In a hall with a hundred instruments, a symphony plays, but the conductor is hidden behind a screen. The scientist hears only the notes the instruments play and wants to understand who sets the rhythm. Scientific data resemble such an orchestra: we record instrument readings but do not see the hidden causes that drive them. MOSAIC learns from these recordings to find the conductors and explain which instruments they affect.

Previously, it was unclear how to transfer explanations from observed data to hidden causes without losing confidence in their correctness. MOSAIC solves this in two steps. First, it compresses the data and identifies hidden factors, guaranteeing that they are recovered uniquely. Then, for each hidden factor, it finds a small set of observed variables that it influences. This is like being able to say after listening to a symphony: "The conductor raises the baton, the flutes come in, and the violins stay silent."

The method has been tested on various systems. In an RNA molecule, MOSAIC found a factor related to a loop, matching a known mechanism. On the solar wind and climate, the method correctly identified the expected groups of variables in all 12 runs. Today, MOSAIC can be applied to space weather, climate monitoring, and chemical process safety. The symphony of complex systems becomes clearer when we know who is conducting.