 |
A Computational Model of Political Cognition: The Dynamics of Candidate Evaluation
| |
| | Unformatted Document Text:
Introduction
Many scholars have recognized the need to incorporate into our explanations of individualpolitical behavior psychological knowledge about how people obtain, store, retrieve, anduse information to make judgments (e.g. Boynton, 1995; Lodge et al, 1995; North, 1990;Ostrom, 1998; Simon, 1985). Important theoretical progress has been made in politicalscience to address this need, most notably the on-line processing model (see for example,Lodge, Steenbergen, and Brau, 1995; Lodge and Taber, 2002) and the memory-basedmodels (Zaller & Feldman, 1992; Tourangeau, Rips, and Rasinski, 2000).
However, given recent developments in cognitive science and political psychology, our
current models of political judgment are at best partial and limited. Consequently, theydo not adequately account for many well-known phenomena repeatedly found in cognitivescience and political behavior research. For example, while the on-line processing modelis capable of explaining why citizens often remember how much they like or dislike acandidate, it does not provide an account of the oft-noted question order or questionwording effects in the survey response. Similarly, while memory-based models can be usedto explain ’response effects’, they fail to account for such well-known regularities as primingand the finding that citizens can often form and express attitudes without being able to sayhow or why they feel as they do. We believe that a more comprehensive model of humanmental processes is necessary to improve our understanding of ordinary citizens’ politicaljudgments and voting behavior.
In this paper, we present a more comprehensive theory of human mental processes that
incorporates both cognitive and affective mechanisms and both on-line and memory-basedprocessing. We develop this theory axiomatically and then formalize it as a computationalmodel. Using the ACT-R architecture (Anderson et al, 1990, 1997, 2002)
1
, we build a
computational model of political information processing, named John Q. Public (JQP). Wethen demonstrate the internal validity of the model (i.e., run simulations to ensure that themodel properly represents the processes described in our theory) and show that JQP canreproduce several well-known empirical regularities from survey-based electoral researchand experimental studies on political cognition, namely cognitive and attitude primingeffects, question order and wording effects, and the integration of campaign informationinto voter preferences.
The first section of the paper presents and discusses a set of axioms and correspond-
ing mechanisms (algorithms) that underlie our computational model. The second sectiondescribes the computational experiments and discusses results. In the conclusion, we sug-gest possible future directions in developing and applying the model, including sensitivityanalysis (comparative statics), empirical verification, and more complex simulations.
1
ACT-R is written in Common Lisp and is currently one of the most popular architectures in artificial
intelligence. Generally speaking, it is a set of libraries for building a model. It provides a set of classicalcognitive mechanisms but lacks affective mechanisms. Consequently, most of the programming work wasdevoted to building affective mechanisms into the model. For more detail about ACT-R, see Anderson etal (1990, 1997, 2002).
2
|
| | Authors: Kim, Sung-youn., Lodge, Milton. and Taber, Charles. |
|
| |
|
|
Introduction
Many scholars have recognized the need to incorporate into our explanations of individual
political behavior psychological knowledge about how people obtain, store, retrieve, and
use information to make judgments (e.g. Boynton, 1995; Lodge et al, 1995; North, 1990;
Ostrom, 1998; Simon, 1985). Important theoretical progress has been made in political
science to address this need, most notably the on-line processing model (see for example,
Lodge, Steenbergen, and Brau, 1995; Lodge and Taber, 2002) and the memory-based
models (Zaller & Feldman, 1992; Tourangeau, Rips, and Rasinski, 2000).
However, given recent developments in cognitive science and political psychology, our
current models of political judgment are at best partial and limited. Consequently, they
do not adequately account for many well-known phenomena repeatedly found in cognitive
science and political behavior research. For example, while the on-line processing model
is capable of explaining why citizens often remember how much they like or dislike a
candidate, it does not provide an account of the oft-noted question order or question
wording effects in the survey response. Similarly, while memory-based models can be used
to explain ’response effects’, they fail to account for such well-known regularities as priming
and the finding that citizens can often form and express attitudes without being able to say
how or why they feel as they do. We believe that a more comprehensive model of human
mental processes is necessary to improve our understanding of ordinary citizens’ political
judgments and voting behavior.
In this paper, we present a more comprehensive theory of human mental processes that
incorporates both cognitive and affective mechanisms and both on-line and memory-based
processing. We develop this theory axiomatically and then formalize it as a computational
model. Using the ACT-R architecture (Anderson et al, 1990, 1997, 2002)
computational model of political information processing, named John Q. Public (JQP). We
then demonstrate the internal validity of the model (i.e., run simulations to ensure that the
model properly represents the processes described in our theory) and show that JQP can
reproduce several well-known empirical regularities from survey-based electoral research
and experimental studies on political cognition, namely cognitive and attitude priming
effects, question order and wording effects, and the integration of campaign information
into voter preferences.
The first section of the paper presents and discusses a set of axioms and correspond-
ing mechanisms (algorithms) that underlie our computational model. The second section
describes the computational experiments and discusses results. In the conclusion, we sug-
gest possible future directions in developing and applying the model, including sensitivity
analysis (comparative statics), empirical verification, and more complex simulations.
1
ACT-R is written in Common Lisp and is currently one of the most popular architectures in artificial
intelligence. Generally speaking, it is a set of libraries for building a model. It provides a set of classical
cognitive mechanisms but lacks affective mechanisms. Consequently, most of the programming work was
devoted to building affective mechanisms into the model. For more detail about ACT-R, see Anderson et
al (1990, 1997, 2002).
2
|
|
 Convention | | Need a solution for abstract management? All Academic can help! Contact us today to find out how our system can help your annual meeting. | | Submission - Custom fields, multiple submission types, tracks, audio visual, multiple upload formats, automatic conversion to pdf. | | Review - Peer Review, Bulk reviewer assignment, bulk emails, ranking, z-score statistics, and multiple worksheets! | | Reports - Many standard and custom reports generated while you wait. Print programs with participant indexes, event grids, and more! | | Scheduling - Flexible and convenient grid scheduling within rooms and buildings. Conflict checking and advanced filtering. | | Communication - Bulk email tools to help your administrators send reminders and responses. Use form letters, a message center, and much more! | | Management - Search tools, duplicate people management, editing tools, submission transfers, many tools to manage a variety of conference management headaches! | | Click here for more information. |
|
|
|
| |
|
|
|
