Nadia Drake, Science Journalist
FROM ME TO WE: ADDING THE PARTS TOGETHER, PLUS SOME SECRET SAUCE
On this planet, most things of consequence are the result of collective behavior. A solitary cell cannot squeeze blood through a valve; a single honeybee cannot craft an elaborate hive. The skies are not blackened by the undulating waves of a solo starling or a lonely locust.
One person can neither ignite nor halt global warming.
Group behaviors are what press their fingerprints into our landscapes, from the Ice Bucket Challenge to wars, from eusocial insects supporting billion-dollar crops to the destruction wrought by fire ants and termites, from having a brain that cracks codes to developing a destructive tumor.
Those groups could be clusters of cells, a colony of insects, or humans interacting on a social network. They could even be swarms of robots programmed to achieve a larger goal.
The question is, how do those behaviors emerge? Can science explain how individual parts eventually add up to something with more functionality? Do the same principles describe how groups behave on scales spanning cells to cyborgs? How do communities maintain that fragile balance between cooperation and competition, individuality and conformity?
And what triggers the collapse of collective behavior?
Scientists trying to better understand and harness such powerful forces convened in Irvine, California, to discuss these and other questions at the 2014 National Academies Keck Futures Initiative Conference on Collective
Behavior. Emerging as a theme at the meeting was the issue of scale, and the desire to define principles that can leap from microbes to humans to robots.
“Even organisms that defy any definition of organismality can display social behavior,” said keynote speaker Joan Strassmann, an evolutionary biologist at Washington University in St. Louis. Strassmann studies a social amoeba called dictyostelium discoideum. Frequently referred to as a slime mold, these eukaryotes begin their lives as single-celled organisms, then merge into complex, multicellular groups. Within those groups, Strassmann sees evidence for familial altruism and cooperative interactions with bacteria—characteristics that are normally associated with primates or humans or things that have brains.
“Really everything is social, isn’t it? If you broaden the definition enough,” Strassmann said, then defined interactive behavior as “simply any time what you do is influenced by someone else.”
During the conference, participants met in small groups to discuss nine different challenges determined in advance by the steering committee. Some of the groups were given discrete goals, while others were left to ponder seemingly unanswerable questions. For two days, teams wrestled with questions aimed toward understanding the fundamental underpinnings of collective behavior. It’s not a directional arrow that science is normally comfortable drawing: Rather than breaking a thing apart and studying how all the pieces work, the conference challenged participants to write the rules that glue those myriad pieces into a functioning whole.
Adding a bit of unintentional irony to the proceedings was the fact that any time a large number of people get together, collective behaviors are going to emerge.
Some of these behaviors were puzzling, like the long, growing lines of people waiting to squeeze onto the same conference bus—even though a perfectly good, wait-free bus was just a few steps away. Some of the behaviors were predictable, like the often-awkward group dynamics that squelch or champion ideas. And then there was the friendly, intergroup competition. “We have the inevitable emergence of team rivalry,” steering committee chair Gene Robinson observed wryly, after one team took a playful shot at their (slightly more organized) counterparts during the first set of group presentations.
Others were more amusing, like the growing popularity of Play-Doh sculptures, produced using tiny tubs of fluorescent clay left for each participant in their team rooms. At first, only a few groups included slides of their neon creations in their team updates; by the meeting’s conclusion, those
brightly colored 3-D doodles featured prominently in almost every report. Some even used the sculptures to represent behavioral strategies, such as avian brood parasitism.
In other words, the conference itself was, at times, so perfectly meta.
At the end of the meeting, some progress had been made in tackling the challenges. There is much still to be learned. But more importantly, interdisciplinary collaborations emerged among the participants, who included physicists, computer scientists, microbiologists, and primatologists. Maybe soon, we’ll have more luck infiltrating colonies of antibiotic-resistant microbes, crafting smarter machines, or designing social networks that can positively influence human interactions.
Sometimes, It’s Not As Simple As 1+1=2
The first step in studying group behavior is understanding what a “group” is—how it forms, functions, and comes apart. Tackling this problem from the bottom up meant asking Team 6 to identify the principles governing how the simplest group—a group of two—works.
Dyads are everywhere, the team noted, but they’re also fluid, with identities morphing between “one,” “two,” and “many” with regularity.
Consider the difference between “one” and “two.” It may sound obvious, but take, for example, a human. That one human grew from a dyad, the two gametes that merged during fertilization. One human may be singular, but it’s a constellation of countless cells and microbes working together to form a discrete unit. Likewise, a pair of humans could be “one,” just as many humans employed by the same company could also equal “one.” So, on some levels, a single human is both more and less than one. But on another level, one human is simply “one.”
The lines defining “one” and “two” are sufficiently blurry that Team 6 had trouble identifying characteristics of dyads applied to everything from protein dimers to merging corporations. They also wrestled with the question of whether a group of two displays truly collective behavior or is merely cooperative, and questioned whether a dyad really is the building block for larger groups.
In the end, the team settled on defining a dyad as something involving two interacting individuals, but noted that in some situations, such as organ transplants, dissolving the “two” and reverting to “one” is crucial. Understanding how dyads function, the team noted, should be possible
with technology that can track how individual cells, fish, dancers, and others behave in different size groups.
FROM ME TO ME
John Donne wrote that no man is an island. Could the same be said of cells? None of the trillions of cells in a human exists in isolation. They all contain the same basic set of genetic instructions, and yet can be vastly different from one another. So how do cells form groups that evolution can act upon, and how do those groups then become organisms?
More simply, what governs the transition from cellular me to more complex me?
In tackling this question, Team 8A presented a hypothetical situation: Say you had a bag of cells. You grab one at random. Would you be able to figure out if that cell came from a multicellular organism? Does organismality write its signature into individual cells? It might, the team argued. Cells from a human may be slightly less hardy, and definitely more specialized than a unicellular organism’s cells. The group then designed some experimental approaches to determine which elements of cells contribute to multicellularity.
Team 8B asked whether cells could be looked at as small, organic Turing machines, capable of storing information that can be read and modified by the environment. In keeping with that idea, the team turned to computer science, where tradeoffs between communication, space, and time dictate how costly it will be for a computer to accomplish a certain task. The same could be true for cells in search of energy, the team said, and plotted the positions of various organisms on a three-axis fitness landscape that evolution can act upon.
The last group considered this challenge through a more applied computer science lens, and looked how tradeoffs between speed and flexibility allow organisms to evolve. Team 8C compared multicellular assemblies, such as bacterial bioflims and tumors, to the hierarchy of hardware, software, and apps that make an iPad so flexibly functional. In the team’s analogy, the cells in both biofilms and tumors are the equivalent of the system’s hardware. Like hardware, they run quickly but are expensive to make. Operating systems, on the other hand, are cheaper but may not be exceptionally diverse. In a biofilm, bacterial cells are all running the same software that tells them how to communicate. That’s why microbial colonies can adapt so easily to changing environments. Tumors, however, are made of cells that
may not be running the same operating systems. This means they might be a bit slower to evolve, but are also much less easy to hack. Apps in this case are cheerful things like invasion and metastasis.
THE FOREST AND THE TREES
Sometimes we’re so focused on details that we miss the forest and only see the trees. But the truth is, there would be no forest without those trees—just like there would be no groups without individuals. And except in rare cases, those individuals are not identical clones of one another. They operate according to different rules and response thresholds. So, to what extent do individual differences benefit a group behavior? When do those differences harm behaviors?
Both teams challenged to answer these questions crafted complex models that began to describe how individuals affect group activity. Team 4A considered how network characteristics such as group size, connectedness, and fluidity alter the impact of individual variation on group activity. The team hypothesized that, as each of these parameters increases, individuality matters less. Take, for example, cardiac muscle—or a biofilm in which cells are packed together tightly. In these situations, a malfunctioning cell might have a greater impact on total group activity than in a school of fish, where an errant individual swimming against the grain is unlikely to derail the rest of the fish.
Team 4B also considered schooling fish as an example, and laid out three hypotheses. The first is that schools where fish all follow a different set of rules can behave similarly to highly cohesive schools where there’s no individual variation. But, Team 4B said, in a school of fish comprising dramatically different individuals, maintaining that similar outcome depends on high levels of interactions among the fish. Next, groups that are both highly interactive and include high levels of individual variation have a better chance of performing well on complex tasks in shifting environments. Finally, the team suggested that the interaction between variation and connectivity could be tested using a swarm of programmable robots.
COORDINATION, FROM CELLS TO CYBORGS
Cooperative behavior exists on every level, from single-celled organisms in microbial mats to empathetic, bipedal primates, to robots working an
assembly line. But do the same principles describe how cooperation works, regardless of whether the components have brains or not?
Team 1 grappled with the challenge of defining principles that transcend this brainy divide, using cooperative systems in organisms with brains as a model.
Or rather, coordinated systems.
Choosing to make a distinction between coordination and cooperation, the team came down in favor of defining principles governing how individual actions create collective behavior rather than why that behavior ultimately exists. Cooperation, the team argued, carries implications—such as intent—that may not apply to mitochondrial–host cell interactions or the behaviors of bacteria in a biofilm.
Coordinated actions, however, sum to produce emergent behaviors, regardless of the intent behind that action. The term bridges the differences between cells and organisms, the team said, and can be extended to describe the behaviors of artificial life forms. Governed by a set of rules, coordinated behaviors can arise from commands coded into robots, from an organism’s genetic instructions, and from the chemical gradients that direct signaling molecules to their targets.
A better understanding of how these rules allow organisms to interact with their environment should make it possible to translate organic behaviors into inorganic, brainless objects, the team said. Thus, it might be possible to transform a swarm of robots into something that acts very much like an ant or a honeybee colony, where the sums of coordinated behaviors produce spectacularly complex structures.
Cooperation is the fire over which emergent behaviors smolder, transform, and emerge. Without it, there would be no group achievement. But what is the kindling for that fire? And what lights it?
“Find the very spark that ignites cooperation,” Frans de Waal, primate ethologist and conference steering committee member, challenged Team 3. First, the team defined cooperation as an interaction that, on average, benefits participants. “You might participate in cooperation and not always win,” the team said. “On average, you will benefit from this. But any given time you don’t have to.”
Identifying the cooperative spark meant comparing a range of cooperative behaviors, from predator avoidance to pack hunting, and pulling out
the must-have proximate conditions involved in each. Among those, the team listed environmental variability, plasticity in response to interactions, communication, and goal-directed activity.
But the group wasn’t satisfied with merely creating a list. So, participants turned to evolutionary robotics and discussed an experimental framework to test how important each factor is for the development of cooperative behavior. Called the Artificial Test-bed for Experimentation into Cooperation and Helping, the experiment would involve manipulating these must-have conditions and then monitoring the fitness and performance of robot test subjects engaged in foraging or defense challenges. In particular, the team is interested in studying how the strength and type of emerging cooperative behavior varies with different signaling intensity or behavioral flexibility. Using the results of those experiments, the team would ultimately like to develop a model that can predict how organisms cooperate outside the lab.
A FRAGILE BALANCE
Eusociality, as seen in honeybee and termite colonies, is among the more successful behavior strategies, conference steering committee chair Gene Robinson said in a preconference interview. And yet, he noted, it’s a remarkably rare system. In some cases, it even appears as though eusocial organisms have reverted to solitary or parasitic lifestyles. What tipped the balance out of favor?
Two teams were asked to evaluate the extent to which successful collective behaviors rely on both cooperation and competition. Is it enough to simply have an absence of conflict? Or is some level of conflict needed for optimal performance?
Conflict is essential, said Team 7A, which began by stating that some level of conflict is needed for efficient cooperation. In trying to determine the optimum balance between conflict and cooperation, the team decided to focus on The Point of No Return—a point where the balance is so out of whack that a group is destroyed. To do this, the team took a computational approach that involved plotting various groups (chimpanzees, humans, honeybees) on a triaxial grid of conflict, cooperation, and fitness, and then identifying where those tipping points might be.
Team 7B first noted that cooperation and conflict can occur simultaneously. To better understand how a shifting balance between the two contributes to group behavior, the team decided to study instances in which
groups quickly form and dissolve—the schooling of fish, a murmuration of starlings, the rise and fall of nations. By looking at these edges of group existence, the team reasoned it could tease out the factors responsible for pushing groups in one way or another—and then begin to mine those data for factors spanning cells to societies.
NETWORKED NETWORKS OF NETWORKS
Most of the time, the human brain is the epitome of networked efficiency. It’s a mega-network of networks, and Team 9 was asked to address the ways in which cooperation and competition help those networks generate things like perception, movement, and thoughts.
The group chose to use epilepsy as a model. Seizures are the outcome of network connectivity gone rogue, where the balance between cooperation and competition has tilted. Normally, neurons live in networks that, through intricate communication channels, regulate one another. But sometimes those channels get jammed; other times, they’re too open.
Epileptic seizures occur when too many cells are talking to one another and the synaptic traffic lights are stuck on green. They’re the whole-brain manifestation of a breakdown in network communication on a microscopic level. But, as Team 9 notes, approaches to studying and treating such disorders usually consider the problem at the whole-brain level, rather than looking at the smaller scales on which networks connect. Now, the group says, technologies exist that can help bridge that gap and determine how competition and cooperation regulate connectivity on a cellular level.
Among the mechanisms the team proposed studying are competition for metabolic resources, signaling through excitatory and inhibitory networks, and trauma repair. For each of these areas, the team designed experimental approaches to investigate their contribution to epilepsy at both the whole-brain and cellular levels. Together, the team said, the results should help construct a computational model linking signaling through neural circuits to the widespread malfunction present in epilepsy.
THE $100,000,000 QUESTION
There’s a flip side to every coin. Many emergent behaviors result in positive outcomes, like accomplishing a task that no single individual could complete. But others, like the formation of terrorist networks and the
spread of destructive ideologies, are decidedly negative. In a perfect world, we’d ditch the negative and promote the positive.
Let’s say you were given $100 million, over the next five years, to study and harness the power of social networks for the public good. What would you do with that money? One might think devising a well-funded research program that could combat anti-vaxxers or mob mentalities would be a fun challenge—but neither team tackling the challenge seemed to find it particularly easy.
Team 2A began by considering various ethical angles involved the research, including privacy, data access, and whether “harnessing networks for the public good” is something that scientists are ethically able to do. The members of Team 2B struggled to conceptualize a question they deemed vast enough to merit $100 million, and began by taking a close look at different kinds of “social networks”—such as Facebook, microbial communities, or even responsive nanomaterials.
In the end, Team 2A focused on studying how social networks could increase resilience to negative influences, such as rumors or misinformation, and suggested crafting a user-generated platform for sharing scientific knowledge. Group members were also interested in studying social networks on multiple scales—individual, group, societal—and wanted to better understand how multispecies networks function (such as pollinator communication or humans working with artificial intelligence). Team 2B decided to ask whether there were unifying principles that describe how information flows through dynamic networks, and then look at how those networks, and the individuals involved, respond to perturbations. Ultimately, the team suggests, understanding such complex systems might provide insight into human behavior.
WHEN THINGS FALL APART
Just as it’s important to study how the whole exceeds the sum of its parts, it’s crucial to understand how the wholes can fall apart.
What triggers the collapse of a beneficial collective? Team 5 addressed this issue by wrestling with a problem known as the Tragedy of the Commons—the dissolution of a functioning economy wrought by individuals prioritizing their needs over group interests. Or, as summarized by the group, “Everybody could be better off without hurting anybody. And yet we don’t do that.”
It’s a problem that has confounded economists, environmental scien-
tists, criminologists, and others for decades. And the team’s challenge was immense: Solve it.
Not surprisingly, that didn’t quite happen. But Team 5 did start identifying factors that are present in different Tragedies of the Commons, such as the overuse of antibiotics (and the subsequent rise of drug-resistant microbes) and the depletion of our fisheries and pastures. The team was particularly interested in identifying variables that transcend scales and disciplines, and disciplines and might point the way toward a solution. Participants also noted that it’s important to clarify the differences between tragedy, collapse, and resilience—tragedy being the first step toward collapse, and resilience being the first step toward a solution. “How do we move from the tragedy to the resilience of the commons? How do we start to solve this big problem?” the group asked.
Ultimately, the group said, they’d like to write a mathematical equation describing how all these variable interact; though that didn’t happen at this conference, the team mentioned plans to submit a paper on the topic soon.