Trust in artificial intelligence to extract geospatial features for topographic databases

National mapping agencies (NMAs) produce topographic databases (TDBs), which provide structured, country-wide spatial datasets used in, for example, maps, spatial planning, and decision-making. Traditionally, TDB features have been manually digitised from remote sensing data, but recent advancements in artificial intelligence (AI), particularly deep learning, have shown potential for increasing automation in spatial data production. However, AI adoption raises concerns about trust among TDB users. Distrust could have adverse impacts on data acceptance, the reputation of NMAs, and the credibility of AI systems. We conducted a questionnaire for public-sector TDB users in Finland to investigate their concerns about, and trust in, utilising AI for TDB data production. The results suggest that maintaining a human component in AI-based workflows is critical for user trust. The data users expressed various concerns about AI, extending beyond data quality. The results highlight the importance of transparency and explainability of AI systems, the reputation of the data producer, and clear documentation of quality metrics to accompany the data. Future studies should expand to include respondents from the private sector and a diverse set of countries, as well as investigate new solutions for explaining AI models’ decision-making for non-experts.