Pattern preferences, information theory, and the evolution of signal design

动物的求偶行为通常涉及复杂的视觉信号，因此理解塑造信号设计的力量是动物行为中的一个中心问题。这个项目的中心假设是，求爱信号的设计是由模式偏好驱动的，而模式偏好又由当地栖息地的模式塑造。这是感觉驱动力的扩展版本，这是一个描述环境如何影响动物感觉系统从而影响偏好的强有力的假说。到目前为止，感觉驱动强调了求爱信号的可检测性，而拟议的研究强调了视觉系统处理信号的效率。该研究计划结合了行为研究和前沿的计算图像处理，以测试信号模式和模式偏好由当地栖息地塑造的中心假设。重点关注的动物是彩虹达特(Etheostoma Caeruleum)，这是北美东部一种丰富的、视觉上引人注目的鱼类。该项目的更广泛影响包括为学生和博士后助理进行国际合作，培训来自代表性不足群体的学生，为信号设计研究人员举办研讨会，以及在昆士兰大学开发动物行为高级实验室模块。拟议的研究针对信号设计提出了一个新的假设，在信息论的更广泛背景下构建了强大的感觉驱动模型，而不仅仅是信号检测理论。传统的感觉驱动假设可探测性是对动物信号的驱动约束，而基于信息论的感觉驱动更广泛地强调感觉系统处理模式的效率。人类对有效处理的信息的偏爱已经被证明，他们通常喜欢与自然场景的空间统计相匹配的模式。因此，基于加工偏差的感觉驱动预测，动物将更喜欢与当地栖息地的统计数据相匹配的信号模式。重要的是，另一种伪装假说也做出了同样的预测。伪装和性信号传统上是自然选择和性选择之间对立关系的缩影；然而，从信息论的角度来看，服务于这些功能的设计可能是相互加强的。实验将使用生成性人工智能，这是一个快速发展的深度机器学习领域，以经验方式测试动物信号设计的相互包容的替代假设。这种方法提供了一种令人兴奋的技术创新，可以极大地提高我们探索接收者心理和信号偏好的能力。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Courtship behavior in animals often involves elaborate visual signals, so understanding the forces that shape signal design is a central question in animal behavior. The central hypothesis of this project is that courtship signal design is driven by pattern preferences that are in turn shaped by patterns in the local habitat. This is an expanded version of sensory drive, a robust hypothesis that describes how the environment influences animal sensory systems and therefore preferences. To date, sensory drive has emphasized the detectability of courtship signals, whereas the proposed research emphasizes the efficiency with which signals can be processed by the visual system. The research plan combines behavioral studies and leading-edge computational image manipulation to test the central hypothesis that signal patterns and pattern preferences are shaped by local habitats. The focal animal is the Rainbow Darter (Etheostoma caeruleum), an abundant and visually striking fish species in eastern North America. Broader impacts of the project include international collaboration for students and a postdoctoral associate, training of students from underrepresented groups, a workshop for signal design researchers, and development of upper-level laboratory modules in animal behavior at UMBC.The proposed research addresses a novel hypothesis for signal design, framing the powerful model of sensory drive in the broader context of information theory rather than signal detection theory alone. Whereas traditional sensory drive posits detectability as the driving constraint on animal signaling, sensory drive based on information theory emphasizes more broadly the efficiency with which patterns are processed by sensory systems. A bias for efficiently processed information has been demonstrated in humans, who generally prefer patterns that match the spatial statistics of natural scenes. Sensory drive based on processing bias therefore predicts that animals will prefer signaling patterns that match the statistics of local habitat. Importantly, an alternative hypothesis of camouflage makes the same prediction. Camouflage and sexual signaling traditionally epitomize the antagonistic relationship between natural and sexual selection; however, from an information theoretic perspective, the designs that serve these functions can be mutually reinforcing. Experiments will use generative artificial intelligence, a rapidly developing field of deep machine learning, to empirically test mutually inclusive, alternative hypotheses of animal signal design. This method offers an exciting technological innovation that could dramatically improve our ability to explore receiver psychology and signal preferences.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.