Physiological and neurophysiological mechanisms for exploration and mistakes

为什么我们会犯错？我们有能力确定最有回报的行动路线，但我们并不总是选择这样做。有时，“错误”并不是真正的错误，而是一种探索的形式--意味着发现新机会和了解世界的动力。在一个不确定的环境中，一些探索是至关重要的，即使它在当下不会产生最有回报的选择。然而，我最近的工作表明，当错误没有任何可以想象的好处时，同样的探索性机制也会导致错误。我认为，这在一定程度上是因为我们的大脑进化到了一个总是不确定的世界，在那里进行一点探索几乎总是有利的。这项研究计划的长期目标是阐明内源性的生理过程如何改变大脑在相同环境下的反应。这一建议将通过研究内生的生理过程如何影响探索，为这一目标奠定基础。这一提议的中心假设是，探索机制在有用和不有用的时候都会导致我们犯错误，因为探索是由决策回路调整中的内生、与唤醒相关的波动驱动的。我们将在两个人类行为目标和一个灵长类电生理学目标中测试这一假设的不同方面。我们的方法是统一的，因为观察到生理探索机制将实时运行，这意味着它将随着我们身体变化的几秒钟或几分钟而变化。相比之下，大多数当前的决策模型假设，行为只在试验时间内变化--以试验和奖励结果等离散任务事件为单位。目标1将首先通过操纵审判时间和实时之间的联系，并询问这种操纵是否系统地改变了勘探的时间，来确定勘探是在实时进行还是在审判时间进行。然后，在我和其他人之前将错误与自主神经唤醒联系起来的工作的基础上，目标2将确定内生的实时唤醒周期是否足以解释探索中的时间变化。最后，目标3将利用我们从神经元群体中大规模记录的能力来确定唤醒的实时波动是否可以预测我们在探索期间发现的神经元调谐中断。如果我的中心假设是正确的，这些研究将表明，错误出现在唤醒和与唤醒相关的大脑活动模式的自然波动中。不管结果如何，这些研究将测试许多决策模型的隐含假设，并在认知神经科学和生理学的交叉点建立一个研究计划，我和我的团队将在数十年的时间里建立这个研究计划，因为我们将阐明认知的不同方面--包括工作记忆、注意力和执行控制--是如何被内源性的实时生理过程改变的。

Why do we make mistakes? We have the capacity to identify the most rewarding course of action, yet we do not always choose it. Sometimes, "mistakes" are not really mistakes, but rather a form of exploration-meaning a drive to discover new opportunities and learn about the world. In an uncertain environment, some exploration is critical, even if it does not produce the most rewarding choice in the moment. However, my recent work suggests that the same exploratory mechanisms can cause mistakes when mistakes have no conceivable benefit. I believe that this happens, in part, because our brains evolved for a world that is always uncertain, where a little exploration was almost always advantageous. The long-term goal of this research program is to elucidate how endogenous, physiological processes alter how the brain responds in identical circumstances. This proposal will lay the foundation for this goal by investigating how endogenous, physiological processes influence exploration. The central hypothesis of this proposal is that exploratory mechanisms cause us to make mistakes both when they are useful and when they are not because exploration is driven by endogenous, arousal-linked fluctuations in the tuning of decision-making circuits. We will test different aspects of this hypothesis in two human behavioral aims and one primate electrophysiology aim. Our approach is unified by the observation that a physiological exploratory mechanisms would operate in real time, meaning it would vary over the seconds or minutes across which our bodies vary. In contrast, most current models of decision-making assume that behavior only varies in trial time-in units of discrete task events like trials and reward outcomes. Aim 1 will first determine whether exploration occurs in real time or trial time by manipulating the link between trial time and real time and asking whether this manipulation systematically changes the timing of exploration. Then, expanding on previous work by myself and others that links mistakes to autonomic arousal, Aim 2 will determine whether endogenous, real time cycles of arousal are sufficient to explain temporal variation in exploration. Finally, Aim 3 will harness our capacity for large scale recordings from populations of neurons to determine whether real-time fluctuations in arousal predict the disruptions in neuronal tuning that we find during periods of exploration. If my central hypothesis is correct, these studies will show that mistakes emerge from natural fluctuations in arousal and arousal-linked patterns of brain activity. Regardless of outcome, these studies will test an implicit assumption of many decision-making models and establish a research program at the intersection of cognitive neuroscience and physiology that my team and I will build on for decades as we elucidate how different aspects of cognition-including working memory, attention, and executive control-are changed by endogenous, real time physiological processes.