Political Science Experimental Lab
|
|
Coordinator Professor Mathew McCubbins Department of Political Science |
Lab Description
The University of
For more information on the experimental laboratory, contact Professor McCubbins.
For a sample of experiments conducted in the lab,
see the The
Democratic Dilemma by Arthur Lupia and Mathew
McCubbins (1998, Cambridge University Press).
|
|
Political
Science Experimental Lab Data Archive PDF Slide Show
Presentation of our Work in Progress as of 12/31/2006 Movie
of our current EEG experiments (mpg-file) |
|
|
We are presently undertaking
experiments on Communication, Learning and Choice. Here is a
quick synopsis of our present work: Overcoming
Overlearning
Under
what conditions will humans change their beliefs? In many political and economic
settings, citizens must make choices based upon their beliefs about
particular phenomena. For
example, jurors must decide the fate of criminal defendants based upon their
beliefs about the defendant’s guilt or innocence and voters must choose
candidates based upon their beliefs about who will best represent their
interests once in office.
Although many citizens may hold correct beliefs about defendants,
candidates, and products in markets, recent research demonstrates that
citizens all too often hold mistaken beliefs and fail to change them
even in the face of new information.
For this reason, McCubbins and his coauthors
seek to identify the conditions under which humans will (and will not) learn
new information. To this end, he
suggest that human learning is characterized by overlearning—that
is, the tendency for humans to ignore new information and to become
increasingly fixed in their beliefs once they have learned something. Given this property of human learning
(that extends from simple pattern recognition such as the words on this page
to the adoption of a partisan identification) he asks whether and under what
conditions humans can overcome their overlearning
and change their beliefs. Specifically,
he and his coauthors are conducting behavioral, functional magnetic resonance
imaging (fMRI), and event-related potential (ERP)
experiments that assess whether the provision of immediate feedback enables
humans to recognize their mistaken beliefs and to change them. Their reason for studying the role of
feedback stems from their interest in both political and economic decision
making, as well as the differences between them. For example, when making a purchase
(such as buying a hamburger for lunch), people typically receive immediate
feedback about their choices and, therefore, know whether their beliefs were
correct or not. When making
political choices, however, people often
do not receive immediate or direct feedback about the consequences of their
choices (for example, when citizens choose a candidate for president, they
often do not immediately know what consequences their choice has for the
state of the economy or foreign policy).
Thus, it seems that overcoming overlearning
in political contexts is more difficult than in many market contexts. That said, if experiments demonstrate
that the provision of immediate feedback enables humans to overcome their overlearning, then this may suggest a number of lessons
for how we design political institutions. Indeed, it may suggest that instead of
creating forums where citizens can deliberate about political issues, that we
should instead design institutions in a way that provides citizens with
immediate feedback about their political choices. Or, it may suggest that the political
choices that we ask citizens to make must be more varied or varied more
often, which suggests that we alter our electoral system. When Are Citizens Competent Cue-Takers?
Many scholars emphasize that
citizens use cues as substitutes for detailed knowledge about politics. However, relatively few scholars
assess the conditions under which cues actually improve the decisions
that citizens make. Thus, this project analyzes experimentally whether and
under what conditions one particular cue (namely, the statements of a speaker
or endorser) enables both sophisticated and unsophisticated citizens to make
better choices than they would have made without this cue. McCubbins
and his student’s experimental results demonstrate that this cue only
improves the decisions of both sophisticated and unsophisticated citizens
when the speaker’s incentives are crystal
clear. Once the speaker’s
incentives become less transparent (as is often the case in the real world),
citizens are no longer able to consistently improve their decisions. In this way, these results demonstrate
the fragility of this cue, as only under idealized circumstances does
it consistently allow both sophisticated and unsophisticated individuals to
improve their decisions.
Among the many cues that individuals receive, which do they choose to learn from and which do they ignore? Under what conditions do they follow the better cue(s)? A Poll is Worth a
Thousand Words
One of the many cues that
citizens have available to them is information from polls. For example, in political contexts,
citizens may learn that 59% of voters favor the Democratic candidate and that
41% of voters favor the Republican candidate. Similarly, in economic
contexts, citizens are bombarded with advertisements stating that 3 out of 4
dentists recommend a particular brand of toothpaste, that 9 out of 10
consumers prefer a particular beverage, etc. Given the plethora of polls that
both candidates and companies conduct and given the widespread dissemination
of the results of those polls, it is important to ask 1) whether and under
what conditions citizens use polls as substitutes for detailed knowledge
about particular candidates or products and 2) whether and under what
conditions polls help citizens to improve their decisions. McCubbins and his coauthors address both of these questions by
using laboratory experiments. Specifically, we polled 66 UCSD undergraduates
about what they thought the answers to 10 different math problems were. We then use the results of these polls
as a key manipulation in our experiments. That is, they ask subjects in their
experiment to solve these same 10 math problems, but before they answer each
problem, they may choose to receive information from the poll that the
experimenters conducted. For example, before solving a particular problem,
subjects may learn that of the 66 UCSD undergraduates who were polled, 60
undergraduates chose answer “a,” 4 undergraduates chose answer
“b,” and 2 undergraduates chose not to answer the problem. McCubbins and his coauthors vary four different things in this
experiment. First, they vary whether the poll is helpful or not. For example, if the majority of the
undergraduates that we polled chose the correct answer to a math problem,
then that is considered a helpful poll. However, if the majority of
undergraduates that polled chose the incorrect answer to a math problem or
chose not to answer the problem at all (both of these scenarios occur in our
sample), then that is considered an unhelpful poll. Second, they vary whether
access to the polling information is costly or free; that is, in some
experimental conditions subjects must pay a cost to receive the information,
which is designed to be analogous to the opportunity costs that citizens face
when they pay attention to and seek out polling information. Third, they vary
the difficulty of the decisions that subjects must make by using math
problems of different levels of difficulty. Fourth, they examine whether and
when polling information helps both sophisticated and unsophisticated
subjects. That is, they collect subjects’ SAT math scores prior to the
experiment, so they are able to assess whether particular types of polls help
both sophisticated and unsophisticated citizens. Their preliminary results demonstrate
that polls only help citizens to improve their decisions under very narrow
circumstances and that they often lead them to make
incorrect choices. When Does
Deliberating Improve Decision Making? For decades, legal scholars, social
scientists, political theorists, and others have extolled the virtues of
deliberation, arguing that we should incorporate more deliberative practices
into our political institutions.
Indeed, deliberation is proposed as a major palliative and, for some,
even a panacea for nearly all that is wrong in society—a procedural
“cure all” for self-interested decision making, for an uninformed
citizenry, for the oppression of minorities, for social fragmentation, and
for low levels of confidence in our government, among other ills. In all of this frenzy to cure societal
ills through deliberation, however, scholars have either explicitly or
implicitly assumed their own conclusion: that deliberation has social welfare-enhancing
effects. From a cognitive science
perspective, the conclusion that deliberation improves social welfare is
peculiar. Indeed, much of the
cognitive science literature causes us to
question not only the welfare-improving effects of deliberation, but also the
core assumptions upon which theories of deliberation rest. Specifically, scholars who advocate
deliberation assume that the participants
will speak to each other, listen to each other, respect each other, and then
learn from one another’s statements. As the cognitive science literature
reminds us, however, speaking, listening, and learning are all costly
behaviors; that is, when we speak or listen to one person, we must forego the
opportunity to do something else.
Further, because humans have limited energy, they are able to pay attention
to and remember only a small fraction of the information available to them.
What these cognitive limitations imply for successful deliberation is this: Whatever the relationship between idealized
theories of deliberation and social welfare, deliberation in practice is
unlikely to improve social welfare because it is improbable that groups of
people will be willing to speak, listen, and learn from one another. In order to assess whether and
under what conditions deliberation can improve social welfare, McCubbins and his coauthor conducted a series of
laboratory experiments. These
experiments represent a simple setting in which discussion among equally
motivated individuals would unquestionably enhance both individual and social
welfare. Thus, the experiment’s
design is a close analogy to scholars’ theories of deliberation. Specifically, their experiment
incorporated the costs associated with speaking and listening in a collective
setting because subjects are allowed to exchange information with one another
throughout the experiment, but they must sometimes pay a small fee to speak
and listen (ranging from $2 to $0.25).
Their experimental results demonstrate that when speaking and
listening are costly, deliberation does not improve social welfare, but
rather, decreases it. They also
investigate whether deliberation has different effects on social welfare in
small versus large groups. Their
results demonstrate that even in groups as small as 4 people, deliberation
does not improve social welfare.
Further, when the number of individuals in a group increases to 8, 9,
10, 12, or 15 participants, then deliberation again brings about significant
declines in social welfare, with larger groups suffering larger declines. |
This web site is made possible through funding from the National Science Foundation and the Public Policy Research Project
| Public Policy Research Project |
Mathew D. McCubbins || UCSD Home Page |
| Political Science Department |