Artificial intelligence:
technology applied to the
investigation of the authenticity of Modigliani's works

Artificial intelligence

Artificial intelligence methodology is based on quantifying the characteristics of individual traits in drawings and comparing them with the distinctive traits of other artists, using techniques of statistical inference and machine learning. Such empirical methodology can classify individual traits with an accuracy of 70% - 90% and for drawings with an accuracy of more than 80%, reaching an even higher percentage. The Amedeo Modigliani Foundation has strongly desired and encouraged the research for the realization of this tool to enhance its activity of study and scientific investigation of the artistic context linked to the figure of one of the pivotal authors of the twentieth century, investing in innovation and, wisely, on the authenticity of the mass of works that includes it. The idea is to equip the scientific community with a universal language capable of enriching and communicating truthfulness to the user. This project of artificial intelligence is able to obscure what emerged from erroneous appraisals that have led to declare authentic many false works of the artist. Where the human eye has failed to reach, there is today the contribution of a system of practices that involve a combination of algorithms attesting to the originality of the work. The authentication of artworks is a stylistic analysis and an operation traditionally carried out by expert figures, and that has very ancient roots; we remember the Morelli method, dating back to the nineteenth century, based on the study of the human body. The twentieth century, however, sees the development of the Van Dantzig method, based on the examination of the trait that has its own shape, tone and length.

Fondazione
Amedeo Modigliani

Over time, variations have been made in this approach to art, in the recognition of truth, always with the sole objective of annihilating wrong expertise and hindering the parallel work of counterfeiters. With the contribution of artificial intelligence adopted by the Amedeo Modigliani Foundation of scientific research, we start from the methodology launched in the twentieth century, expanding it and implementing methods for the quantification of the traits focusing on the same concepts suitable to be quantified by the machine. A new algorithm is proposed for the segmentation of the individual strokes, the type of brushstroke on the canvas, the pressure of the brush in the work phase, the drafting of the color and the distinctive feature of the artist. It will be created, as a result of data acquired, neural networks, a computational model, database that includes exclusively recognized collections, of certain origin and museum works. The intent is to achieve the achievement of absolute values of authenticity, able to put a halt to the market of counterfeits that concerns the work of Amedeo Modigliani.