of Artificial Intelligence (AI), AI is mostly referred to as intelligent systems, possibly embedded in larger systems, with capabilities that make AI systems achieve complex goals. The rapid advancement of AI has an impact on human-computer interaction, including the design of information technology. When designing information technology, having a human-centered approach enables User Experience (UX) practitioners to design systems with an emphasis on human beings and their needs. The term UX is used as an umbrella term in system development in which UX practitioners perform activities related to user research, problem setting, design conceptualization, and testing.

Through a review of related work, two directions of existing literature were discovered: 1) designing AI systems, and 2) designing with AI. In the thesis, the focus is on the latter, as there was a research gap in studies on practices of applying AI to design information technology. Through an explanatory sequential mixed-method study combining quantitative and qualitative approaches, a survey (N=64) served as a preliminary exploration of the phenomena, identification of the best candidates, and informing the qualitative research. The qualitative approach (N=6) consisted of contextual inquiry and semi-structured interviews which enabled further exploration and understanding of the UX practitioners’ subjective experiences of applying AI in their work practices. This underlines the pragmatic stance of acquiring knowledge about the practical application of AI in UX-related work practices. To analyze the collected data, inductive thematic analysis was applied due to an interpretivist approach.

Based on a survey, we found that approximately half of the UX practitioners use AI systems across the design process. The findings show that UX practitioners use AI systems as a designerly tool to get inspiration, as a starting point, and for sparring purposes in stages mainly related to problem setting and design conceptualization. The most common AI application is generative AI which augments the UX practitioners’ abilities and supports them in solving trivial and tedious tasks to provide more time to focus on UX tasks with a higher level of importance and abstraction. The findings suggest that the perceived advantages of AI in our study align with related work in other domains.

The findings reveal that the generative AI output is merely used as inspiration rather than being used directly in the design. Due to ethical concerns of trust, biases, and lack of transparency, the output needs to be validated and/or edited before being directly incorporated into the design. In general, Visual generative AI is used to a low degree by UX practitioners, compared to text generative AI, because the overall user experience is not accounted for in the AI output. Based on the UX practitioners’ perceived challenges of using AI in the design process, the findings suggest that AI is not able to design autonomously in a human-centered approach because the AI output must be balanced with human creativity, intuition, presence, and empathy. This suggests that AI can support UXPs to perform HCD activities rather than replace them. However, the UXPs’ perceived challenges of AI might be false due to a lack of AI literacy, or because the UXPs applied AI systems that did not fit the purpose of their tasks.

A discussion of the findings reveals that AI’s efficiency enables an agile and iterative design process while reducing time, budget, and effort resources. However, the discussion of AI’s efficiency and automation of certain UX activities suggests a demotion of certain competencies of UX practitioners. In addition, there is a perception among UX practitioners that human-centered design without real users is not human-centered because AI can not account for the contextual and intangible in real-life situations of human beings. Overall, AI can support UX practitioners in line with existing UX practices, methods, and tools despite the perceived advantages and challenges of AI. Based on the findings and discussion, there is a need for further research on AI in relation to its ability to be creative, innovative, and empathic in a human-centered approach to determine if the UXPs’ perceived challenges of AI are valid.
Publication date1 Jun 2023
Number of pages105
ID: 532472656