Você está na página 1de 30

Summary

Experimentation in Social
Psychology
AR ON SO N , W IL SO N A N D B R E W E R (199 8)

Y LO N A M AK
S 234 950 7
Introduction
This article is about the workhorse of social psychological research; the
experimental method.
The experimental method: The researcher randomly assigns people to
different conditions and ensures that these conditions are identical except for
the independent variable (the one believed to have a causal effect on
peoples responses).
Two main missions:
Discuss why the experiment is the method of choice, advantages and
disadvantages.
Provide detailed instructions in how to do it for those new to this method.
Why do experiments?
Valuable insights can be gained by careful observations of everyday behavior. But
this study has many limitations, when journalists report the results of their
observations it is difficult to gauge the typicality of what they report.
Observational versus Correlational versus Experimental studies
Observational: Naturalistic behavior is systematically observed and recorded
Correlational: Two or more variables are systematically measured, and the
relationship between these variables is assessed. If there is a correlation between
the variables, then we can predict one from the other Correlation does not prove
causation
Experimental: Controlling all factors except the independent variable and
randomly assign people to condition Causal relationship can be determined
with much greater certainty
Why choosing an Experimental Design:
The ability to control variation to insure that the stimuli in experimental conditions
are similar
The ability to randomly assign people to conditions
Random assignment is the great equalizer: as long as the sample size is sufficiently
large, researchers can be relatively certain that differences in the personalities or
backgrounds of their participants are distributed evenlhy across conditions.
It is impossible to measure all variables in a correlational design

Field versus Laboratory Research:


Field Studies Laboratory Studies
People are unaware of being observed Conducted in a laboratory, people know they
are being observed
Advantage: behavior is best understood if Advantage: to gain enough control over the
observed in a naturally occured context situation to be able to make causal inferences
Mostly used for correlational studies Mostly used for experiments
Why Laboratory research for experimental designs?
To gain enough control over the situation to be able to make causal inferences
A critical value of experimental designs is that the situation is identical for all people except for
the independent variables of interests, this is more difficult in naturalistic settings
Laboratory makes it easier to random assign people to conditions
Laboratory permits researcher to manipulate independent variables more precisely and to
elimnate or minimize the intrusiveness of extraneous variables.

Problem-Oriented versus Process-Oriented Research


Basic research: investigators try to find the best answer to the question of why people behave
the way they do, purely for reasons of intellectual curiosity. No direct attempt is made to solve a
specific social or psychological problem. -> Interested in the mechanisms underlying an
phenomenon (process-Oriented)
Applied research: to solve a specific problem, constructing theories about why people do what
they do and find ways of alleviting them. (Problem-Oriented)

Proces-Oriented studies do not study the original problem at all, focusing instead on general
mechanisms that produce many different effects. To really get at the heart of a problem, to
understand its cuases (experimental), process-oriented studies are often the method of choice.
Planning and conducting a
laboratory experiment
The best way to describe how to conduct an experiment is to take a real study and
dissect it carefully, examining how it was done and why it was done that way.
Choosing the type of experiment to perform:
The first decision you would want to make is whether to design your experiment for
the laboratory or the field. The next decision is whether the experiment is to be an
impact or a judgement type.
Impact: people are active participants in an unfolding serie of events and have to react to
these events as they occur
Judgement: participants are more like passive observers; they are asked to recognize,
recall, classify, or evaluate stimulus materials presented by the experimenter.
Judgement experiments are usually easier to do, because they require a less
elaborate setting of the stage to involve the participants in an impactful situation.
The four stages of laboratory
experimentation
1. Setting the stage for the experiment
2. Constructing the independent variable
3. Measuring the dependent variable
4. Planning the postexperimental follow-up.
1. Setting the stage
In designing any laboratory experiment, a great deal of ingenuity and
invention must be directed toward the context, or stage, for the manipulation
of the independent variable. It not only must be consistent with the
precedures for presenting the independent variables and measuring their
impact but also can and should enhance that impact and help to justify the
collection of data.
Many experiments involve deception: the setting must include a sensible,
internally consistent pretext or rationale for the research as well as a context
tht both supports and enhances the collection of the data and reduces the
possibility of detection -> Cover story
In a well-designed experiment, the cover story is an intricate and tightly
woven tapestry. In a judgement experiment, the cover story is typically less
elaborate and more straightforward than in an impact experiment.
2. Constructing the independent
variable
The independent variable is the experimental manipulation. It is a variable
that is independent of all sources of variation except hose specifically under
the control of the experimentor.
The experimenter begins with the conceptual variable: a theoretically important
variable that he or she thinks will have a causal effect on peoples responses
Any characteristics that the participants bring to the experiment cannot be
regarded as independent variables. In internal analysis needs to be done to
test and check the manipulation.
Whether to manipulate the independent variable on a between or within-
subject basis?
Between-subject design: people are randomly assigned to different levels of the
independent variable
Within-subject design: all participants receive all levels of the independent variable
often preferred
Avoiding participant awareness: the challenge of doing experiments in social
psychology lies in the quest to find a way to circumvent or neutralize the
theories that the participants walk in with so that we can discover their true
behvior under specifiable conditions rather than being left to ponder
behavior that reflects nothing more than how the subjects think they should
behave in a contrived attempt to confirm their own theory.
Demand characteristics: features introduced into a research setting by virtue of
the facts that it is a research study and that participants know that they are part of it.
Optimizing the impact of the independent variable:
Present instructions not too briefly
Ensuring that instructions are expressed precisely
Ensure that participants attend throughout the course of the experiment to the
relevant stimulus
Ensure the right amount of pretesting that is required to have the desired impact on
the independent variable
Ensure that the hypothesis is right

Choosing the number of independent variables: an experiment should be only


as complex as is required for important relationships to emerge in an
interpretable manner.
3. Measuring the dependent
variable
Decision to rely on participants self-reports or observations by others as the
means of assessing a persons responses to the experimental situation.
Behavioroid measure: close to the behavioral end of the contiumm would be a
mesure of the participants commitment to perform a particular action without
actually performing it.
Social psychologists often measure other things than actual behavior: (1)
convencience (2) problem or process orientend (3) intervening processes
Disguising the measure: disguise the fact that a particular collection of data is
actually the measurement of the dependent variable.
Dependent measures that are uncontrollable: dependent measures that by their
very nature are uncontrollable, obviating th eneed to disguise measure.
4. Planning the postexperimental
follow-up
The experiment does not end when the data have been collected. The
experimenter will want to remain with the participants to talk and listen in
order to accomplish four important goals:
1. To ensure that the participants are in a good and healthy frame of mind
2. To be certain that the participants understand the experimental
procedures, the hypotheses, and their own performance so that they gain
a valuable educational experience as a result of having participated
3. To avail themselves of the participants unique skill as a valuable
consultant in the research enterprise
4. To probe for any suspicion on the part of the participants, such as whether
they believed the cover story
Validity and Realism in
experiments
Experimentation almost always involves a trade-off between competing goals: control and
realism.
Internal validity: the extent to which acausal conclusion based on a study is warranted, which is
determined by the degree to which a study minimizessystematic error(or 'bias').
External validity: the validity of generalized (causal) inferences in scientific research, usually
based on experiments as experimentalvalidity. In other words, it is the extent to which the results of
a study can be generalized to other situations and to other people. Mathematical analysis of external
validity concerns a determination of whether generalization across heterogeneous populations is
feasible, and devising statistical and computational methods that produce valid generalizations.
Construct validity: the degree to which a test measures what it claims, or purports, to be
measuring."In the classical model oftest validity, construct validity is one of three main types of
validity evidence, alongsidecontent validityandcriterion validity.Modern validity theory defines
construct validity as the overarching concern of validity research, subsuming all other types of
validity evidence.
Mundane versus Experimental
versus Psychological realism
Mundane realism: the extent to which experimental findings can be generalized to
the real world.
Experimental realism: if the situation is involving to the participants, if they are
forced to take it seriously, if it has impact on them
Psychological realism: a particular literary method in which a piece of fiction
focuses on the interior motives, psychological processes, and characters' mental
narratives instead of simply telling a story. Psychological realism focuses on why
something happens and not just the scenario that plays out. It focuses on the mental
processes of the characters and often includes their inner thoughts and feelings along
with their motives for behavior
Ethical concerns in
experimentation
The field of experimental social psychologists is constructed in an ethical
dilemma. This is due to two sets of values to which most social psycholgists
subscribe:
A belief in the value of free scientific inquiry and a belief in the dignity of humans and
their right to privacy
A researcher must decide on a course of action after giving careful
consideration to the importace of the experiment and the extent of the
potential injury to the dignity of the participants.
Ethical problems arise even in the absence of either deception or extreme
circumstances. The use of deception, combined with the possibility of self-
discovery, presents the experimenter with a special kind of ethical problem.
Field experiments

Advantages:
Research setting in a real shopping situation,
shoppers in their natural environment (more
external validity)
Consumers unaware they are being observed (if
conducted properly!)
Disadvantages:
Less experimental control: field experiments
typically come with problems with alternative
explanations/confounding variables (less internal
validity)

2/17/17 | 16
2/17/17 | 17

Lab experiment: chocolate


Studying the choice overload hypothesis in a
controlled situation (see Iyengar & Lepper, 2000; study 3
for a similar set-up):
Lab study manipulation: single participants exposed
to different number of flavors (6 vs. 30)
Truly random allocation across conditions
Dependent variables:
Social presence kept constant
Desirability of choosing (enjoy decision)
Product satisfaction (more satisfied with
choice)

2/17/17 | 17
Random allocation to conditions
Allows for experimental control
Given big enough sample, groups
should be equivalent on any
dimension
Any resulting group differences
are caused by manipulation,
rather than external factor
Random assignment allows
researchers to make causal
inferences
2/17/17 | 18
2/17/17 | 19

Lab experiments

Core features: manipulation and random


assignment
Advantage: experimental control/internal
validity
Allows for inferences about causality
Disadvantage: realism may suffer from
experimental control (external validity)
This forces us to be creative!

2/17/17 | 19
2/17/17 | 20

Creative experiment

How is food consumption affected by other


consumers around us?
Specifically: how does the body type of and
food consumed by others affect how much we
eat?
2/17/17
McFerran, Dahl, Fitzsimons, and Morales (2009), Journal of Consumer Research, 36 (April),
| 20
915-29.
2/17/17 | 21

Creative experiment

From the RQ, you can already identify the


relevant variables:
How does the body type of and food
consumed by others affect how much we
eat?
IVs: body type and food type
DV: quantity eaten
Two IVs rather than one
How does that work?
2/17/17 | 21
McFerran et al. (2009, JCR): Study design
2 (body type: thin vs. obese) x 2 (food: healthy
vs. unhealthy) between-participants factorial
design
Independe
2 independent variables with each 2 levels
Independen nt variable
t variable 1 2

Body Type Food

Thin Obese Healthy Unhealthy

Levels Levels

2/17/17 | 22
McFerran et al. (2009, JCR): Study design
Factorial design: every level of every IV crossed
with every level of every other IV
All possible combinations included
Food

Healthy Unhealthy

Body Thin 4
Type experiment
Obsese al
conditions

2/17/17 | 23
2/17/17 | 24

Creative experiment

Participants watched another participant


(actually a confederate!) take a large quantity
of food while watching movie
Manipulation of food: confederate took large
portion of M&Ms or granola in view of the
participant

2/17/17 | 24
2/17/17 | 25

Creative experiment

How do we manipulate confederate body type?

In doing so, how do we avoid confounds?


Confounds threaten internal validity: does effect on DV
result from my manipulation, or from somewhere else
Solution McFerran et al (2009): Using the SAME person
across body size conditions

2/17/17 | 25
2/17/17 | 26

BMI =
19

BMI =
33

2/17/17 | 26
2/17/17 | 27

Findings

Comparison
s made
between
experiment
al
conditions

2/17/17 | 27
Experimental designs
2 (body type: thin vs. obese) x 2 (food: healthy
vs. unhealthy) between-participants factorial
design
Between-participants design: every participant
serves in only one condition
Within-participants design (repeated measures
design): every participant serves in more than
one (or all) conditions

2/17/17 | 28
Experimental designs
2 x 2 design: two main effects and 1
interaction effect possible
Body type main effect: people likely to
emulate thin confederate
No interaction effect: the above is true
regardless of type of food confederate
consumes

2/17/17 | 29
Experimental designs
Interactions make predictions more specific:
they tell us when a particular effect will occur
(and when it will not occur) -- they add
specificity
E.g.: celebrity endorsers are more effective
than non-celebrities, but ONLY when
observers are not involved

2/17/17 | 30

Você também pode gostar