Recently a new preprint from our lab has been uploaded to arXiv with Pascal Klamser as first author. In this work we put the so-called “criticality hypothesis” to the test, which states that biological collective information processing systems in order to optimize their performance, should operate close to a phase transition, i.e. a critical point representing the border where the collective dynamics undergoes a qualitative change. It has been suggested that animal groups, which can be also viewed as collective information processing systems, should also operate in this special regime. We test this hypothesis in the context of collective predator response using a generic agent-based model. We show that indeed the collective response appears to be optimal at the critical point, but not due to optimal information transfer as predicted by the criticality hypothesis. However, the critical point turns out to b evolutionary highly unstable due to strong spatial self-sorting effects, This shows that individual-level evolutionary adaptation is not a robust mechanism for self-organized criticality.
If you want to learn more, check out the full preprint including SI: Klamser & Romanczuk (2020), arXiv:2009.02079
Our new paper on vision-based flocking is out! Renaud Bastien and Pawel Romanczuk formulate a mathematical framework for social interactions in groups of agents based only on information available in the visual input received by individuals. Then, following a bottom-up approach, we formulate a minimal vision-based model: Agents only see black & white – something is there or not – and respond only to blobs and edges in their visual field. Surprisingly, with this minimal mechanism we observe different modes of collective movements observed in nature and in “classical” phenomenological flocking models. Our work offers a change of perspective for modeling collective movements, potentially allows to build links to sensory neuroscience and provides new inspirations for swarm robotics.
R. Bastien & P. Romanczuk, Science Advances 05 Feb 2020: Vol. 6, no. 6, eaay0792, DOI: 10.1126/sciadv.aay0792
Checkout also the Twitter threads by Renaud Bastien and Pawel Romanczuk providing more information, animations and a download of the simulator.
Today a new preprint by Parisa Rahmani, Fernando Peruani and Pawel Romanczuk, has been published on arXiv. Using a generic flocking model we investigate the role of limited attention on collective information processing in complex environments. The main result is rather surprising and counters the established knowledge in flocking behavior as well as more generally in network science: Less connected flocks are better at reaching global consensus and achieve higher migration accuracy in complex environments. This effect can be explained by a collective isolation from distracting environmental cues through self-organization. However, the high accuracy comes at the cost of being unresponsive to environmental cues. The collective becomes more responsive than solitary individual to the environment only at high individual attention capacity, where collective accuracy and global consensus break down. Thus, we observe a fundamental trade-off in collective information processing, which is potentially of relevance to other systems beyond animal collective behavior. Big thanks to Parisa for her hard work on this project!
P. Rahmani, F. Peruani, P. Romanczuk, “Flocking in Complex Environments – Attention Trade-offs in Collective Information Processing”, preprint, arXiv:1907.11691
Check out the new bioRxiv preprint by Renaud Bastien, Department of Collective Behavior, MPI Konstanz and Pawel Romanczuk. In this manuscript, we introduce a mathematical framework for purely vision-based social interactions in agent-groups, and how it leads to self-organized emergence of collective movement. Already the simplest models – without any representation of distance or object classification lead to various modes of collective behavior.
Check out the new arXiv preprint by Bryan C. Daniels and Pawel Romanczuk. In this paper we explore the role of higher order network structure in collective decision making. More specifically, we investigate the speed-accuracy trade-off in collective binary decisions, how it is affected by degree heterogeneity across nodes, e.g. a so-called rich-club structure, and how it can be quantified in terms of the participation ratio as a core graph property.
On Thursday, Sep 27th 2018, we got the amazing news that our application for the cluster of excellence Science of Intelligence got selected for funding within the German excellence funding program. Pawel Romanczuk and Jens Krause will contribute their expertise to explore principles of social and collective intelligence within the consortium of in total 21 PI’s from such different fields as biology, robotics, psychology, philosophy, neuroscience, educational and computer science. The cluster will be also collaborating through strategic partnerships with the MIT Center for Brain, Minds and Machines and the MIT Center for Collective Intelligence , as well as various industry partners such as Amazon Development Center Germany. We are looking forward to seven years of cutting-edge research investigating principles of intelligence and building bridges between the various disciplines of intelligence research.
Check out our new preprint entitled “Searching for structure in collective systems” by Colin Twomey (U Penn) as the main author, co-authored by Pawel Romanczuk. The manuscript proposes a novel, information theoretic method for quantifying collective systems. The method does not make any assumptions about the nature of interactions in the collective, e.g. it does not rely on assuming pairwise interactions. Our approach is complementary to both, a bottom-up approach focusing on microscopic (binary) interactions (e.g. analyzing pairwise correlations), and a macroscopic approach focusing on aggregated collective variables. Thus it can provide important insights with respect to the, possibly dynamic, meso-scale structure of collectives, from fish schools to neurons.
This week Pawel Romanczuk is greatly honored to be an invited speaker at the Ising Lectures, at the Institute for Condensed Matter Physics in Lviv/Ukraine on invitation by Prof. Holovatch.
In parallel, there is a great ICTP conference in Trieste/Italy on Collective Behavior. From our lab Parisa, Pascal and Yinong are attending it, and Pawel rushes there as well to give a talk on Thursday, May 10th. The program looks amazing and the conference looks like a great opportunity to get up to date with cutting-edge research on collective behavior and meet a lot of great people.
Recently the paper “How ecology shapes exploitation: a framework to predict the behavioural response of human and animal foragers along exploration-exploitation trade-offs.”, co-authored by Pawel Romanczuk appeared in Ecology Letters. Therein, we propose a novel framework for understanding how spatio-temporal dynamics of ressources shape the social interaction between competing agents by combining theory with a review of empirical observations. The work started in the course of an exciting collaboration supported by the Princeton-Humboldt strategic partnership. Many thanks to all co-authors – in particular Chris Monk who did a great job putting everything together!
A new paper from the Robofish team, just appeared online! The Robofish project is an exciting interdisciplinary collaboration between fish biologist around David Bierbach & Jens Krause, bio-roboticist around Tim Landgraf and our lab. In this specific paper the robotic fish was used probe the social behavior of surface and cave-dwelling fish. Thanks to all co-authors, especially to David Bierbach for the great work!
Bierbach, D. et al: “Insights into the social behavior of surface and cave-dwelling fish (Poecilia mexicana) in light and darkness through the use of a biomimetic robot”, Front. Robot. AI (2018), doi: 10.3389/frobt.2018.00003