Physiological and neurophysiological mechanisms for exploration and mistakes

我们为什么会犯错误？我们有能力找出最有价值的行动方案，但我们并不总是选择它。有时，“错误”并不是真正的错误，而是一种探索的形式--意味着一种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，即使它不会产生最有价值的选择。然而，我最近的研究表明，当错误没有任何想象中的好处时，同样的探索机制也会导致错误。我相信，这在一定程度上是因为我们的大脑是为一个总是不确定的世界而进化的，在这个世界里，一点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生理过程如何改变大脑在相同情况下的反应。这项建议将通过调查内源性生理过程如何影响探索为这一目标奠定基础。这一观点的核心假设是，探索性机制会导致我们在有用和无用的时候犯错，因为探索性机制是由决策回路调整过程中与唤醒相关的内源性波动驱动的。我们将在两个人类行为目标和一个灵长类动物电生理目标中测试这一假设的不同方面。我们的方法是统一的观察，一个生理探索机制将在真实的时间，这意味着它将在几秒钟或几分钟内，我们的身体变化。相比之下，目前大多数的决策模型都假设行为只在试验时间内变化，即以离散的任务事件为单位，如试验和奖励结果。目标1首先通过操纵试验时间和真实的时间之间的联系，并询问这种操纵是否系统地改变了探索的时间，来确定探索是发生在真实的时间还是试验时间。然后，扩展我和其他人之前将错误与自主觉醒联系起来的工作，目标2将确定内源性的、真实的觉醒时间周期是否足以解释探索中的时间变化。最后，目标3将利用我们对神经元群体进行大规模记录的能力，以确定唤醒的实时波动是否预测了我们在探索期间发现的神经元调谐的中断。如果我的中心假设是正确的，这些研究将表明，错误是由觉醒的自然波动和与觉醒相关的大脑活动模式产生的。无论结果如何，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上建立一个研究计划，我和我的团队将在几十年内建立这个研究计划，因为我们阐明了认知的不同方面-包括工作记忆，注意力和执行控制-是如何被内源性的，真实的时间生理过程改变的。

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.