The learning process of accessibility instrument developers: Testing the tools in planning practice

Many planning support tools have recently been developed aimed at measuring and modelling accessibility (Accessibility Instrument or AI). The main difficulty for tool developers is designing an AI that is at the same time technically rigorous and usable in practice. Measuring accessibility is indeed a complex task, and AI outputs are difficult to communicate to target end-users, in particular, because these users are professionals from several disciplines with different languages and areas of expertise, such as urban geographers, spatial planners, transport planners, and budgeting professionals. In addition to this, AI developers seem to have little awareness of the needs of AI end-users, which in turn tend to have limited ability for using these tools. Against this complex background, our research focuses on the viewpoint of AI developers, with two aims: (1) to provide insights into how AI developers perceive their tools and (2) to understand how their perceptions might change after testing their AI with end-users. With this in mind, an analysis of 15 case studies was performed: groups of end-users tested different AI in structured workshops. Before and after the workshops, two questionnaires explored the AI developers' perceptions on the tools and their usability. The paper demonstrates that the workshops with end-users were critical for developers to appreciate the importance of specific characteristics the tool should have, namely practical relevance, flexibility, and ease of use. The study provides evidence that AI developers were prone to change their perceptions about AI after interacting directly with end-users.