CHALLENGE SUMMARY
Individuals in social groups have their own experiences and preferences, but must work together cooperatively in order for the social group to remain cohesive and function optimally. Do these individual differences contribute positively or negatively to the social group? Can individual variation be positive under some circumstances but negative under others? Do individuals retain their preferences and tendencies after joining the group, or does the “group mentality” and social environment override these differences? To address these questions, we must be able to track individuals and their behavioral preferences before and after joining the group, understand the mechanisms that create individual variation, examine the effects of social environment on individual variation, and study the consequences of individual variation for the group. Understanding the role of individual variation in group function will not only provide insights into the proximate mechanisms regulating the function of groups of biological organisms, but can also be applied to develop better functioning computational programs and robotics.
There is considerable evidence that there are relatively stable differences among individuals in the types of behaviors they perform and when they perform these behaviors. Individuals can vary in their response thresholds to certain stimuli; thus, some individuals may respond rapidly to cues presented at low intensities, while other individuals may require prolonged exposure to high intensities of these cues before they respond. Response thresholds to different cues can co-vary, and thus individuals can be cat
egorized into specific “behavioral syndromes” or “personality types,” such as bold or shy. This individual variation in behavior can be maintained across environmental conditions or throughout the lifetime of the animal.
The mechanisms underpinning individual variation are not yet fully characterized. Genetic variation among individuals is clearly critical in establishing individual variation in behavior. The environment experienced during development, juvenile and adult stages also shape behavior. Physiological conditions, including the individual’s reproductive and nutritional state, can alter behavioral response thresholds. Finally, the experience of the individual—particularly learning and the positive and negative reinforcement received when performing the behavior—can modulate response thresholds. However, the molecular and physiological mechanisms by which these different factors cause changes in behavior, and whether they operate on the same pathways at these different timescales, remains to be determined.
Individual variation appears to play a largely positive role in establishing and maintaining successful social groups. Based on their differing response thresholds, individuals will segregate themselves among different tasks, with certain individuals preferentially performing specific tasks. Partitioning individuals among different tasks (division of labor) can improve the efficiency of task performance and the overall productivity of the group. Several studies have demonstrated the positive impacts of diversity on group performance, in both social insect societies and human groups. However, the optimal amount of diversity in behavioral task preferences, the proportion of individuals performing a given type of behavior, and the type of diversity needed (the ability to perform specific tasks well or the ability to flexibly move between tasks) may vary depending on the type of group, the duration of the group, and the goals of the group.
In contrast, the role of individual variation in collective behaviors has been woefully understudied. In collective behaviors, multiple individuals, regardless of their current activities, spontaneously perform a coordinated behavior, such as migration in locusts or birds, schooling in fish, or applause in audiences. Thus, it appears that the signals to perform the collective behaviors elicit responses from the majority of individuals, though there can still be variation in response thresholds to join the behavior. Furthermore, these collective behaviors can be quite novel, and distinct from any behavior the individual normally exhibits. The collective behavior also appears to generally override individual variation in behavior, such that all individuals perform essentially the same identical behavior. However, response thresh-
olds among individuals must still exist in order to allow the group to form, disperse, and, in some cases, function optimally. In the case of swarms of honeybee colonies seeking a new nest site, there is a distinct group of individuals (scouts) that have more information and coordinate the behaviors of the other individuals.
The challenge to the working group is to develop general principles to understand the interplay between individual variation and group function: is individual variation beneficial or harmful to the functioning of the group and under what circumstances, how much is individual variation subsumed by the social environment, and what are the underlying mechanisms?
The interplay among robots or artificial life units is another area for exploration. Generally speaking the word robot may refer to both physical robots and virtual software agents. Talking to the experts in this field it appears that there is no universal agreement on which machines/devices represent the robots; however, one has the general appreciation that robots could perform tasks including moving around, operating a mechanical limb, sensing and manipulating their environment, and exhibiting intelligent behavior. Particularly, one likes to see behavior that resembles intelligent beings such as humans or other animals. There exists consensus that “robot” signifies an apparatus (machine/device) that can be programmed to perform a variety of physical tasks or actions.
It is perceived that the two distinct ways that robots are different from actual beings are in the arena of cognition and biological features. Regardless of its human-like or dislike appearance robots need programming to function properly. The advent of modern feedback control systems along with advances in computational powers of miniaturized electronics has made modern robots a lot more artificially intelligent. This allows them to perform tasks based on their own sensing of their surroundings and potentially perceived outcomes. This may be performed either by an individual robot or a group (or swarm) of robots. In the swarm scenario, the concept of particle swarm optimization technique could be used to control the movements of the robots. In this manifest the robots rely on their own individual sensing experiences (self-awareness) of the environment and also utilize the collective knowledge (swarm knowledge) of the environment among all of them to guide the robots to move to the desired direction and/or desired outcome, i.e., the dream land! The human-like issues such as jealousy, sympathy, becoming number one, etc., could also be programmed into robots if it becomes necessary in performing the desired tasks more effectively. The field of artificial intelligent is a rapidly growing field and modern electronics
and miniaturized mechanical actuators have allowed the robot designers to make their robots amazingly powerful and self-supporting.
Key Questions
How much flexibility does an individual have in varying its behavior? What is the basis of behavioral plasticity and switching strategies in the context of sociality and under what circumstances should this be deterministic or stochastic?
Different factors underpin individual variation (genetic diversity, pre-adult and adult environmental conditions, learning). Do any have a greater impact on behavior? Do these factors operate on similar genes, gene networks, and physiological pathways?
What are the relative roles of genetic diversity, imprinting, and epigenetic mechanisms on regulating long-term differences in behavior?
What is the optimal balance between variation and uniformity is establishing successful social groups? (leaders vs followers, introverts/weakly connected vs extroverts/highly connected, information gatherers vs receivers, generalists and specialists)? Does this balance vary depending on the size of the group, the duration that the group remains together, the goals of the group, and the environmental conditions? For example, is more variation better in unstable environments?
In collective, group-level behaviors (swarming, migration), do individuals have inherent biases to join the group or do the collective behavior and its associated signals simply override any individual behaviors? Once in the group, are there still individual differences? Are pre- and postcollective differences among individuals conserved?
At what point does individual variation prevent integration into the group and have negative fitness consequences? How common is this? Under what conditions does this arise?
Is behavior flexibility more important for long-term versus short-term groups?
How do we measure individual variation and its consequences on groups? How do we model this?
Are robots going to become a complete human?
Are some robots more powerful than the human in some tasks that necessitate multitasking and fast computations?
Could robots perform tasks beyond what is programmed in them?
Reading
Beshers SN and Fewell JH. Models of division of labor in social insects. Annual Review of Entomology 2001;46:413-440.
Jandt JM, Bengston S, Pinter-Wollman N, Pruitt JN, Raine NE, Dornhaus A, and Sih A. Behavioural syndromes and social insects: Personality at multiple levels. Biological Reviews 2014;89(1):48-67.
Jeanson R and Weidenmuller A. Interindividual variability in social insects—proximate causes and ultimate consequences. Biological Reviews 2013;89(3):671-687, doi: 10.1111/brv.12074.
Sih A, Bell AM, Johnson JC, and Ziemba RE. Behavioral syndromes: An integrative overiew. Quarterly Review of Biology 2004;79:241-277.
Webster MM and Ward AJ. Personality and social context. Biological Reviews 2011;86:759-773.
Whitman DW, Agrawal AA. What is phenotypic plasticity and why is it important? In Whitman DW, Ananthakrishnan TN, editors. Phenotypic Plasticity in Insects: Mechanisms and Consequences, pp. 1-63. Science Publishers 2009.
Because of the popularity of this topic, two groups explored this subject. Please be sure to review the other write-up, which immediately follows this one.
IDR TEAM MEMBERS—GROUP A
Guy Bloch, The Hebrew University of Jerusalem
Iain D. Couzin, Princeton University
Delphine Dean, Clemson University
Deborah M. Gordon, Stanford University
Michael Greshko, Massachusetts Institute of Technology
Brian K. Hammer, Georgia Institute of Technology
Chen Hou, Missouri University of Science and Technology
Fushing Hsieh, University of California at Davis
Thomas Mueller, Biodiversity and Climate Research Center (BiK-F)
Oded Nov, New York University
Christopher N. Topp, Donald Danforth Plant Science Center
Michael Greshko, NAKFI Science Writing Scholar Massachusetts Institute of Technology
Team 4A was asked to generalize the principles governing the interplay between individual variation and group function.
With predictable frequency, humans—a disparate collection of diverse individuals—act collectively, whether it is a round of applause at the end of a show or an undulating “wave” of sports fans cheering at a game. From ants to naked mole rats, assemblages of diverse individuals throughout nature often navigate social environments, producing behaviors best understood on the group level. Studying a single ant in isolation, for instance, yields little insight into an entire ant colony’s maintenance regimen, but studying a teeming anthill in action does. Analyzing an individual cancerous cell may not fully reveal what happens on the level of the tumor.
These behaviors sit atop a stratum of diverse individuals that vary in their experiences, preferences, occupations, environments, personalities, and biology. How does individual variation impact group functions, including collective behavior, and vice versa?
Individual Variation and Labor Partitioning
Team 4A began by assessing how individuals vary from one another. Individuals’ genes—and how those genes interact with the environment—play an important role in helping to form a person’s predispositions. Throughout development, moreover, individuals’ social and physical environments affect their life histories by changing sensitivities to various stimuli. And from “viral” content on Twitter to unremarkable singers’ improbable ascensions to fame, random effects are also nontrivial, a principle the team pithily dubbed “the Justin Timberlake effect.”
In an effort to discuss behavioral variation in groups, conversation centered on what is known about partitioning of labor among social insects. Instead of characterizing divisions of labor as hard and fast, team consensus pointed to a more fluid concept of “task allocation,” in which sufficiently plastic individuals can change tasks in order to address a given group’s needs. However, there are costs associated with switching tasks and with “doing the wrong thing,” so in some groups, individual flexibility is sometimes disadvantageous to the group or organism.
Group Activity versus Group Function
In the pursuit of general principles, Team 4A engaged in spirited debate over what constitutes group “function,” in light of the fact that some groups appear to have identifiable, task-oriented purposes, while others do not. In an ant colony, for instance, individuals may behave collectively to defend or maintain a colony, while a school of fishes’ coordinated movements decrease individuals’ risk of being harmed. While this “function”—the collective accomplishment of a certain task—is intuitive and accessible to human observers, it may not reflect the totality of meaningful group activity, because observers may be biased toward seeing group-wide behaviors as optimal.
There are many evolutionarily stable groups in nature, however, that are far from optimized. Sparrows in groups may display different foraging strategies: Some sparrows actively seek out new food sources, while others simply wait around until new food is found, at which point they make a beeline for it. One can imagine that these different behavioral types generate appreciable group activity, and they do. The group-wide ratio between go-getter and freeloading sparrows remains stable through evolutionary time. However, the sparrows’ observed use of available resources vastly differs from the optimal resource use, making it hard to say that the group “functions” in any way other than “poorly.” For the purposes of a reliable common language, then, Team 4A considered “group activity.”
A Path Toward Generalizing: Individual, Network, and Environment
The team hypothesized that variations in group-level activity depend upon the interplay among three different entities: the individual, the social network, and the external environment in which the individuals and networks reside. To understand schools of fish, then, one must understand how individual fish can act, how a school of fish is structured through time, and how both are situated within a body of water full of resources and threats.
Individuals’ response thresholds for various stimuli—from both the social network and the nonsocial environment—depend upon the interplay between an individual’s genetic predispositions and how they interact with the individual’s world. Factors like epigenetic modification and learning allow this interplay to be embodied by the individual through development, leading to an individual having various response thresholds for various stimuli at a given time.
Not only is the individual’s experience cumulative, but past experiences
and the changes they imbue can also influence present behavior. Neglect of a child, for example, not only may affect a child’s physical and psychological health in the short term, but its lasting effect may also influence the child’s future behaviors and decisions. On the other hand, some children seem to be remarkably resilient. The explanation for this is open to more research.
As the preceding example suggests, differing interactions within social networks also introduce an important source of individual variation. An example from the microbial world is also illustrative. In a system in which individuals are largely identical at time zero—a clonal population of bacteria with managed microenvironments, for instance—the distribution of those microbes within networks creates differences in access to information, exposing the microbes to varying kinds and strengths of stimuli.
Social networks’ structural features, dependent on individuals’ ways of sensing one another, also play an important role in how individuals impact one another. Ants primarily communicate by “smelling” one another’s chemical secretions, which can persist in the environment for a considerable amount of time. A scout ant, for example, can lay down a chemical trail that leads to a food source it discovers, allowing other ants in the colony to visit the food source in the future. Schooling fish, on the other hand, rely on less persistent cues—vision and pressure waves in the water—to detect other fish in the school. These different ways of signaling among individuals likely contribute to fundamental differences in how schooling fish and ant colonies can organize themselves.
Network Features That Impact the Effect of Individual Variation
Within this conceptual framework, the team discussed how features of network structure may influence individual variation’s effects on group activity. The team focused on three prominent, intuitive features of networks: group size, network connectivity, and fluidity. Group size is the number of individuals in the group in question. network connectivity is the number and distribution of connections per node—in other words, the diversity of ways that an individual can exist and connect to other individuals within a network. Fluidity is how much and how rapidly an individual’s place within a network can change through time.
Multicellular tissues and biofilms, for instance, exhibit complex dynamics—a result of individuals’ varying response thresholds and diverse environmental stimuli—but these systems network structures are fairly static and grid-like. An individual heart cell, for instance, cannot apprecia-
bly change its place within a network composed of its fellow cells, locked into place next to its neighbors by packing and the extracellular matrix. The packing geometry, moreover, places limits on the possible number of neighbors a cell or bacterium can have: heart cells have anywhere from two to ten neighbors, simply based on how they can fit together. These systems display low fluidity and low network connectivity, as the team defined them.
Other systems studied by team members seemed to fit along the continuum of increasing network connectivity and increasing fluidity. At any given moment in time, a school of fish appears fairly grid-like, and packing geometry places limits on the number of neighbors a given fish can have in the network. However, in contrast to heart cells, schools of fish are highly fluid, with individual fish moving within the school with ease. Social insects like ants display even higher measures of fluidity within their networks while also displaying massive increases in network connectivity, introducing greater structural diversity—and greater randomness—within the network. Similarly, human online networks also display high connectivity and high fluidity. Wikipedia, for instance, enables individuals to create and modify a mind-bogglingly large number of articles, and the types of contributions an individual can make vary widely with respect to content. Within this context, the team then asked how much an individual’s variation matters in affecting group activity, hypothesizing that individual variation matters less with increases in group size, network connectivity, and fluidity. A small group of cardiac cells, for instance, will be more affected by a given individual’s response to a stimulus than a massive ant colony would be. The greater stochasticity inherent in larger, more interconnected, and more temporally fluid systems may introduce more noise to the system, rendering any one individual’s response less important.
Conclusions and Future Directions
Developing general principles for how diverse individuals generate group activity affords greater insight into how complex systems function at all scales, whether they are tumors composed of individual cells or online communities abuzz with millions of people. Greater understanding of these systems not only empowers researchers to develop better medical treatments and social institutions, but new insights also could lead to vast improvements in the function of robots and other systems of artificial intelligence.
The team’s discussions point toward a unifying conceptual framework for assessing changes in group activity, recognizing the importance of
how individuals vary, how these individuals are situated within networks, and how the external environment affects both individuals and networks. New experimental approaches could employ changes to any of these three components—alterations in individuals’ response thresholds by genetic manipulation, for example—in order to observe changes of group activity. The team also considers this framework ripe for new computational modeling of group dynamics.
IDR TEAM MEMBERS—GROUP B
Ehab Abouheif, McGill University
Yehuda Ben-Shahar, Washington University in St. Louis
Thierry Emonet, Yale University
Tim Gernat, University of Illinois at Urbana-Champaign
Lyndon A. Jordan, The University of Texas at Austin
Claudia Lutz, University of Illinois at Urbana-Champaign
Nicholas T. Ouellette, Yale University
Noa Pinter-Wollman, University of California, San Diego
Jeremy Van Cleve, University of Kentucky
Claudia Lutz, NAKFI Science Writing Scholar University of Illinois at Urbana-Champaign
Develop general principles to understand the interplay between individual variation and group function.
In collective behavior, the combined actions of the members of a group yield outcomes that differ from the sum of those same individuals acting alone—sometimes in desirable or dramatic ways. Seething populations of bacteria in biofilms jointly produce chemicals that allow them to survive in challenging environments. Colonies of bees, ants, or other social insects gather and store valuable food stores through coordinated foraging efforts. Understanding the hidden mechanics of collective behavior that enable these and other groups to function could suggest solutions to current societal challenges: What is the best way to assemble and manage a committee or a team? Could many small robotic devices be designed to work together,
in a “swarm,” to perform complicated tasks in disaster sites or dangerous environments?
One intriguing aspect of the examples of successful groups given above is how much they differ in their composition. Some groups, including communities of bacteria or body tissues comprising somatic cells, are made up of nearly identical individuals. The members of other groups, such as colonies of bees or ants, packs of wolves, or primate societies, may vary greatly in their physical traits, personality, genetic background, motivations, or other characteristics. If both groups with low variation and groups with high variation can successfully function, does this variation matter? What are its costs and benefits to collective behaviors, and do these change depending on the environment, or the task being performed?
A clearer understanding how individual variation among a group’s members affects the collective behavior of that group could lead to better management and conservation of group-living species, and the better design of teams, whether of humans or machines. IDR Team 4B’s goal was to hypothesize general principles for how individual variation affects group function.
Generalizable Terms for Addressing the Challenge
IDR Team 4B considered a broad range of examples of collective behavior, including both groups of living organisms and groups of robots. To come up with principles of group function that were generalizable across all cases, the team focused on the most basic elements of collective behavior:
Individual variation is the degree to which individual members of a group vary in their performance of a behavior. For example, in some species of ants, individuals within a colony vary greatly in size; size, in turn, affects an ant’s ability to carry food or fight predators.
Level of interactions within a group describes how much influence the behavior each member of a group has on the behavior of other members. Interactions might occur through intentional communication, such as meerkats responding to a neighbor’s alarm call, or, like birds in a flock following those in front of them, through passive signaling.
Group function is a collective behavior of a group with some measurable outcome. That outcome is expected to depend on the two terms defined above: individual variation among group members, and the level of interactions between those members.
In other words, individuals have preferences or rules for how they will behave in isolation; individuals forming a group have the opportunity to interact and influence the behaviors of other members. The combination of individual preferences and interactions between individuals will determine how the group behaves.
Considering the Simplest Case
To form hypotheses for the effect of individual variation on group function, IDR Team 4B considered a simplified case, in which the group behavior could be quantified by averaging the behavior of the group’s members. In this simple case, many factors, including the effect of the environment on group members, the ability of group members to change the environment, and past experiences of individuals, are all ignored. One hypothetical example is the movement of a school of fish within an otherwise empty pool of water. The movement of the school could be measured by averaging together the movement of all the fish.
Based on the framework suggested by IDR Team 4B, changes in individual variation (differences in velocity of the individual fish), level of interactions (how much the fish influence each other’s velocity), or both of these factors are expected to influence group function (how the school moves). The team proposed two hypotheses. First, a group with no individual variation and no interactions may produce an output that is indistinguishable from a group with high individual variation and a high level of interactions.
Consider a school of clone fish, in which each individual behaves according to the same set of rules. The fish will all go in the same direction and maintain a cohesive school. In a second school, each fish in isolation would have a unique set of rules for generating its movement, but interactions ensure that when in the school, the fish maintain a certain distance from each other. This group of “well-connected” fish will maintain a cohesive school indistinguishable from the first group of identical fish, even though the mechanisms that produced this result are very different.
Considering the dynamics of the well-connected school suggested a second hypothesis: Increased individual variation requires an increased level of interactions to maintain group cohesion. Without some type of interaction to coordinate behavior, the fish in the “well-connected” school would scatter as each followed its individual swimming preferences, and the school would dissolve. Individual variation can disrupt the stability of the group, while interactions can preserve stability.
These hypotheses, if tested, could produce useful guidelines for improving the way that groups of humans, animals, or machines work together. Hypothesis 1, for example, suggests that an engineer who is programming robots that must work together to perform a task could choose between two design options: simple, identical machines that cannot interact, or more complicated machines with a variety of abilities that coordinate their behaviors with each other. In most cases, though, there is not a simple relationship between the behavior of each individual within the group and the resulting collective behavior.
Introducing Complexity of Group Function
In real examples of collective behavior, many complex factors affect the behavior of group members and influence group function. The environment in which the group acts may be temporally dynamic or unpredictable, and may be altered by the behavior of group members. Size, organization or other properties of the group may change over time. The result of collective behavior might also be very complex; building a termite mound, for example, is not easy to break down into the behavior of many individual group members.
The team considered the impacts of these sources of complexity on the relationship between individual variation in behavior and group function, and proposed a third hypothesis: In group functions with greater complexity, increased individual variation can increase the likelihood that at least one member of the group possesses behavioral rules needed to perform that function. Combining this new hypothesis with hypothesis 2 predicts that in complex situations, just as in the simplified example, an increased level of interactions must accompany increased individual variation to maintain group cohesion. For example, if one honeybee in a colony is able to find a new flower patch, the bee cannot bring all of the nectar in that patch back to the hive by itself. However, interactions with other bees allow the information to be shared with and used by many other bees in the colony.
Pareto optimality theory, a concept from economics, has been used to describe a situation in which not all aspects of a behavior or a task can be optimized simultaneously. Pareto optimality was proposed by the team as a way to formalize predictions about specific systems in which groups engage in complex functions under various environments. In this framework, sources of complexity are modeled as constraints that limit what the group
can do. The group must divide up its resources or abilities to satisfy these constraints.
In some cases, the trade-off between multiple constraints is weak—it is easy to partially satisfy multiple constraints simultaneously. In this situation, a generalist strategy, being a “jack of all trades,” is most efficient. A group that consisted of identical generalists could do well in highly variable environments or in situations with a relatively low level of interaction. If a generalist encounters a task, it may complete it without much coordination from members of its group.
When the trade-off between constraints is strong—satisfying one constraint creates a big loss in the ability to satisfy another constraint—a generalist strategy is less efficient than a specialist strategy. The best strategy then is to invest only in maximizing the ability to satisfy one constraint. In this situation, a group with high individual variation and a high level of interaction to ensure that the group acts together could have the advantages of a generalist strategy (as it is able to satisfy multiple constraints through the individual strengths of its members), while escaping some of the costs (each individual within the group is still a specialist).
Conclusions
The hypotheses and framework proposed by IDR Team 4B can be used in the future to create predictions about the role of individual variation and interactions in specific examples of collective behavior. They provide a starting point for organizing ideas or data related to group function.
Understanding and predicting the impact of individual variation on group function is important for research efforts in both natural and human-designed systems. Ecologists and conservationists could use these theoretical tools to engineer environments that promote variation in populations of threatened species, helping those populations to survive and grow. Robots that will perform complex group functions could be designed in a more systematic way. This framework could even enable the construction of a more robust and peaceful human society, one in which better communication is used to reduce conflict and misunderstanding, and to allow us to take advantage of our great diversity of strengths and ideas.
