AI IN RECONSTRUCTING LOST MODERNIST ARTWORKS

The disappearance of modernist artworks through the historical conflict, material degradation and partiality of archival regimes poses a continuous problem in terms of art history and cultural heritage conservation. It is a foregone conclusion that such works are difficult to reconstruct, since modernist aesthetics is focused on abstraction, fragmentation, and conceptual purpose over representational devotion. In this paper, the author suggests a reconstruction model based on AI that views artistic reconstruction as a probabilistic inference problem, as opposed to a deterministic recovery problem. The system is already based on the framework of representation learning, conditional generative modeling, uncertainty-aware sampling, and human-in-the-loop validation to produce several possible discovery hypotheses based on the existing evidence and style samples. The hierarchical style modeling strategy gives reconstruction possibilities at artist and movements levels, whereas explicit uncertainty signs circumvent overconfidence in interpretation. The framework is tested by using representative case studies which are partially documented, partially fragmented and entirely lost modernist artworks. The quantitative metrics of consistency, the analysis of perceptual styles, and uncertainty analysis and the qualitative evaluation process performed by experts prove that the reconstruction fidelity is proportional to the strength of evidence and that the uncertainty is explicitly magnified in the case of sparse constraints. The findings indicate the significance of AI as a supportive analysis tool that can aid the judgment of curators and art-historians but does not sacrifice the interpretative responsibility.