Visibility of resource utilization : influencing user behaviour when interacting with AI systems

Abstract

Artificial Intelligence tools and models have reached a point where they have become an integral tool in the daily lives of students and IT professionals. Simplifying and optimizing users’ workloads, extracting and analysing large amounts of data, digital assistants, automating development, and automated code debugging are a few of several other reasons why it has gained such popularity. Although AI is associated with numerous benefits, over-reliance can lead to unsustainable use of AI tools. Especially, when the users are not aware of the underlying resource costs associated with their over-reliance. This research aims to understand and promote user awareness and attitudes towards the sustainable use of AI models by displaying resource costs associated with their choices. Design science research methodology was followed to create an artifact that presents resource utilization information, and it was evaluated by students and IT professionals. Quantitative analysis was conducted to understand how users choose AI models when there is no visibility of resource utilization versus when there is visibility. Qualitative thematic analysis is performed to understand how displaying resource costs affected their experiences and attitudes. A total of 38 participants which consisted of both students and working professionals participated in the experiment and the quantitative results show a 5.3% improvement in users preferring sustainable choice when resource utilization costs are made visible. In addition, the qualitative analysis reveals that users had clarity and felt easy to understand environmental cost. Users were also motivated to use AI more carefully and only when needed and choose sustainable options when possible. Users agreed that software providers should be transparent about the environmental costs, thereby promoting user awareness and sustainable software use.