4th Collaborative Workshop on Modeling Collective Behavior

Where: Pioneer Works, Brooklyn, NY

When: June 10-12, 2019

Organizers: Jason Graham (Scranton University), James Curley (UT Austin) & Simon Garnier (NJIT)

Funding: NSF OAC Public Access Initiative grant #1838955

Application deadline: March 31, 2019 (Apply here)

The goal of this workshop is to provide researchers coming from mathematics, physics, engineering and biology with experience in working collaboratively to model and analyze data produced by research in collective behavior. This year we are also including an interdisciplinary art/science project (see Project 4 below) that will use virtual reality to explore and interact with data and models of collective behaviors.

During the workshop, interdisciplinary teams will work toward constructing mathematical and/or computational models of biological data. Each team will be tasked with exploring, analyzing and modeling original data sets provided by biologist attendees. The event will be similar in spirit to a hackathon in computer science, with the goal of finding new ways to understand the provided biological data. Regular debriefing sessions will be organized throughout the 3 days of the workshop in order for each team to provide updates to and receive feedback from the rest of the participants on their progress.

This should provide valuable training and experience for researchers coming from all fields, as well as an opportunity to create new collaborations beyond the workshop itself. Ultimately, we hope this workshop to impact research on collective animal and human behavior by promoting close collaboration between scientists working on understanding the organization and functioning of animal and human societies.

Project list:

The architecture of a social network is an emergent property of the social behaviour of the individuals within it. As such, the composition of individuals within a population, in relation to their social traits, determines the structure of the society. One trait commonly known to influence individuals’ social behaviour across a range of systems is age. However, how these age related differences in sociality scale up to shape emergent networks, and how changes in age composition affect society structure and functioning, remains largely unknown. Using a tracked wild bird system (great tits – Parus major) containing thousands of individuals that show age-related changes in social behaviour, this project will address how age compositions can predict emergent local social dynamics across space and resultant global social structures through time. This will be carried out using vast observational data, as well as fine-scale experiments that manipulated age compositions under natural settings. In this way, this project aims to provide new insight on how simple individual level differences shape emergent spatio-temporal social dynamics, and the causal significance of age compositions within societies.

–

Maintaining group cohesion is essential in order to reap the benefits of sociality. However, wide distances between individuals’ positions, in terms of physical locations, behaviors, motivations, ideas and values, can threaten the short-term and/or long-term cohesion of a group. In some cases, group members can continuously and simultaneously monitor the positions of some or all group members and adjust to them accordingly. However, more often, communication between individuals is required in order to monitor and regulate group cohesion. Communication can not only be used to share information on one another’s positions, it can also be used to more strongly influence the positions of others, for instance by advertising strength of conviction to one’s current position. The aim of this project is to create a broad theoretical model of social cohesion and to explore the role that communication can play in its regulation. The model will focus on the idea that group members monitor one anothers’ positions (either physically, behaviorally or ideologically) and, when some individual(s) stray too far from the center position of the group, group members use communication to close the gap by pulling others towards their own position. Potential parameters the model will explore include the level of deviation and clustering between group members’ positions, the number and relative positions of signallers, the spread of information through the group, the decision-making mechanism, the strength of communication on group members’ behavior, and the level of motivation to remain cohesive. From our models, we will measure the time it takes to regain group cohesion, the effect of communication on the change in position of the group centroid, and the contexts in which group fission occur.

–

Project 3: Modeling collective feeding (leaders: Olga Shishkov, Miguel A Fuentes-Cabrera, with external help from David Hu)

Although collective dynamics is a well-established field, the collective dynamics of feeding behavior has not yet been thoroughly investigated. Examples of collective feeding of animals include the group feeding of livestock raised in herds, maggots eating a carcass, and a piranha shoal devouring prey. For some predators, a group feeding behavior can devolve into a “feeding frenzy”, in which the animals eat as much as possible without regard to hurting one another – like the behavior of Black Friday shoppers. These behaviors can serve as inspiration for robots attempting to all recharge at the same charging station, clean up waste from a single source, or mine from the same mine, as this is a similar problem of resource collection by multiple agents. In this workshop, we will investigate how different feeding patterns of individuals and food arrangements affect the group dynamics. We will start with individuals in 2D gathered around a single food source, expand the model to include different configurations and a quasi-2D state, and apply it to specific scenarios, such as charging robots and fish feeding frenzies. Concepts such as the time each animal spends eating, how soft its body is, and the pressure generated by each individual will be considered. This will contribute to a theoretical basis for experimental work on the collective behavior of fly larvae, fire ants, and other animals

–

Project 4: The Realm: modelling collective behavior phenomena in virtual reality (leaders: Heather Barnett, Daniel Strömbom)

Inspired by the coordinated motion of animal groups, ‘The Realm’ combines visual art and computational design with behavioral science to create aesthetic and immersive collective encounters. In a virtual environment, participants interact with imaginary bio/digital creatures, whose individual and collective actions are driven by rules drawn from biological data (such as shoaling fish and flocking birds), integrating SPP models, boids and generative code to create biodigital agents which can be interacted with. Computational models and player interactions interconnect to create a dynamic co-evolving system, affected by bodily gestures, sounds and social signals. The primary intended output from the project is an art installation but there is also potential for developing other applications such as educational or scientific data visualisation tools. We invite others to hack the modelling systems and explore alternative realities for visualizing and engaging with collective behavior phenomena in virtual reality. Challenges which could be presented for the workshop to wrangle include:

model algorithms and interaction (how to program agent behaviour which is biologically ‘authentic’ and requires minimal processing power to create real time affect and interaction).
how to elicit phase transitions into behavioural models (i.e. the agents changing behaviour dependent on human cues and/or environmental change)
evolving a gestural language of stimulus / response between real and virtual agents in the system.
alternative applications into educational /scientific tools.

–

Project 5: Mechanistic modeling of decision-making in the slime mold Physarum polycephalum (leaders: Subash Ray, Abid Haque, Jason Graham)

Physarum polycephalum is a single-celled protist capable of showing complex problem-solving and decision-making behaviors despite lacking a neural architecture. Previous studies have suggested that the problem-solving abilities of P. polycephalum is driven by the coupled-oscillator based sensorimotor system linked to its membrane. The membrane is composed of multiple rhythmic contractile regions that constantly interact with one another and lead to the emergence of a complex pattern of contraction-relaxation cycle at the cell level. In a recent study by Ray et al. (2019, in press), the authors found that when a tubule-shaped cell is given a choice between two food sources, the pattern of information transfer between the contractile regions varies with the difference in quality between the food sources. In particular, when the food sources differed in quality, the contractile regions near the rejected food source acted as the information sources and the regions near the chosen food source acted as the information destinations. In this project, we propose to create a mechanistic model to explain this behavior and compare its predictions to the results in Ray et al (2019).