Network Science and Cognitive Neuroscience: Explaining How the Brain Works?


May 23-25, 2016
Enchantment Resort
Sedona, Arizona, USA


Recent years have seen two largely independent scientific developments. On one side, there has been the rise of “network science” and “networks in science,” whether using the structure and dynamics of networks to gain a new perspective on complex systems, or learning and employing explicit causal networks to understand mechanisms and processes in scientific domains. On the other side, new methodologies and techniques for mapping and recording large-scale neural systems have sparked empirical and computational work that focuses on the brain as a quintessential complex causal network. Indeed, understanding how the wondrous capabilities of the human brain emerge from the intricate multi-level temporal-spatial organization of anatomical structure, signaling function, and causal connectivity is one of the most intriguing and perplexing challenges facing us today.

The 2004 TICS review by Sporns et al. provided an overview of the state of understanding of the emerging efforts to merge complex systems network science with cognitive neuroscience. In the review, Sporns et al., also highlighted a number of open questions and suggested ways the field could continue moving forward. The ensuing decade has seen an explosion of publications using increasingly sophisticated approaches from network science, resting state functional connectivity measures, and large multi-site connectome projects aimed at expanding our understanding of how the brain normally functions and how dysfunction can arise due to network perturbations.

At the same time, causal networks (including Bayesian networks and other graphical models) are increasingly used to model brain connectivity and communication structures, in no small part because of the many machine learning algorithms for extracting graphical structure from fMRI data. These networks have helped to provide insight into the pathways by which different brain areas communicate with one another, and in some cases, have been (preliminarily) connected with “higher-level” (e.g., psychological) descriptions of particular cognitive tasks. There have been some initial attempts to span these two different types of (uses of) networks, but little systematic unification.

Through its Studying Complex Systems program JSMF is supporting the ongoing development of theoretical and methodological approaches for the study of complex networks. And via the JSMF Understanding Human Cognition program JSMF is supporting collaborative grants using networks—both complex systems and causal networks—in studies of neural development, action, emergence from unconsciousness, recovery from injury and disease, and how the brain may or may not respond to rehabilitation interventions.

The time seems right to bring together representatives from these different groups for a workshop discussion about what we are currently learning about complex causal brain networks and how the field has changed as a result of the recent focus on networks and connectivity. The workshop will also explore what new opportunities and challenges network science poses for the field of cognitive neuroscience, what new concepts and methods are emerging and what they could be contributing to our understanding of the human brain.

Some (non-exhaustive, non-exclusive) guiding questions:

  • What are the advantages and limitations of different network modeling frameworks for understanding aspects of brain connectivity, causality, and dynamics?
  • How should we understand or interpret the networks that are generated through automated, machine learning methods? How do these interpretations vary as a function of the input data?
  • Are there common principles or explanatory schema that unify (some of) the many successes from using network science in neuroscience? For example, are there particular measures of network structure that are more likely to be explanatory in developmental or rehabilitation contexts?
  • In the context of neuroscientific questions, can we integrate dynamical models (from network science) and causal models (from machine learning)? For example, can we construct multi-level causal models, or develop an “intervention logic” for network science?
  • Do network science methods or frameworks make key assumptions that are arguably violated in neuroscientific contexts? If so, can we adjust our algorithms and models so that those assumptions can be weakened or eliminated?
  • Are there “privileged” timescales for neural functioning at different levels of brain organization?
  • How can we use network models and ideas to better explicate, model, and test claims of localization, neural reuse, modularity, and related notions?