Using 'virtual reality'-techniques to dissect important characteristics of highspeed decision-making in archerfish

该项目旨在利用最近建立的技术来解决一些令人兴奋但以前无法解决的关于狩猎射水鱼的快速启动决策的问题。在用瞄准好的射击驱逐猎物后，这些鱼很快就决定启动一个启动动作，将鱼向右转向猎物，然后以正确的速度将其推到猎物身边。令人惊讶的是，在最初下降运动的40毫秒内采样的信息是鱼做出决定所需的全部信息。在过去的几年里，我们发现射水鱼的启动决策联合收割机结合了巨大的速度和令人印象深刻的复杂性。我们希望在适当设计的行为实验中解决的开放性问题是：射水鱼真的有可能在100 ms内设计出一条完整的迂回路径，最佳地引导它们通过障碍物吗？他们如何能够如此迅速和足够精确地访问内部地图，以将与障碍物相关的信息输入到他们的启动决策中？基于什么样的信息，射水鱼可以缓冲它们的“开环”开始，以防止后来可能对猎物路径的干扰？是什么线索让鱼单目，并在短短40毫秒检测其猎物的初始距离和垂直速度？什么样的“反馈”可以让鱼评估他们过去的决定是正确的还是系统性的错误？决策过程可以独立处理多少个初始值组合？一项新建立的技术使所有这些问题都触手可及。射水鱼决策的复杂性和速度的不寻常组合将使研究结果特别有趣。它们将与驱动复杂行为真正需要什么类型的网络这一基本问题相关，并且它们将直接为直接检查网络中的处理奠定基础。认识到速度和复杂性可以以惊人的方式结合在一起，这对人类决策的概念提出了质疑，但也可以在技术系统中加以利用。

The project aims to exploit a recently established technique to address a number of exciting but previously inaccessible questions about the rapid start-decisions of hunting archerfish. After having dislodged prey with a well-aimed shot these fish very rapidly decide to launch a start manoeuvre that turns the fish right towards were their prey is later going to land and pushes it off with just the right speed to arrive simultaneously with its prey. Surprisingly, information sampled in just 40 ms of the initial falling motion is all that the fish needs for its decision. Over the past years it turned out that the archerfish's start-decisions combine enormous speed with impressive complexity. The open questions we would like to solve in appropriately designed behavioural experiments are these: Is it really possible that archerfish can design a complete detour path, that optimally guides them past obstacles, in just 100 ms? How can they so rapidly and with sufficient precision access internal maps to feed obstacle-related information into their start-decisions? On the basis of what information can archerfish buffer their 'open-loop' starts against potential later disturbances of their prey's path? What is the cue that allows the fish to monocularly and in just 40 ms detect initial distance and vertical speed of their prey? What kind of 'feedback' allows the fish to evaluate if their past decisions were appropriate or systematically wrong? How many initial-value combinations can the decision-making process handle independently? A newly established technique brings all these questions within reach. The unusual combination of complexity and speed in the archerfish decisions will make the findings particularly interesting. They will be relevant for the fundamental question of what types of networks are really needed to drive complex behaviour and they will directly lay the foundation for a direct examination of the processing in the network that elicits the starts. Appreciating that speed and complexity can go together in remarkable ways questions concepts of human decision-making but could also be exploited in technical systems.