Sensitive Pictures was an experience designed for the Munch Museum in Norway in which visitors experienced an emotional trajectory as they firt viewed Edvard Munch’s famously emotional paintings before self-reporting their emotional responses, engaging which computer vision technology that tried to assess their emotions, before finally being presented with their own emotional data as they left the museum with the intention of provoking them to make new meaning of their experience.

My subsequent CHI 2024 paper “Machine Learning Processes as Sources of Ambiguity”, co-authored with ITU Copenhagen, extends this thinking, revealing how artists employing generative AI to create visual art deliberately invoke various forms of ambiguity throughout the machine learning pipeline. It further challenges AI’s conventional understanding of uncertainty, dependability and explainability, calling for AI methodologies that embrace ambiguous artistic thinking.

Here’s an example of visualising visitors’ data as a series of emotion labels for different paintings within the museum.

And here’s an example of an emotional trajectory.

Papers

Sivertsen, C., et al, Machine Learning Processes As Sources of Ambiguity: Insights from AI Art. Accepted to appear in Proceedings of the 2024 ACM Conference of Computer-Human Interaction (CHI 2024), ACM. https://www.dropbox.com/scl/fi/scj6qd1x86aj07ex3fhvk/CHI24-ML-ambiguity.pdf?rlkey=5fa6hvshuh5dewedo2furogcc&dl=0  

Benford, S., et al, Sensitive Pictures: Emotional Interpretation in the Museum. Proceedings of  Proceedings of the 2022 ACM Conference of Computer-Human Interaction (CHI 2022), pp. 1-16). ACM. 2022. https://doi.org/10.1145/3491102.3502080