A chat with Prof. Guy Theraulaz

Share this story on

Facebook icon Twitter icon

Prof. Guy Theraulaz is a leader in the field of self-organisation and animal collectives. He visited the Centre for Ecological Sciences, IISc in October 2022 as an Infosys Chair Professor. Dr Guy Theraulaz is the research director at the Research Center on Animal Cognition, CNRS - University of Toulouse Paul Sabatier, France (https://crca.cbi-toulouse.fr/). He studies collective animal behaviour in social insects, fish schools, bird flocks and ... human crowds.

Here are excerpts from an interview with him:

Good morning, Dr Theraulaz. Why did you decide to work on collective animal behaviour?

Thirty five years ago, I was trained as a neuroscientist, and at that time, it was extremely difficult to study the brain. So I took the chance to work in a social insects lab. There is indeed a kind of analogy between a brain and a social insect society. In a brain, neurons interact with each other just as individuals interact with each other in a social insect society. The main difference between those two systems is the fact that in social insects we can have direct access both at the individual and the collective scales. We then can more easily understand the emergent behaviors at the scale of a colony. When we talk about a social insect colony, we are talking about a superorganism - an organism made up of tiny individuals that interact with each other. Under certain conditions, they develop a kind of collective intelligence. This collective intelligence is a byproduct of these interactions [1]. And so to understand these phenomena you have to characterize these interactions, you have to identify the information which is exchanged when individuals interact with each other, you must also understand how these individuals combine and integrate multiple interactions and finally you have to analyze the effects of these interactions on the behavior of each individual. And this is what now constitutes the essence of my research.


You have studied fish, as well as birds and insects. Your experiments show that fish follow their ‘most influential’ neighbours, and do not pay much attention to other fish.[2] In contrast, ants and other social insects look at many of their nestmates and take a sort of ‘majority opinion’.[1] Why do fish, despite their larger brains, use a simpler system?

Ants mainly interact indirectly with their conspecifics. They mainly use chemical traces left on the ground, and it is a much simpler way to integrate multiple information than when you have to communicate directly with multiple individuals. We find some kind of cognitive economy in ants and in fish, and also in other social animals like sheep, because sheep also interact with only one other individual as it has been recently shown. Fish only interact with one or two neighbours because they cannot track the behaviour of all individuals around them. This way, they just have to pay attention to a few of them.


The fish that became ‘influential’ were usually the ones in front of, and a little to their side, right?

No, it depends on three parameters - the distance between fish, their relative position, and relative orientations[3]. These three parameters play equivalent roles and combine their effect in determining which neighbour will be the most influential for a given fish at a given moment.


Do you think similar schooling patterns would be found in all fish species?

Generally, you can observe similar collective behaviour at least for some species of fish. But there are a few species where the collective behaviour is different. What we are doing now is studying different species of fish under the same conditions and measuring and modeling the social interactions. The next step is creating a table that will summarize and compare that information on each species - if there is attraction, repulsion, alignment, no alignment, etc. Collective behaviour arises from this set of rules.


In your studies on crowd behaviour, you noted that humans form subgroups within the overall crowd, and try to maintain communication within the group, even at the cost of speed [4]. Did you see such subgroups in any other species?

No, because in human crowds, this behaviour results from people wanting to talk to each other. Creating sub-groups that adopt specific configurations is just the easiest way to communicate. But fish don’t talk to each other!


Perhaps in higher animals, such as birds?

Birds are very difficult to study in terms of collective movements. Also, I don’t know if the collective movements would be affected by the sex ratio. So, there are a lot of open questions.


In a study on humans, you found that many people resisted misinformation, and that misinformation could even reduce the negative effect of cognitive biases.[5] Isn’t this counter-intuitive? Can you comment on this finding?

We humans have a lot of cognitive biases. For instance, we have a tendency to underestimate quantities [5,6]. The interesting thing is that by providing humans incorrect information that something is larger than it is, the error caused by cognitive bias is reduced. This is a consequence of a large proportion of individuals compromising with social information, i.e. partially following it. By doing so, individuals are able to benefit not only from relatively accurate social information but also from incorrect information that goes against their cognitive bias. It might be interesting to extend this approach and see if by biasing the information that is provided to a group, we can reduce the effect of the cognitive bias, and lead the group towards better collective behaviour.


So where do we go from here? Both for fish schools and for animal self-organisation in general? What are the main questions and challenges you see in the future?

First, there is a big challenge in humans, because we are interacting more and more - with the internet, with smartphones - and this creates many new forms of collective behaviour. In particular, with the digitalization of society and economies, social information has increasingly taken the form of digital traces, which are the data individuals leave either actively or passively when using the Internet. So, we really need to understand how people interact with each other in these systems, and how collective behaviour depends on the way information is transmitted and provided in such artificial interfaces.

The other big question is the evolution of social interactions. This is a difficult issue because one individual needs to produce a signal and another must produce a specific response to that signal. Then there are two different levels of selection - the collective level (i.e., the properties that emerge thanks to the interactions) and the individuals that benefit from the collective properties. Generally, the functional model of evolution considers these two levels differently. So, we have to find a way to connect these two levels and develop new models of evolution.


Based on your experiences, what advice do you have for students in ecology and behaviour?

I think now behavioural science is becoming a ‘big data’ science. We have many tools to collect data automatically, so we have to be able to manage large amounts of data.  This means that biologists have to study a little bit of computer science. And a little bit of physics also, because statistical physics is the only way that provides the theoretical tools to handle such large amounts of data.


One final question - how do you think your findings would help people? Would it impact industries - such as fisheries and robotics - or conservation efforts?

I was involved, at the beginning of my work, in the field of swarm intelligence [7, 8, 9]. Swarm intelligence is a way to provide efficient solutions to many problems in artificial systems - in developing optimisation algorithms to solve complex problems in telecommunication networks, applied science, and robotics [10]. Our current research on collective intelligence in human groups may also help to design new systems to improve the way information is processed collectively by humans, so as to help groups make better decisions.


Computer scientists and robot programmers are using models based on the behaviour of social animals like ants and wasps[7, 8]. But how good are animal societies at finding the globally best solution? Aren’t there many examples of solutions that are locally optimal but globally poor?

The problem with optimality is that while it is a theoretical framework which might be useful, we usually have a huge amount of possibilities to explore. And the environment is ever-changing. So, it is difficult to get something which is optimal at each time point, since we have so many dimensions to optimise in. So, the aim is not to get an optimal system, but to get a system which is viable and adaptable to different states of the environment. It is better for a system to be able to adapt, and this is why I think self-organisation is interesting [11]. It can provide many different kinds of solutions with the same individual behaviour and the same kinds of interactions. So, we come back to your question in the beginning. With the minimum amount of information, how do we develop an adaptive system? It is a little different from designing and optimising everything at each level.


References and further reading:-

  1. Garnier S, Gautrais J, Theraulaz G (2007) The biological principles of swarm intelligence. Swarm Intell 1: 3–31. https://doi.org/10.1007/s11721-007-0004-y
  2. Lei L, Escobedo R, Sire C, Theraulaz G (2020) Computational and robotic modeling reveal parsimonious combinations of interactions between individuals in schooling fish. PLoS Computational Biology 16(3): e1007194. https://doi.org/10.1371/journal.pcbi.1007194
  3. Calovi DS, Litchinko A, Lecheval V, Lopez U, Pérez Escudero A, Chaté H, Sire C & Theraulaz G (2018) Disentangling and modeling interactions in fish with burst and coast swimming reveal distinct alignment and attraction behaviors. PLoS Computational Biology 14: e1005933. https://doi.org/10.1371/journal.pcbi.1005933
  4. Moussaïd M, Perozo N, Garnier S, Helbing D, Theraulaz G (2010) The walking behaviour of pedestrian social groups and its impact on crowd dynamics. PLoS ONE 5(4): e10047. https://doi.org/10.1371/journal.pone.0010047
  5. Jayles B, Escobedo R, Cezera S, Blanchet A, Kameda T, Sire C, Theraulaz G (2020) The impact of incorrect social information on collective wisdom in human groups. J R Soc Interface 17(170):20200496. https://doi.org/10.1098/rsif.2020.0496.
  6. Jayles B, Kim H-R, Escobedo R, Cezera S, Blanchet A, Kameda T, Sire C & Theraulaz G (2017) How social information can improve estimation accuracy in human groups. Proceedings of The National Academy of Sciences USA, 114: 12620-12625. https://doi.org/10.1073/pnas.1703695114.
  7. Bonabeau E, Dorigo M & Theraulaz G (1999) Swarm Intelligence: From Natural to Artificial Systems. Oxford University Press. https://doi.org/10.1093/oso/9780195131581.001.0001
  8. Bonabeau E, Dorigo M, Theraulaz G (2000) Inspiration for optimization from social insect behaviour. Nature 406: 39–42. https://doi.org/10.1038/35017500.
  9. Bonabeau E & Theraulaz G (2000) Swarm Smarts, Scientific American, 282 (3): 72-79
  10. Dorigo M, Theraulaz G & Trianni V (2020) Reflections on the future of swarm robotics. Science Robotics, 5: eabe4385.. https://doi.org/10.1126/scirobotics.abe4385
  11. Camazine S, Deneubourg JL, Franks N, Sneyd J, Theraulaz G & Bonabeau E (2001) Self-Organization in Biological Systems. Princeton University Press. https://press.princeton.edu/books/paperback/9780691116242/self-organizat...


Thanks to Arpit Omprakash, Kajal Kumari and Vishwesha Guttal for help in framing questions and conducting the interview.

The interview was conducted by Jose Mathew.