Honeybee Democracy - Thomas D. Seeley (2010)
Chapter 10. SWARM SMARTS
…for so work the honey-bees,
Creatures that by a rule in nature teach
The act of order to a peopled kingdom.
—William Shakespeare, Henry V, 1599
Let us now consider what lessons we humans can learn from honeybees about how to structure a decision-making group so that the knowledge and brainpower of its members is effectively marshaled to produce good collective choices. This is an important subject, for human society relies on groups to be more reliable than individuals when it comes to making weighty decisions. This is why we have juries, boards of trustees, blue-ribbon panels, and nine justices on the U.S. Supreme Court. But as we all know, groups don’t always make smart decisions either. Unless a group is properly organized, so that the face-to-face deliberations of its members result in collective reasoning that is broadly informed and deeply thoughtful, the group is apt to be a dysfunctional decision-making body. If so, then the judgments issued by the group can produce fiascoes for the affected community. Fortunately, the house-hunting bees show us a brilliant solution to the puzzle of what gives rise to good group decision making. It is a solution that has been honed by natural selection for many millions of years—fossils from the Oligocene epoch indicate that honeybees have existed for at least 30 million years—so it is certainly a time-tested method for achieving collective wisdom.
Of course, employing insects as management gurus has its limitations, and we should not blindly imitate their methods. Nevertheless, I will claim that the bees demonstrate to us several principles of effective group decision-making and that by implementing them we can raise the reliability of decision making by human groups. The latter part of this claim is not merely hypothetical, for I have applied what I’ve learned from the bees to humans, especially to my colleagues at Cornell. In 2005, just as the shape of the bees’ decision-making process was becoming clear, I became head of the Department of Neurobiology and Behavior. For fun, and as an experiment, I decided to introduce some of the ways the scout bees go about choosing a home to the ways my fellow professors and I hold deliberations in our monthly faculty meetings. Unlike the swarm bees, we do not face life-or-death decisions, but we do face ones that are difficult enough: choices about hirings, promotions, and other matters with long-term consequences for our tight-knit academic community. I am probably blissfully blind to my colleagues’ true thoughts about our collective decision making, but I think that they have been satisfied with the tough decisions we’ve made, even though things certainly haven’t always gone the way that each professor wanted. And I’d like to think that their apparent satisfaction refects how our decisions have been based on open and fair discussions. In any event, I will explain below how I have tried to put the “Five Habits of Highly Effective Groups” that I have learned from the bees to work in a university setting.
To strengthen further my case that what has been learned from the bees has relevance to humans, I will discuss some intriguing resemblances between honeybee swarms and New England town meetings in how they are organized to produce good decisions. Why use the New England town meeting as a point of comparison? It is because this special form of small town government, which has existed for more than three centuries and is arguably the most authentic form of human democracy in the world, uses a collective decision-making process not unlike that used by the swarm bees. Once a year, on Town Meeting Day—traditionally the Tuesday following the first Monday in March—the citizens in a town come together in an open, face-to-face assembly and render binding collective decisions (laws) that govern the actions of everyone in their town. Each town meeting is a fascinating blend of communal ambience and individual enterprise, as is each honeybee swarm. We will see that there are compelling similarities in the inner workings of these two proven forms of democracy. I don’t think it is a happenstance that what works well for bee swarms also works well for town meetings.
Lesson 1: Compose the Decision-Making Group of Individuals with Shared Interests and Mutual Respect
For the members of a decision-making group to work together productively, they must have a fair amount of alignment of interests so that they are inclined to form a cooperative and cohesive unit. It is also helpful if the group’s members have a fair amount of mutual respect so that they will constructively debate the proposals offered by one another, consider the other individuals’ points of view, and refrain from bruising egos and arousing anger when it comes time to critically evaluating one another’s ideas. Certainly a decision-making group composed of clashing curmudgeons is unlikely to have the morale and working relations needed to function effectively.
The house-hunting bees exemplify a group whose members have shared interests and mutual respect. Biologists now understand that the genetic success of each worker bee in a honeybee colony depends on the fate of the entire colony; no individual bee succeeds unless the whole colony survives and reproduces. Furthermore, it is now understood that the virtual absence of reproduction by the worker bees in a honeybee colony means that these bees all propagate their genes through one shared channel: the reproductive offspring of their mother queen. And since these reproductive offspring—the queens and drones produced in the spring—contain an unbiased sample of the colony’s genes, the colony propagates the workers’ genes with a high degree of fairness. So, because the workers have a common need for their colony to thrive, and because a thriving colony passes the workers’ genes into the future with near perfect impartiality, it is not surprising that the workers of a honeybee colony cooperate strongly to serve the common good.
The humans in a community rarely share a singularity of purpose like the bees in a swarm, so humans are less inclined than bees to be highly cooperative when tackling a problem they must address together. Nevertheless, there are certain things we can do to encourage ourselves to work together. One is for a group’s leader to remind the members at the outset that they all have a large stake in the welfare of the group. At the start of the annual town meeting in Bradford, Vermont, for example, the moderator—Larry Coffin, who has served in this role for 38 years and probably knows more about this job than anyone else in the state—begins the meeting in the traditional way by asking for a moment of silence “out of respect for the exercise in democracy that we are about to engage in.” This gently reminds everyone in the auditorium that they have assembled to make decisions and pass laws for their community.Similarly, at the start of the monthly faculty meetings in my department at Cornell, I generally make a few remarks reminding everyone that our overarching goal is to make decisions that will strengthen our department and so ultimately benefit us all.
A second way to foster good working relations within a human group charged with a decision-making task is to stock it with genuinely reasonable people, ones who are known to be respectful of others and constructive in their comments while at the same time good at spotting hidden problems and engaging in vigorous debate. Often, however, one cannot choose the members of a decisionmaking group. But even when the personality mix of a working group cannot be shaped, norms of behavior and procedural practices that foster good morale can be promoted. In Bradford, Vermont, for instance, Larry Coffin reminds everyone at the start of the yearly town meeting to address their comments and opinions directly to him, the moderator, rather than to other citizens. This helps keep tempers from faring and the debate moving forward. Likewise, in my department’s faculty meetings, I sometimes find myself gently cutting of demoralizing stalemates by noting when contrary viewpoints are being needlessly repeated. And twice I have had to cool an overheated exchange between two faculty members by nudging them of a personal quarrel. Such things reawaken my appreciation of the marvelous absence of corrosive relations among the debating bees.
Lesson 2: Minimize the Leader’s Influence on the Group’s Thinking
One of the most striking features of the swarm bees’ decision-making process is that it is a perfectly democratic endeavor, one in which the power is evenly diffused among all the scout bees in a swarm. In other words, the swarm bees choose their new home without a leader integrating information from different sources or telling the others what to do. Even the all-important queen, who is certainly the genetic heart of a swarm, is merely a bystander. Indeed, in many of the experiments described earlier in this book the swarm’s queen was confined in a small cage (around which the swarming bees clustered), so she was physically separated from the scout bees’ deliberations, and yet the swarm skillfully chose its new home. By operating without a leader, the scout bees of a swarm neatly avoid one of the greatest threats to good decision making by groups: a domineering leader. Such an individual reduces a group’s collective power to uncover a diverse set of possible solutions to a problem, to critically appraise these possibilities, and to winnow out all but the best one.
Unlike honeybee swarms, most human groups operate with a leader. So clearly, a prominent question we must address is how the leader of a decision-making body should behave to promote sound thinking by the group. I suggest the answer is that the group’s leader should act as impartially as possible, so that his or her influence on the outcome of the decision-making process is minimized. Only then can the group fully exploit the power of collective choice. This means that at the start of deliberations the leader should limit his or her comments to neutral information about such things as the scope of the problem, the resources available to solve it, and the rules of procedure. Also, the leader should refrain from advocating any solutions he or she would like to see adopted and instead should show an open-minded desire for fresh ideas. By functioning not as a proselytizing boss but as an impartial information seeker, the leader creates an atmosphere of open inquiry that helps the group tap its summed knowledge to assemble a wide range of possible actions. In addition to conducting meetings in a nondirective way, the leader should encourage the airing of doubts and disagreements, even ones that are critical of the leader. This fosters the free discussion and careful debate that the group will need to thoroughly evaluate its options.
If a leader shows partiality at the outset of the deliberations, or expresses displeasure if the discussion is not going in a certain direction, then he or she is likely to subvert good group decision making. One problem with both of these leadership practices is that they can lead to a premature consensus by the group as its members, consciously or unconsciously, seek to please their leader. An example of this phenomenon is the decision made by President George W. Bush and his foreign policy team to invade Iraq in 2003. As explained by Scott Mc-Clellan, deputy White House press secretary at the time, Bush’s style of leadership was headstrong. He told his foreign policy advisers of his deeply held belief that Saddam Hussein was an international pariah who possessed weapons of mass destruction and so should be removed. Evidently, Bush’s foreign policy advisers, including his national security adviser, Condoleezza Rice, went right along with his thinking as they sought to please the president. They did little to question his thinking, engage in extended debate about the possible policy options, or delve deeply into the consequences of going to war. In short, they squandered their opportunity to use group intelligence. Thus we now know that the hasty and flawed decision to invade Iraq was based largely on the gut feelings of just one man, George W. Bush.
Larry Coffin, the gentleman who has served for nearly four decades as the moderator of town meetings in Bradford, Vermont, shows us how an impartial leader can promote the emergence of a group’s collective wisdom. Although the moderator has sole authority for running the town’s annual meeting, he or she must always remember that the will of the people comes first. One way that Coffin avoids influencing the townspeople’s will is the way he starts the discussion for each question, or Article, on the meeting’s published agenda. After reading the Article—for example, “Shall the Town of Bradford purchase a fire truck for an amount not to exceed $306,000?”—Cofn asks the crowd, “What is your pleasure?” Soon a townsperson will raise a hand, Coffin will recognize him or her to speak, and the process of open deliberation on this Article is underway.
In a New England town meeting, the moderator is responsible for making sure that every registered voter is allowed to speak, that the competition among competing views is conducted fairly, and that the group makes its decisions in a timely manner. To fulfill these duties, moderators are instructed not to rely on personal authority, but instead to lean on Robert’s Rules of Order, which Major Henry M. Robert, an engineer in the U.S. Army, published in 1876 as a “guide to fair and orderly procedures in meetings.” If a town’s moderator defers to these rules, and so acts with some humility, then the general will of the town’s citizens will emerge on each problem they address.
Lesson 3: Seek Diverse Solutions to the Problem
Sometimes a problem’s architecture defines the possible solutions—to open a door, we know our choices are limited to pushing or pulling—but other times the available options are not well defined. Then the logical first step toward solving the problem is to uncover a profusion of possible solutions in the hope that one will prove excellent. And here is where a democratic group can vastly outperform a despotic individual, since a group’s power to explore for options can greatly surpass that of a lone individual. This is especially true if the group’s members are numerous, diverse, and independent. With many individuals bringing unique experiences to the problem and searching independently for possible solutions, the chances are high that someone will come up with a radically new option, which might be just what is needed.
The house-hunting bees provide a beautiful demonstration of the efectiveness of a large and diverse search committee whose members explore on their own. As we have seen, a swarm sends out hundreds of scout bees that explore for potential homesites over an area stretching five or more kilometers (at least three miles) from the bivouac site. Each intrepid scout bee works by herself, diligently poking around tree trunks and rock outcrops in search of small, dark openings that might lead to a suitably roomy and protective nest cavity. Whenever a scout chances upon a possible dwelling place, she scrutinizes it and, if it proves acceptable, she returns to the swarm and freely reports her discovery with a waggle dance. This puts an option on the table for further consideration. It is as if each scout that announces a new site says to her fellow scouts, “Shouldn’t we give some thought to this possibility, which is located X degrees to the right (or left) of the sun and Y meters away?” The distributed reconnaissance process of the scout bees often continues for hours or days, so it is not surprising that a swarm typically uncovers 10 to 20 or even more possible places to live. Clearly, the house-hunting process of a honeybee swarm is open to the widest possible array of choices, and this gives the bees a strong start in selecting the best available living quarters.
What can we humans do so that our own decision-making groups, when faced with complex problems, also develop a broad set of alternatives from which to choose? Considering what the bees do, I suggest four things. First, make sure the group is sufficiently large for the challenge it faces. Second, make sure the group consists of people with diverse backgrounds and perspectives. Third, foster independent exploratory work by the group’s members. And fourth, create a social environment in which the group’s members feel comfortable about proposing solutions. If a group implements all four suggestions, then it is likely to achieve a thorough exploration of its options.
Often one cannot shape all of these four elements of a group’s search for alternative solutions, but it will still help to improve some of them. For example, in organizing a faculty meeting in my department at Cornell, I cannot adjust its size or composition. I can, however, encourage creative thinking about possible solutions and can foster the reporting of them to the group. To help get new ideas, I will present the problem to my colleagues well in advance of the meeting. This way, each one can wrestle with it privately prior to the meeting. And to encourage everyone to contribute their ideas, I will begin the meeting by suggesting that we start tackling the problem by getting a wide range of options on the table. My colleagues are always good “scout bees,” and most are as uninhibited as a dancing bee about sharing their knowledge, so this brainstorming phase quickly yields a broad set of proposals. But to be sure that we are looking at all the conceivable options, I will ask each person who has not spoken if he or she has something to add. Often, the quieter folks will further broaden the list of options with thoughtful proposals.
Given the importance of endowing a decision-making group with diverse knowledge, it is noteworthy that Robert’s Rules of Order, which provides the exact rules of procedure for a New England town meeting, includes a nifty rule that helps ensure that every participant at the town meeting gets to express his or her thoughts on each issue; no one may speak more than twice on a particular issue until everyone who wants to speak on the issue has had an opportunity to do so once. So long as a meeting’s moderator strictly enforces this rule, nobody can dominate a discussion. This certainly boosts the fairness of a debate. It also boosts the debate’s effectiveness, because this rule helps ensure that a town meeting’s decision making benefits from the full range of facts and opinions held by its participants. Evidently, Major Henry M. Robert understood the importance to a community’s deliberations of tapping fully the collective knowledge of its members.
Lesson 4: Aggregate the Group’s Knowledge through Debate
Probably the greatest challenge faced by a group that makes decisions democratically is to know how to turn the knowledge and opinions of its many members into a single choice for the group as a whole. Indeed, this is a problem that has challenged social philosophers and political scientists for centuries. We humans have devised a variety of voting procedures to single out one option from a list of possible choices: majority rule, plurality wins, weighted-voting schemes, and others. However, the problem of social choice is not unique to humans. In many other species, the same problem arises: how should a democratic group’s members reach a decision when they strongly disagree?
The house-hunting process of honeybees provides us with an intriguing answer to this question, one that has been shaped by natural selection over millions of years. We have seen that the heart of the bees’ decision-making process is a turbulent debate among groups of scout bees supporting different options (potential nest sites). These groups compete to gain additional members from a pool of scout bees who are not yet committed to a site. Whichever group first attracts a quorum of supporters wins the competition. The winning group then goes on to build a consensus among the scouts, so that when it comes time for the scouts to pilot the swarm to its new home, they are in complete agreement about the flight plan.
What is perhaps most impressive about the bees’ system of social choice is its ability to distinguish good options from bad ones so that almost always a swarm selects the single best site from among the dozen or more possible homesites that its scout bees have discovered. And what I find most noteworthy about a swarm’s skill in decision making is how it arises from a truly ingenious balance between interdependence and independence among the debating scout bees.
The scouts operate interdependently in that they communicate with one another about their swarm’s options. This communication is crucial, because it is what makes it possible for one scout bee’s news about a dream homesite to percolate among the hundreds of scout bees in a swarm. We have seen how a committed scout can advertise “her” site to uncommitted scouts by performing a waggle dance. The uncommitted scouts who follow another scout’s dance are then recruited to the advertised site, and these recruited bees can in turn advertise the site and thereby recruit still more scouts to this particular site. Thus there is potential for runaway growth—positive feedback—in the number of scout bees visiting each site. And we have seen that the better the site, the stronger the dances advertising it, hence the greater the positive feedback for this site. Thus, by grading their dance advertisements according to site quality, the scouts adaptively bias the competition for more supporters in favor of the superior sites. And once a bias in favor of the better sites is established, it will grow and grow—the rich will become richer—as the process of positive feedback amplifies the starting bias. Sooner or later the supply of uncommitted scouts will dwindle, further intensifying the competition among groups of scouts and leading ultimately to the bees’ interest shooting up at one site while fading away at all the others. Almost always, the best site prevails in this winner-takes-all contest. The system works so well that even when the best site is discovered several hours after all the others, it can still quickly dominate the competition. Such come-from-behind success is made possible by the strongest advertising, and thus strongest positive feedback, coming from the crowd championing the tip-top site.
The interdependence among the communicating scout bees is certainly a crucial part of the social machinery by which they aggregate their many pieces of information about potential homesites. Their capacity for recruitment communication, and the positive feedback it engenders, is what gives their decision-making system the ability to concentrate its attention—the buildup of scouts—on one site. But what guarantees that the scout bees focus on the best site is one small but utterly critical piece of independence among them; each one decides whether to advertise a site, and if so how strongly, based on her own, independent evaluation of the site. No scout bee, not even one that has encountered a wildly exuberant dancer, will blindly follow another scout’s opinion by dancing for a site she has not inspected. This is critical. If scout bees were to blindly copy dancers, then their decision-making system would be prone to catastrophic amplifcations (again, through positive feedback) of errors in the reports by the first scouts who discover potential homesites. It would be much like what happened in the stock market bubble in the late 1990s, when investors bought stocks in telecommunication and technology companies based on watching what others were buying—the “conventional wisdom”—rather than on checking carefully for themselves the fundamentals of these companies. Mindlessly joining a stampeding herd, investors sunk hundreds of billions of dollars in companies that lacked solid value and eventually went bankrupt.
So, rather than perform slavish imitations of dancers, the scout bees perform judicious imitations. A scout will copy the dance that informed her of a site, but only after she has scrutinized the site herself and has concluded it truly deserves to be promoted. Thus the scout bees make use of the power of communication to help good ideas spread while at the same time they avoid the risk of creating an information cascade about an inferior site. By evaluating sites independently, they invest their attention wisely.
How can humans use what the bees have demonstrated about aggregating the knowledge and opinions of a group’s members to make good choices for the group as a whole? I suggest three things. First, we use the power of an open and fair competition of ideas, in the form of a frank debate, to integrate the information that is dispersed among the group’s members. Second, we foster good communication within the debating group, recognizing that this is how valuable information that is uncovered by one member will quickly reach the other members. And third, we recognize that while it is important for a group’s members to listen to what everyone else is saying, it is essential that they listen critically, form their own opinions about the options being discussed, and register their views independently.
These three principles will be familiar to residents of New England towns where the annual town meeting is conducted in the old-fashioned way, with citizens coming together on Town Meeting Day and making decisions through face-to-face deliberations. Just as the bees engage in courteous but freewheeling debates, so townspeople hold civil but spirited exchanges of views. Just as the bees share their knowledge and opinions about nest sites with concise dances, so townspeople contribute their facts and feelings about fire trucks, bridge repairs, and tax rates with brief speeches. And just as the bees show support (by dancing and visiting) for sites based on their independent evaluations, so townspeople show support (by shouting aye or nay, standing, or using handwritten ballots) for Articles based on their personal judgments. In both bee swarms and town meetings, the heart of the decision-making process is an open competition of ideas that are publicly shared but privately evaluated.
How do the faculty in my department at Cornell use what has been learned about the bees’ means of deliberating to make smart decisions in our meetings? First, just like the scout bees that begin their hard work by searching widely for possible dwelling places, we begin tackling a tough issue by looking broadly at our options (as described above). Next, we use the same method as the scout bees for turning the diverse pieces of information in the minds of many individuals into one chosen course of action: a friendly competition of ideas. To get the give and take of deliberation going, I usually say something like, “Well folks, I ’d like us to kick these ideas around a bit.” This works. Most of my colleagues are comfortable sharing their thoughts, and those who are quieter will be drawn into the discussion when I go around the room and ask them to share their views. One of the best things about this approach is that it lets different individuals contribute different pieces of the puzzle. One person will mention something that we’ve overlooked so far about one of the proposals. Somebody else will say he or she doesn’t understand the last person’s point, and another person will provide clarifcation. Somebody will then say there is something about one of the proposals that bothers him or her, and others will agree or disagree with this and explain why. If the meeting is going well, there is a noticeable forward movement to the discussion.
Once it feels like everything that needs to be said has been said, and my colleagues indicate that they have the information they need to make a decision, we will take a vote. In the past, we generally took votes by show of hands, but now we use secret ballots. Some people balked at first. “We’ve never done this before except on tenure decisions!” But after I explained that I really want to get each person’s independent opinion, free of peer pressure that could produce conformity, they realized that voting by secret ballot is the best way to know our true collective judgment on an issue.
Lesson 5: Use Quorum Responses for Cohesion, Accuracy, and Speed
One might think that when a democratic group has to make a decision that will apply to everyone in the group, it is best to let the group’s debate continue unhindered until the opinions of the participants have coalesced around a unified choice. After all, if a problem has an underlying correct solution, then it might pay to argue things through until everyone accepts this solution. This would both ensure that an accurate decision is made and promote broad acceptance of the decision. Sometimes, however, there isn’t one solution that serves everyone’s interests, in which case more discussion is unlikely to produce agreement and the best thing is probably to cut of a bitter debate with a vote. But even when there is an optimal solution—likely for groups whose members have common interests—it may not be worthwhile arguing on to reach complete agreement. Usually there are costs associated with investing more time in the decision process, and the accumulating costs of further debate can eventually outweigh the benefits.
The house-hunting honeybees show us a clever way for a decision-making group to make an accurate consensus decision and also save some time. Their trick is to have the scout bees make quorum responses, that is, to have these bees make sharp changes in their behavior when a threshold number (quorum) of individuals support one of the alternatives. Let’s review how this works. We have seen that the bees in a swarm must choose accurately to survive, and that they must stay together to survive, so they need to reach an accurate consensus decision about their new home. We have also seen that these bees will invest heavily, up to several days, in searching for possible homesites and publicly debating which one is the best. And we have seen that once the population of scouts at one of the potential homesites exceeds a threshold, or quorum, the scouts visiting this site will abruptly change their behavior and return to the swarm to perform piping signals. Their piping induces the many thousands of nonscout bees to warm their flight muscles in preparation for the swarm’s flight to the chosen site. This piping probably also tells the scouts from the nonchosen sites (the ones without a quorum) that they should cease advertising and visiting these sites, which in turn will speed up the consensus building among the scout bees. Thus, because a quorum of scouts at the winning site triggers key changes in the behavior of the scouts from this site, the decision-making system of honeybee swarms has a means of accelerating consensus formation once evidence sufficient to guarantee an accurate decision has accumulated at one of the sites. Brilliant! An additional benefit of the scout bees’ sharp quorum response is that it enables the thousands of nonscouts in a swarm to start their flight preparations long before the scouts reach their consensus, which further shortens the time the swarm spends hanging precariously from a swaying tree branch.
Quorum responses can also help human decision-making groups that need to find agreement do so with high accuracy and all possible speed. For example, in the faculty meetings of my department, when we face a major decision that should be resolved with a unanimous vote one way or the other—such as whether or not to recommend an assistant professor for promotion with tenure—we will take straw polls (by secret ballot) periodically during the discussion to see how close we are to consensus. If a poll reveals that we are far from unanimity, then we all know that further careful debate is needed for everyone to become of one mind. But if a poll reveals that we are close to agreement, the few folks supporting the minority position usually will realize that a collective decision has essentially been reached, that prolonging the debate is pointless, and that it is best to switch to the majority position to build the needed consensus. Thus the device of taking straw polls can give the members of a decision-making group the information they need to make quorum responses that will accelerate their consensus building. Of course, in a human group, as in a bee swarm, individuals should operate with a high threshold when making a quorum response to avoid sacrificing the accuracy of the group’s decision making. I believe this is how it works in our faculty meetings. Although I don’t know for sure, I estimate that my colleagues will change their votes (and minds?) for the sake of achieving consensus only when at least 80 percent—approximately 16 out of 20—of us are already in agreement. E pluribus unum through quorum responses? Yes, but do so carefully, using a quorum that is sufficiently large to ensure accurate decision making by the community.