chapter 2: how social psychologists conduct research Flashcards
what is the observational method of research
Researcher observes people (research is based on making observations of other people)
-Used to describe behavior
-Realistic information (why they are good)
two examples of observational research
ethnography and archival analysis
what is ethnography
Description from an “insider’s point of view”
-Try to immerse yourself into the culture of others to gain information about others you might not otherwise be able to get
-Ex. People join cults to learn more about the cult
what is archival analysis
Researcher examines accumulated documents (archives)
-Interested in how the weather impacts voting behavior, looking in previous years to see the weather that day and how people actually vote
what are three limits of the observational method
1) Some behaviors difficult to observe
-Occur rarely
-In private
-Ex. If someone was interested in sexual behavior this is hard to gain access to because most people are not willing to let you see them having sex
2) Archival analysis
-Missing info
-Record keeping is often not done perfectly, incomplete information, could be variables you are interested in that you cannot access
3) does not show causation
what is the correlational method
Two or more variables are systematically measured and the relation between them is assessed
how can you do correlational method with one group
Only one group – measure two things at the same time
what are the two kinds of correlations you can find
positive and negative
what is the correlation coefficient
A statistical technique that assesses how well you can predict one variable from another.
what two things must you consider when looking at a correlation coefficient
- strength: indicated by absolute value
-Absolute value near 1 – strong correlation, good prediction
-Absolute value near 0 – weak correlation, poor prediction - direction: indicated by sign
-Positive – as one variable increases the other increases
-Negative – as one variable increases the other decreases
what is a survey (typically used for in correlations)
-Representative sample of people asked about attitudes or behavior
-Correlations computed using responses to questions
when are surveys the best/ most representative
Best when random selection is used
-Random selection is when each person in a population has the same probability of ending up in your sample
-But you have to be careful
-We almost never have true random selection which can create problems
what can happen if you do not use a representative sample and give the FDR / Landon example
The Perils of using a Nonrepresentative Sample
-Failure to Use Random Selection = Misleading Results
Overwhelmingly the polls showed that Landon would win the election,
-1936 – polled people from phone directory and car registrations and randomly selected people off of the list, showed that Landon was definitely going to win
-Why did this happen? People who had more money had phones and cars and people who have more money tend to vote for republicans, not an accurate sample to pull from (not representative of the population)
-Probably would not happen today because most people have cars and phones nowadays
what is good about correlations
-Realistic
-Can be cheap and easy
-Can investigate things you otherwise could not
Ex. The correlation between smoking and lung cancer showed that smoking can cause lung cancer, never can randomly assign people to a smoking condition because of the IRB
what is bad about correlations
Typically self-report
-People are biased and not always honest when they self-report, various motivations for lying
Cannot demonstrate causation
-You can never know when there is a confounding variable (another variable that is causing the two variables being measured or that explains the relationship between the two of them, ex. Stress and smoking and lung cancer)
-You can never tell which variable is causing the other, which direction the correlation is
explain correlation does not equal causation and the three possible reasons behind a correlational relationship
Three possible causal relations when a correlation is found (e.g., you find that cigarettes smoked is correlated with lung cancer)
-Smoking causes lung cancer
-People with lung cancer smoke more
-Not caring about health leads to…
-Smoking
-Cancer
(direction of relationship or a third variable causing both)
explain the birth control and STD example and another explanation
A study found a correlation between the type of birth control women used and their likelihood of getting a sexually transmitted infection (STI). Those whose partners used condoms were more likely to have an STI than were women who used other forms of birth control. Does this mean that the use of condoms caused the increase in STIs? Not necessarily.
-Could be that: People in committed relationships are less likely to use condoms because they are only sleeping with one person and might use other forms of birth control
what is an experimental design and its three key features
Research design with 3 key features (you need these three things to have a true experiment)
-Manipulation of IV
-Measurement of DV
-Random Assignment of participants to groups (this allows for control of outside variables as well)
Only way to demonstrate that one variable causes another variable
Ex. You flip a coin to determine if people will exercise in a group or by themselves for a week. At the end of the week, you give them a satisfaction with life survey.
internal validity in experiments?
Internal validity is typically high in experiments (if the experiment is measuring what it is supposed to be measuring/ does it demonstrate causation)
-Can control variables
-Randoms assignment
external validity in experiments?
External validity is often low in experiments (whether or not the findings can be generalized to other populations or situations)
-People
-Situations
why is external validity low in experiments
It is low because in real life things are not as controlled as they are in experiments, there are a lot more things going on in life that might impact the results (a lot of experiments do not resemble real life); also is low because of the kinds of people that are participating in the studies (selection bias, the sophomore problem is people who need to participate in studies to get credit)
how can you deal with low external validity (4 ways)
-psychological realism
-replications
-meta-analysis
-field experiments
what is psychological realism and how does it help external validity
Using Psychological Realism (this is the extent to which the variables in which you are interested in are actually operating in your experiment/ the extent to which you are actually manipulating the variables you are interested in studying, ex. If you are interested in studying fear you have to create a situation that is impactful where people actually feel fear); try to create situations as similar to the real world as possible
-Perhaps with a cover story
-Ex. People interested in drinking behavior will set up their lab to look like a bar
explain replications and how they help external validity
No research results rest on one particular study, people study the same thing over and over so we can have more confidence in what we are finding
explain meta analysis and how they help external validity
-Where you take a bunch of studies looking at the same phenomenon and analyze all of the results together, the participants in your study are other studies that people have done
-These things give you a good idea about how strong the findings are in a whole body of literature
explain field experiments and how they help external validity
Getting out of the lab and doing experiments in the environment in which your variables operate
why does random selection not work to help external validity
(unfortunately, not by using random selection)
-This would be the best way to solve this issue (using true random selection), but this is very seldomly done because it is essentially impossible to do, you would need a list of everyone in the country or in the world and pulling random people off, this is typically not feasible
explain replications and why replication studies rarely occur in psychology
Replications:
-Repeating a study, often with different subject populations or in different settings
-Ultimate test of external validity
Replication crisis in psychology
-Few people actually try to do replications
-No incentives for this (does not typically get published)
-And when they do, many findings don’t replicate
-Especially for social psychology
explain the open science collaboration and the replication crisis (why things dont replicate)
Open Science Collaboration
-Wanted to reproduce research findings on a large scale
-Tried to replicate 100 recent research findings
-Used original materials, contacted original authors
-97% of original studies had significant results
-36% of replications had significant results
why did the studies not replicate in the replication crisis
-Biases in reporting and publication (studies with significant results are more likely to get published in the first place, people without significant results they are not likely to get published); we only know how many significant findings there are not how many insignificant findings there are
-Think of alpha level, there is a 5% chance of finding a result that is significant even if there is not a relationship, the type 1 error will still get published because it found significant results
There are also changes in people overtime, ex. The changes in technology overtime can change people as well
P-hacking: doing what you need to do to get a significant result
-Altering data, taking participants out of your study,
-Happens a lot at universities where you have to publish high quality research or else you lose your job
how can we help the replication controversy
-Questioning things even if they are published, pushing for more replication studies, look at things in totality and not rely on individual studies
-A lot of journals require you to pre-register your study with them so you cannot alter or change data, register hypotheses ahead of time
-Paying more attention to non-significant results
what is good about experiments
-Demonstrates causation
-can control variables
what is bad about experiments
-Expensive and time consuming
-May not resemble real world
what is a quasi experiment
Quasi-Experiment (more similar to correlational studies than experiments)
Uses existing groups
-Anytime you are using any pre-existing groups this is a quasi experiment, if you are not doing random assignment they are not experiments
Lacks at least one key feature of experiments
-Manipulation of IV
-Random assignment
-But still have measurement of DV
Cannot identify causation
ex. You ask people if they exercise by themselves or in a group. Then you give them a satisfaction with life survey.
what is good about quasi experiments
-Realistic
-No one can question whether or not things work this way in the real world
-Can be cheap and easy
-Can investigate things you otherwise could not
what is bad about quasi experiments
Cannot demonstrate causation
what are the two goals in conflict with one another in social psych (ethical dilemma)
1) Create experiments that resemble the real world and are well controlled
-Social psychology is different than cognitive psychology because a lot of the topics we are interested in are relevant in the real world
2) Avoid causing participants stress, discomfort, or unpleasantness
-A lot of the things that happen in the real world can lead to these results
how to we settle the ethical dilemma in social psychology
Protect participants:
-Informed Consent
-Allow participants to discontinue
Create realistic situations
-Good design
-Deception: Must be necessary; Must not be something people would care about; Must debrief; Usually people don’t mind; Research shows that people do not care when deception is involved ; Not all social psychology involves deception
what kick started the guidelines for ethical research
-The syphilis study was incredibly unethical, which led to the creation of the Belmont report which helped establish rules and regulations for future research
-Men who had syphilis were not given treatment for it, and were not aware that they had it, research had the point of seeing the long term effects of the disease
what are the guidelines for submitting ethical research now
Submit to Institutional Review Board (IRB)
-Started in early 1970’s
-For institutions receiving federal funds
-One scientist, one nonscientist, and one person who is not affiliated with the institution
explain Facebook’s mood experiment
-Facebook has an algorithm that influences what you see in the newsfeed which changes all of the time
-Tweaked people’s Facebook feeds, two groups (emotional contagion: if you are exposed to certain moods or emotions are you more likely to take those on), some people saw more posts from friends that included negative words, some people saw more posts from friends that included positive words; measured people’s posts that they made or status updates
-Had a computer program that would go through all of the status updates and see if you were exposed to more negative words did you have more negative words in your status updates (found that this was true and happened)
-Found evidence for emotional contagion
-Got away with this by having people sign the terms of use which stated you can participate in any research they conduct
-Not good from an ethical standpoint
