Poster #21. - Drake eScholarShare

SIMULATING BELIEF PROPAGATION WITHIN A POPULATION VIA AGENT BASED MODELING
Stephen McCray and David Courard-Hauri, Environmental Science and Policy Program, Drake University
Abstract
Typical pre-run mix of opinions:
An agent-based computer model was designed in NetLogo to study the
communication of environmentally relevant scientific information in a
heterogeneous society. The roles of uncertainty, expert interpretation, and
intentional information selection in the maintenance of false beliefs even when the
agent has a personal incentive to hold beliefs that correspond to exogenous reality
were studied. The relative importance and power of these influences in the
emergence of stable or complex dynamic networks of false belief systems were
investigated. In addition, this work has implications for environmental policy and
social activism.
Results are preliminary at this point, but we can make several general conclusions.
 Most parameters lead to consensus with the current model. The consensus is
correct the vast majority (94-100%) of the time, based upon parameters tested.
 Parameters that increase potential interactions also increase the probability of
consensus, but possibly also the probability of incorrect result. In other words, the
probability of multiple stable outcomes is increased by isolation.
Introduction
We are interested in the propagation of incorrect beliefs within a population,
where individuals have access to partial, imperfect data about external reality. The
study was inspired by observations that opinions about global climate change
among the lay public have become increasingly divergent, while expert opinion on
the topic has converged quite dramatically.1 However, there are multiple other
examples of belief divergence even though only one alternative is possible. We
developed a simple model to look into this topic and investigate the specific
factors that lead people to develop a decision. We used evidence from studies on
individual choice within groups to model extremism behavior that people exhibit
when forming decisions in groups.2 By looking into factors such as extremism,
neighbor radius, and a bias factor, it is possible to observe unique emergent
behavior.
Results
 Increased extremism appears to lead to increased potential for consensus, but
there are some indications that very rapid movement toward extreme positions
may solidify entrenched multiple outcomes. More testing on the relationship
between this parameter and others is required.
Two regions of stable opinions have developed
(blue is wrong):
Methods
The core data used to generate the decision model of the agents in our simulation
is based upon the observed voting patterns of appointed judges.3 Researchers
observed that when people of like minds gather together, they are more likely to
reach more extreme conclusions than any individual. By incorporating the
extremism factor of the agents, it is possible to model the decisions that are being
made.
In our model, we assume that
individuals are trying to
determine whether the external
reality corresponds to a sine or
cosine wave. They have made
measurements, but the
measurements are imprecise and
so are normally distributed about
the actual wave. Each agent is
then able to see only some of these points, and from those points each agent tries
to determine whether they are looking at a sine or a cosine wave. Agents also have
a randomly generated bias, predisposing them to expect one or the other. To
simulate social interactions, the agent will then interact with its neighbors, find out
what they believe, and make decisions according to the makeup of the group and
the extremism concept. This will alter their bias in belief. From this point, the
agent will then reevaluate their opinion about whether they see sine or cosine.
Consensus:
Future Directions
An important question that we wish to explore is the effect of expert opinion.
Experts are like the lay public, in that they have access to only imperfect
information. However, they generally have access to more information and can
better interpret apparently conflicting results. The presence of experts ought to
increase movement toward consensus, but we hypothesize that the ability to
choose which experts to pay attention to may strengthen divergence of opinions.
We will model experts as individuals with access to significantly more data points.
Individuals will select from a pool of experts depending upon their findings, and
expert selection will give an individual access to his/her extra data points.
References
1. Doran, P. T. & Zimmerman, M. K. (2009). Examining the Scientific Consensus on Climate Change. EOS, Transactions of the
American Geophysical Union, 90, 22-23.
2. Sunstein, C.R. (2009). Goring to Extremes, How Like Minds Unite and Divide, Oxford University Press.
3. Sustein, C.R., Schkade, D., Ellman, L.M., & Sawicki, A. (2006). Are Judges Political? An Empirical Analysis of the Federal
Judiciary, Brookings Institution Press