獎賞性學習與懲罰性學習交互作用之神經基礎

Abstract

增強式學習的獎勵和懲罰對於人類的行為是很重要的調節因子，然而獎勵學習與懲罰學習是否由相同神經機制反映仍有許多爭論。此外，人類經常在同一個時間點學習如何最大化獎勵與最小化懲罰，但是迄今僅有極少數的研究在探討獎勵學習與懲罰學習的交互作用以及初級增強物與次級增強物的差異。我們利用功能性磁振造影於正常成人進行增強式學習任務時檢測其腦部反應，欲探討獎勵學習與懲罰學習之交互作用，並觀察初級增強物和次級增強物有無差異。目前研究支持獎勵學習與懲罰學習為不同的神經機制，此外，在交互作用下僅有懲罰學習會影響獎勵學習，且初級增強物和次級增強物似乎並無明顯差異。這些結果表示當獎勵學習與懲罰學習具有競爭性關係，懲罰學習會較獎勵學習優勢。此研究中可以增強我們對於增強式學習的神經科學知識。

Reward and punishment are important modulators of reinforcement learning to guide human behaviors. However, it remains elusive whether reward learning and punishment learning would share the same neural mechanism. Although humans usually learn to maximize reward and minimize punishment at the same time, the mutual influence between reward and punishment learning and the difference between primary and secondary reinforcers remain largely unknown. The current research investigated the interaction between reward learning and punishment learning as well as the differences between primary and secondary punishment, by using functional magnetic resonance imaging in combination with probabilistic instrumental learning tasks and genetic screening in healthy adults. Our findings support the view that reward learning and punishment learning may have different mechanisms. In addition, we found reward learning was interfered by punishment learning but not vice versa, and it seemed that there was no significant difference between primary and secondary punishment. These findings suggest that when reward learning conflicted with punishment learning, punishment learning would be more dominant. Results obtained from these investigations can enhance our understanding about reinforcement learning.