Award-Winning Researchers Work to Predict Human Behavior and Emotions

[arizona.edu] Turning speculation into predictability is the objective of a human
decision-making model University of Arizona researchers have developed
– an effort that has earned international recognition for the team.

Young-Jun Son, an associate professor in systems and industrial
engineering at the UA, worked with one of his doctoral students,
Seungho Lee, to create a computer model to predict how groups and even
individuals will react during emergency situations, such as a major
fire or a bombing.

"Individual behavior is not like ant behavior," Son said. "We want
realistic human behavior. And if you look at our model, it is
applicable for individual behavior."

Theirs is the type of research that can help law enforcement and
first responders make better decisions and improve real-time planning
and decision making during a crisis.

"We thought we could make a model that would mimic human decision
making," Lee said. "It’s not just about assuming what would happen, but
we want to understand the psychological behaviors, including the
emotions people would have, including fear and fatigue. And we claim
that we can."

Son and Lee took the Best Paper Award in the area of homeland
security at the Institute of Industrial Engineers Annual Conference
this month for their paper, titled "Integrated Human Decision Behavior
Modeling Using Extended Decision Field Theory and Soar Under BDI
Framework." Son and another team of his students earned the same award
in 2005.

The international institute, the world’s largest professional
society dedicated to industrial engineering, allowed 700 papers to be
presented during the conference in Vancouver, Canada.

"To homeland security, this is very important," Son said. "I think that’s why we got a lot of attention from the judges."

Lee, a doctoral degree candidate in systems and industrial
engineering, took the conference’s Best Ph.D. Scientific Poster Award.

Researchers have for some time studied intelligence, cognition and
how humans make decisions, but have tended to focus on one of three
different models: engineering, psychological and economical. Yet Son
and Lee created a model that integrates them all in an attempt to
advance the way researchers evaluate both human decision-making
processes and planning.

In their paper, Son and Lee expanded on the Belief-Desire-Intention
system, which is used to predict the future actions of humans.

"We believe that in the details of the BDI there are a lot more
details that can address planning and human characteristics," Son said.
"BDI is the guideline structure so as time goes on, we need to add more
and more human structures so that 20 or 30 years later our human
decision model will get closer to the human model."

For now, the model "is very abstract," Son said. "We needed to develop a model with a lot more detail."

During the UA team’s analysis, individuals were asked to record
their perceptions of different emergency situations in the Cave
Automatic Virtual Environment, a configuration of screens at University
Information Technology Services.

Using the CAVE allowed the researchers to study the ways that
participants would react to situations such as the presence of smoke or
a fire.

The team then tried to determine how actions would change under
certain conditions, such as the presence of police of the intensity of
the crowd.

Son and Lee used the data collected from the answers to make a
virtual model, mapping individual perceptions to create a group
environment in a given situation.

The researchers then loaded the responses into a computer simulation
model, testing different configurations that would not only determine
how groups would react but also how individuals would react. For
example, a computer considers a number of factors and determines the
direction in which people would run after they came to an intersection.

Son and Lee said additional research is required to understand the
exact emotional response people would have in any given situation and
the team continues to work on the project. For instance, they want to
study in greater details how an individual’s confidence relates to them
doing what they plan to do, or changing their mind.

Such a model would help determine how many police officers would be
needed, how to predict evacuation times, how to estimate the number of
casualties and how to get information to survivors.

In the paper, the team wrote: "The proposed simulation has a
potential to allow the responsible governmental and law enforcement
agencies to evaluate different evacuation and damage control policies
beforehand, which in turn allows the execution of the most effective
crowd evacuation scheme during an actual emergency situation."

Source: http://lqp.arizona.edu/node/231