Breaking Balance: Encoding local error signals in perturbations of excitation-inhibition balance

Effective learning in neural circuits requires credit assignment—linking local synaptic changes at individual neurons to distant behavioral outcomes—a challenge not addressed by classic Hebbian plasticity. While machine learning models rely on explicit local error signals to guide weight updates, we do not understand the underlying feedback mechanisms in neurobiology. Here, we propose a simple and biologically plausible solution: local deviations from precise excitation/inhibition (E/I) balance encode error signals that instruct synaptic plasticity. Using a computational model derived from an adaptive control theory framework, we show that targeted feedback to inhibitory interneurons perturbing E/I balance can produce neuron or assembly-specific error signals that support credit assignment in multilayer networks. Simulations demonstrate that such a balance-controlled plasticity mechanism is consistent with phenomenological local plasticity models while enabling online learning in hierarchically organized networks. Further, we show this framework is consistent with key features of dis-inhibitory microcircuit dynamics during in-vivo learning experiments. These results suggest that the brain may exploit E/I balance not just for stability, but as a substrate for error-driven learning.