Linking microscopic and mean-field descriptions ofdopaminergic modulation in neural circuits

Next-generation neural mass models capture the collective behavior of brain regions through an exact mean-field reduction of the spiking neural networks, retaining key microscopic properties while remaining computationally efficient. This explicit link to single-neuron dynamics makes them a natural framework for studying how neuromodulators, such as dopamine (DA), shape macroscopic network activity. Integrating its effect is a key step toward biologically grounded brain models. Because the spiking network underlying such a reduction is fully known, the reduced model can be directly validated against it. Here, we perform a systematic comparison between a recently proposed DA-modulated mean-field model and its corresponding spiking network across multiple dynamical regimes, evaluating its ability to capture the qualitative organization of the network activity, and transitions between dynamical states. This comparison identifies the conditions under which the reduction accurately reflects the behavior of the full network, and highlights its theoretical limitations.