How causal perspectives can inform problems in computational neuroscience

원문 초록

Over the past two decades, considerable strides have been made in advancing neuroscience techniques, yet the translation of these advancements into clinically relevant insights for human mental health remains a challenge. This review addresses a fundamental issue in neuroscience - attributing causality - and advocates for the development of robust causal frameworks. We systematically introduce the necessary definitions and concepts, emphasizing the implicit role of causal frameworks in neuroscience investigations. We illustrate how persistent challenges in neuroscience, such as batch effects and selection biases, can be conceptualized and approached using causal frameworks. Through theoretical development and real-world examples, we show how these causal perspectives highlight numerous shortcomings of existing data collection strategies and analytical approaches. We demonstrate how causal frameworks can inform both experimental design and analysis, particularly for observational studies where traditional randomization is infeasible. Using neuroimaging as a detailed case study, we explore the advantages, shortcomings, and implications for generalizability that these perspectives afford to existing and novel research paradigms. Together, we believe that this perspective offers a framework for conceptualizing, framing, and inspiring innovative approaches to problems in neuroscience.