How do neuromodulators encode uncertainty to aid flexible behaviour?

在一个充满不确定性的瞬息万变的世界里，我们如何快速做出决定？健康的人特别擅长在日常生活中做出最优选择；也许他们认为坐公交车是去上班最快、最好的方式。尽管公交车的确切行程时间有很小的不确定性，但如果它是最快的，我们应该坚持我们的选择。然而，如果公交路线上突然出现道路施工，我们可能不得不灵活适应，而不是步行到更远的火车站，以获得最快的旅程。我们可以想象，如果我们不能很好地适应不确定性，我们最终会做出糟糕的选择。在一种情况下，我们可能每天都会因为公共汽车晚点一次而改变交通方式。或者，当道路工作发生时，我们可能无法适应，仍然倾向于留在缓慢的公交车上。这两种情况都会导致更糟糕的选择，表现为旅行时间更长。事实证明，在不确定的情况下做出好的选择所需的计算并不简单--无论是计算机还是机器人，或者患有精神障碍的人都特别擅长灵活地做出决定。目前，机器学习算法和机器人可以惊人地擅长执行单一的、定义的任务；但大多数都无法适应新的环境或执行许多不同的任务。因此，了解大脑中的哪些过程有助于我们做出灵活的选择将是有用的，这可能会使先进的人工智能和机器人领域以及心理健康保健受益。到目前为止，科学家认为，做出适应性的选择后，我们的大脑会对世界以及未来可能带来回报的行动形成“最佳猜测”(预测)，这对我们有帮助。然而，我们不知道在哪个区域的哪些生物信号可能会携带关于我们不确定世界的这些“预测”的信息。在这个项目中，我将研究两种特定的神经化学物质(称为去甲肾上腺素和乙酰胆碱)，以询问它们是否可能起到潜在的生物学‘不确定性信号’的作用。这些神经化学物质将有一个荧光标签，使我能够通过光度显微镜和插入活着的小鼠大脑的光纤来监测它们；所有这些都是在动物解决一个不确定的决策游戏时进行的。我还将在已知参与学习和决策的特定大脑区域(称为海马体和前额叶皮质)中使用药物来操纵这些神经化学物质。我们是否需要这些区域的神经递质来对我们的环境做出好的猜测，并预测好的选择？回答这些问题可以更好地指导哪些药物靶点与未来的精神药物相关，例如改善的抗抑郁药物。在研究了相关的神经化学物质和大脑灵活决策后，我的目标是建立计算模型，测试大脑可能使用的不同潜在机制，并看看哪种模型最适合动物的选择。这种成功的模型可以为未来灵活的多任务机器学习算法提供参考，这些算法适用于从科技和工程行业到金融再到医疗保健的广泛领域。

How do we decide things quickly in an everchanging world with a lot of uncertainty? Healthy humans are exceptionally good at making optimal choices in everyday life; perhaps deciding that the bus is the fastest and best way to get to work. Despite small uncertainties in the exact journey time of the bus, we should stick to our choice if it's the fastest. However, if there is a sudden roadwork on the bus route, we may have to adapt flexibly and instead walk to the train station further away to get the quickest journey. We can imagine that if we cannot adapt well to uncertainty, we end up with bad choices. In one scenario, we may change transportation every day just because the bus was late once. Alternatively, when the road work happen, we may fail to adapt and still favor staying on the slow bus. Both of these scenarios result in a worse choice in the form of longer travel time. It turns out that computations needed for making good choices under uncertainty are not simple - neither computers or robots, nor people with psychiatric disorders are particularly good at flexible decision making. Currently, machine learning algorithms and robots can be amazingly good at performing single, defined tasks; but most fail at adapting to new environments or performing many different tasks. Hence, it would be useful to understand what processes in the brain help us make flexible choices, possibly benefiting both the advancing field of artificial intelligence and robotics, as well as mental health care. So far, scientist believe that making adaptive choices, our brain forms 'best guesses' (predictions) about the world and what actions might lead to something rewarding in the future helps us. However, we don't know in detail what biological signals in what area may carry information on these 'predictions' about our uncertain world. In this project, I will look at two specific neurochemicals (called noradrenaline and acetylcholine), to ask if these may be acting as potential biological 'uncertainty signals'. These neurochemicals will have a fluorescent tag, allowing me to monitor them with a photometry microscope and optical fibers inserted into the brains of living mice; all whilst animals are solving an uncertain decision-making game. I will also manipulate these neurochemicals with drugs in specific brain areas know to be involved in learning and decision making (called Hippocampus and Prefrontal Cortex). Do we need the neurotransmitters in these areas to make good guesses about our environment and predict good choices? Answering these questions could better guide what drug targets are relevant for future psychiatric drugs, such as improved antidepressants. After having investigated the relevant neurochemicals and brain flexible decision making, I aim to build computational models testing different potential mechanism that the brain might be using, and see which model best fits the animal's choices. Such successful models could inform future flexible, multi-task machine learning algorithms, relevant to a broad range of fields; from tech and engineering industry, to finance, to healthcare.