Lee Jaeeon, Sabatini Bernardo L
Howard Hughes Medical Institute, Department of Neurobiology, Harvard Medical School, Boston, MA, USA.
Department of Molecular and Cellular Biology, Center for Brain Science, Harvard University, Cambridge, MA, USA.
Nat Rev Neurosci. 2025 May 7. doi: 10.1038/s41583-025-00925-2.
A hallmark of optimal reinforcement learning is that an agent learns to avoid actions that lead to negative outcomes while still exploring alternative actions that could lead to better outcomes. Although the basal ganglia have been hypothesized to contribute to this computation, the mechanisms by which they do so are still unclear. Here, we focus on the function of the striatal indirect pathway and propose that it is regulated by a synaptic plasticity rule that allows an animal to avoid actions that lead to suboptimal outcomes. We consider current theories of striatal indirect pathway function in light of recent experimental findings and discuss studies that suggest that indirect pathway activity is potentiated by the suppression of dopamine release in the striatum. Furthermore, we highlight recent studies showing that activation of the indirect pathway can trigger an action, allowing animals to explore new actions while suppressing suboptimal actions. We show how our framework can reconcile previously conflicting results regarding the indirect pathway and suggest experiments for future investigation.
最优强化学习的一个标志是，智能体学会避免导致负面结果的行为，同时仍在探索可能导致更好结果的替代行为。尽管已有假说认为基底神经节参与了这一计算过程，但其具体机制仍不清楚。在这里，我们聚焦于纹状体间接通路的功能，并提出它受一种突触可塑性规则的调节，这种规则使动物能够避免导致次优结果的行为。我们根据最近的实验结果来考量当前关于纹状体间接通路功能的理论，并讨论那些表明纹状体中多巴胺释放受抑制会增强间接通路活动的研究。此外，我们强调最近的研究表明，间接通路的激活可以触发一种行为，使动物在抑制次优行为的同时探索新行为。我们展示了我们的框架如何调和先前关于间接通路的相互矛盾的结果，并提出了未来研究的实验建议。
