Contrary to the ideal of a completely engaged electorate, individuals who have the
least interest in a specific outcome can actually be vital to achieving a democratic consensus. These
individuals dilute the influence of powerful minority factions who would otherwise dominate everyone else,
according to new research published in the journal Science.
A Princeton University-based research team reports Dec. 16 that this finding — based on
group decision-making experiments on fish, as well as mathematical models and computer simulations —
can ultimately provide insights into humans’ political behavior.
The researchers report that in animal groups, uninformed individuals — as in those with
no prior knowledge or strong feelings on a situation’s outcome — tend to side with and
embolden the numerical majority. Relating the results to human political activity, the study challenges the
common notion that an outspoken minority can manipulate uncommitted voters.
style=”width: 508px; height: 347px” />
Princeton University-led researchers found that uninformed individuals — those with no
strong opinion or prior knowledge — promote democratic consensus by diluting the power of a
strongly opinionated minority that would otherwise dominate group decisions. Because uninformed individuals
have no feelings on a situation’s outcome, they side with and embolden the numerical majority. The
researchers saw this dynamic play out with golden shiners, a strongly schooling fish. The researchers
trained a large number of groups to swim toward a blue target, while smaller groups were trained to follow
their natural predilection for a yellow target (right image). When placed together, the large trained group
would follow the smaller group to the yellow target. When fish with no prior training (the uninformed
individuals) were introduced, however, the fish increasingly swam toward the majority-preferred blue target
(left image). (Image by Science/AAAS)
“The classic view is that uninformed or uncommitted individuals may allow extreme views to
proliferate. We found that might not be the case,” said lead author http://icouzin.princeton.edu/”
target=”_self”>Iain Couzin, a Princeton assistant professor of http://www.princeton.edu/eeb/”
target=”_self”>ecology and evolutionary biology. He and his co-authors, including
engineering professor Naomi”>http://www.princeton.edu/~naomi/”>Naomi Leonard,
found that even a small population of indifferent individuals act as a counterbalance to the minority —
whose passion even can cause informed individuals in the majority to waver — and restore majority
rule.
“We show that when the uninformed participate, the group can come to a majority decision
even in the face of a powerful minority,” Couzin said. “They prevent deadlock and
fragmentation because the strength of an opinion no longer matters — it comes down to numbers. You
can imagine this being a good or bad thing. Either way, a certain number of uninformed individuals keep that
minority from dictating or complicating the behavior of the group.”
Of course this effect has its limits, Couzin said. He and his co-authors also found that if the
number of uninformed becomes too high, a group ceases to function coherently, with neither the majority nor
the minority taking the lead. “Eventually, noise dominates because there just aren’t enough
informed individuals to guide the group,” he said.
In a related paper, another Princeton-led team translated the computer simulation, which churned
through many variables, to mathematical formulas that can approximate the same behavior. Naomi Leonard, the
lead author of the paper, published online Dec. 19 in the Proceedings of the National Academy of Sciences,
said the goal was to describe the findings in a way “that captures the essential dynamics but
reduces out some of the complexity.”
Leonard, the Edwin S. Wilsey Professor of Mechanical and Aerospace Engineering, said that deriving
mathematical formulas and substituting them for heavy computing allowed researchers to conduct a deeper
investigation into the role of variables such as the size of the uninformed population or the strength of
interactions between individual decision makers. By describing the underlying mathematics, scientists can
not only derive behavioral results, but can also predict how changing variables will affect the final
outcome.
She said that such a detailed examination was “difficult to do when the model is complex
but important to getting at a fundamental understanding of the mechanism.”
Parallels to humans
An important aspect of the findings, said Couzin, is that they are based on experiments on groups
of fish, as well as mathematical models and computer simulations. Though the idea of uninformed populations
benefiting the democratic process seems counterintuitive, the experimental results suggest that this dynamic
is a naturally occurring decision-making process, he said.
The experiments involved golden shiners, a fish prone to associating the color yellow with a food
reward, Couzin said. The researchers trained groups of golden shiners to swim toward a blue target, while
smaller groups were trained to follow their natural predilection for a yellow target. When the two groups
were placed together, the minority’s stronger desire for the yellow target dominated the group’s
behavior. As fish with no prior training (the uninformed individuals) were introduced, however, the fish
increasingly swam toward the majority-preferred blue target, the researchers report.
style=”width: 508px; height: 597px” />
The researchers constructed models of animal groups with a majority and a minority population, each
with a differing preference to move in a certain direction. How strongly the respective groups felt about
their preference could be increased or decreased. As the preference of the minority group became more
intense, the preference of the majority became less likely to be realized until the minority won out every
time. (Image by Science/AAAS)
“We think of being informed as good and being uninformed as bad, but that’s a
human construct. Animal groups are rarely in a fractious state and we see consensus a lot,” said
Couzin, who studies the behavior and communication behind animal movement, swarming and flocking.
“These experiments indicate there is an evolutionary function to being uninformed that
perhaps is as active as being informed,” he said. “Animals may be equally adaptable to
simply going with the majority in certain circumstances because having that quick decision-making capability
is beneficial for survival. We shouldn’t think of it as a bad thing, but look at advantages animals
exhibit to being uninformed in natural circumstances.”
Donald Saari, a professor of mathematics and economics at the University of California-Irvine who
studies voting systems, said he sees parallels to the Princeton-led work in markets and politics.
Highly informed economic forecasters and political activists frequently lose out to the masses of
consumers and regular voters who base decisions on personal preferences and reasons more than on expertise,
said Saari, who is familiar with the Science report but had no role in it.
For instance, he said, the arc from minority domination to pluralism to the potential degeneration
into “noise,” as described in the Princeton study, can be seen in the American electoral
system.
A forceful minority can dominate in circumstances that attract the more politically inclined, such
as midterm elections and primaries. In more popular elections, however, that influence wanes as less
passionate people participate. Situations in which a candidate’s personality or personal life takes
precedent over policy positions in voters’ minds could be an equivalent to the breakdown in
direction Couzin and his co-authors found when there is a glut of uninformed individuals, Saari said.
“This study gives us a new interpretation of group decision making that really flies in
the face of previous opinions. We usually assume that a highly opinionated and forceful group is going to
sway everyone,” Saari said.
“What we have we here is something very different,” he said. “It doesn’t
say whether or not the consensus it good, it just provides a way of understanding when and how the consensus
changes. If the numbers of the uninformed, or people who don’t have a strong opinion, are large
enough, that dilutes the effect of the highly opinionated or knowledgeable in the final outcome. Quite
frankly, I think it’s because the highly opinionated are not in the center and the uninformed, to a
large extent, are.”
Saari said that there might be an additional consideration or factor that uninformed individuals
bring to the group process rather than mere devotion to the majority opinion.
“These results raise a lot of questions for me and present another way of thinking about
and coming up with explanations for what we observe in group dynamics,” he said.
“I think the effect the uninformed have is much more than just number-counting plurality
and that they’re offering something else,” Saari said. “Why are the fish with no
‘opinion’ more effective toward taking the group toward plurality than the fish that only
had some opinion? What is that additional dynamic, what are the real contributions of the uninformed? I don’t
know what it is, but I do know it’s worth investigating.”
The power of the uninformed in simulations and reality
The researchers developed three models that initially revealed and described how uninformed
individuals restore popular power. The modeling work was based on a computational tool developed in Couzin’s
lab that predicts and explains animal group behavior based on various forms of social interaction among
group members. Couzin http://icouzin.princeton.edu/wp-content/uploads/file/PDFs/Couzin%20et%20…
target=”_blank”>first reported the model in the journal Nature in 2005.
For the current work in Science, Couzin worked with, from Princeton, second author Christos
Ioannou, a former postdoctoral fellow in Couzin’s lab who is now a research fellow at the University
of Bristol. In addition to Leonard, co-authors also included postdoctoral researcher Colin Torney and
doctoral student Andrew Hartnett, both in Couzin’s lab; and professor http://www.princeton.edu/~slevin/”
target=”_self”>Simon Levin, the Moffett Professor of Biology and co-author of the
2005 Nature paper. The team also included Güven Demirel, a researcher at the Max Planck Institute
for the Physics of Complex Systems; Thilo Gross, an engineering lecturer at the University of Bristol; and
Larissa Conradt, a visiting researcher at the University of Cambridge.
style=”width: 508px; height: 436px” />
The models showed that even the presence of one or two uninformed individuals caused an immediate
change in the group’s behavior, even in groups with the most adamant minority. This figure shows the
effect of the uninformed on the four animal groups by strength of the minority opinion, with black being the
least intense and red indicating the most. In the red group, the majority nonetheless took back control with
less than 10 uninformed individuals present. Around 20 uninformed individuals, however, all four groups
begin to experience “noise,” which the researchers describe as a state where too many
uninformed individuals result in a lack of leadership. (Image by Science/AAAS)
In this project, Couzin used his model to first simulate animal groups of different sizes with a
majority and a minority population, each with a differing preference to move in a certain direction. He
added the factor of how strongly the respective groups felt about their preference, a variable he could
increase or decrease.
As expected, the researchers report, if the majority’s preference was just as strong or
stronger than the minority’s, the group moved in the direction the majority favored. But when the
intensity of the minority’s preference increased, the animals as a whole frequently caved to that
group’s desires. In the groups with the strongest minority preference, the animals always went with
the minority.
Couzin then added a third group, the uninformed, that had no preference on the direction to move.
The model showed that even the presence of one or two uninformed individuals caused an immediate change in
the group’s behavior. The uninformed individuals were ultimately most effective in the groups with
the least committed minority and those with the smallest total number of members. But even in groups with
the most adamant minority, the majority took back control with less than 10 uninformed individuals present.
“Consensus naturally emerges in these models once uninformed individuals are introduced,”
Couzin said. “There is a sharp transition from minority to majority control. At a certain
threshold, only a few uninformed individuals can alter the entire outcome of group decisions.”
Mathematical models — one created by Demirel and Gross, another by Torney —
helped explain the mysterious pull of the uninformed individuals. These models were based on social
processes in human groups, such as how conventions become established, or how people influence each other’s
opinions, Couzin said.
The calculations indicated that during the decision-making process, all individuals have a tendency
to follow what they perceive as the predominant view, but opinionated individuals are more resistant to
social pressure, Couzin explained. This reluctance to compromise manipulates the perception of what is
popular, meaning that the strong convictions of the minority can make their view seem dominant. Uninformed
individuals, having no strong opinion or preference, tend to inhibit this process because they respond
quickly to numerical rather than semantic differences and curb the influence of forceful individuals.
style=”width: 508px; height: 444px” />
The researchers modeled different sizes of animals groups, always with the majority (N1) having one
more individual than the minority (N2). Uninformed individuals had the strongest and more immediate
influence in smaller groups. Again, disorder sets in when the number of uninformed individuals gets too
high. (Image by Science/AAAS)
The models were used to design the experiments with the golden shiners, which Ioannou, who was not
aware of the hypothesis being tested, conducted over a three-month period. The majority group of fish
trained to swim toward the blue target consisted of six fish; five fish made up the strongly “opinionated”
minority group, which was driven by a natural attraction to the color yellow.
As in the simulations, the minority group won out when uninformed individuals were not present and
the fish swam toward the yellow target in slightly more than 80 percent of the trials where only the
minority and majority groups were present.
The untrained fish, however, which were introduced in groups of five or 10, consistently put the
group on course toward the blue target, Couzin explained. When five were added, the whole group went toward
the blue target half the time. In trials with 10 untrained fish present, the fish made their way to the blue
target nearly 70 percent of the time.
“We saw that the counterweight to a powerful minority can come from the least expected
population — the uninformed,” Couzin said.
“It was extremely rewarding to see this counterintuitive prediction play out in reality
with living organisms,” he said. “Our work is a simplification of reality, but it allows
the underlying mechanics of this type of decision making to be observed and understood.”
The research was supported by grants from the National Science Foundation, the Searle Scholars
Program, the U.S. Navy Office of Naval Research, the Royal Society and the Defense Advanced Research
Projects Agency administered by the U.S. Department of Defense.
