Social Psychology: Research Methods

One Problem with Social Phenomena

Results always seem obvious/explainable post hoc (after the fact)

Hindsight Bias

Tendency to overestimate ability to predict outcomes that have already occurred

One Way to Avoid Hindsight Bias

Making predictions about the future (not so easy/obvious in practice)

Hypotheses

Testable predictions about the relationships between two or more variables

- Mostly based on previously established theories

- Can also be based on initial observations, news reports, etc.

Good Theories

Explain previous-and predict future-outcomes

- Often refined as new findings emerge

Descriptive Research

Often involves a summary of a single variable without indication of its association with other variables

- Involves use of observational methods

What do researchers do in descriptive research?

Observe and systematically record behaviour of target individuals

Ethnography

When the research joins the group and observes its behaviour from the inside to eliminate biases

Unobtrusive Measures

Better minimizes bias in participants

Examples of Unobtrusive Measures

The electronically activated hearing device (EAR) and body cam footage

Operational Definition

Specification of exact, objective, and measurable behaviours/responses that are indicative of certain behaviours like politeness or friendliness

Interjudge Reliability

High level of agreement between judgements of ratings from at least two judges-indicates that the coding is not just based on subjective impression

Archival Method

Examination of accumulated documents and archives of culture/group

Examples of Archival Research

Diaries, novels, magazines, newspapers, social media posts

Big Data

Large samples of research

- Social media platforms are a source of this for archival research

Correlational Method

Predicting behaviours and examining relationships by measuring 2 or more variables

Strong Relationships

Allow for better predictions about one variable from another

Positive Correlation

Higher values on one variable are related to higher values on the other variable

Negative Correlation

Higher values on one variable are related to lower values on the other variable

Strong Correlation

The pattern is tighter and clearer, regardless of direction

- Linked with strong relationships

- r is closer to +1 or -1

Weak Correlation

Forms loose, unclear patterns

- Predictions are less accurate

- r is closer to 0

Correlational Coefficient

A statistic that captures both the strength and the direction in a single value

- Denoted by "r"

- Ranges from -1 to +1

Surveys

Most common approach in correlational research

Benefits of Surveys

Can ask about behaviours/attitudes that are otherwise difficult or impossible to observe

- Easier to obtain representative samples

Representative Samples

Samples that better reflect the composition of the population of interest on certain variables of interest

Random Selection

Ensures that everyone in the population has an equal but random chance of being included in a study

- A way to obtain representative samples

Problems with Surveys

- Survey samples are sometimes not representative

- Lack of accuracy in certain responses (difficulty in accurately predicting behaviour or underreporting negative behaviours)

Major Limitation of the Correlational Approach

We don't know WHY the variables are related

Directionality Problem

We don't know which variable is causing what

Causality/Third Variable Problem

A third, unmeasured variable could be causing both variables

Causal Language

e.g. increases, improves

- Used in experiment research and should be avoided in correlational research

Experimental Method

Explains and manipulates behaviour

- Only way to test causal relationships

Bystander Effect

The phenomena in which an individual is less likely to help a victim when more bystanders are present

Random Assignment

Ensures that participant characteristics are independent of the conditions

- Participants characteristics can still differ across conditions using this, but it becomes less probable with larger samples

Independent Variable

The variable that is hypothesized to influence the dependent variable; participants are treated identically except for this variable (e.g. the number of bystanders is the IV in the experiment on the bystander effect)

Dependent Variable

The response that is hypothesized to depend on the independent variable; all participants are measured on this variable (e.g. helping behaviour is the DV as any differences between conditions would mean that it is dependent on the number of bystanders)

Predictors vs Outcome

The IV and DV in correlational methods

p values

Probability that there is no real effect and that the difference observed occurred by chance alone

- Researchers rely on these to determine the probability that their findings occurred by chance rather than due to manipulation (p < .05 is a typical threshold)

Problem with the Experimental Method

Even if you use random assignment, you can still jump to false conclusions

Confounds

Other features that may explain the effect rather than the IV

- Occurs if more than one feature differs across conditions

Internal Validity

Tight control in which only the variable of interest differs across conditions which better explains cause and effect

Limitations of the Experimental Method

- Greater internal validity in the lab often make situations more contrived (less reflective of the real world)

- Greater internal validity also threatens external validity

External Validity

Extent to which the results of a study can be generalized to other people and other situations

Psychological Realism

Features/designs that trigger the same thoughts, feelings, etc, that would occur in real life

- Sometimes maintained by hiding the true purpose of the study and using cover stories

Replication

Repeating a study with different populations or settings

- Increases generalizability of results

- The top psychological journals (96%) are from WEIRD countries and represent only 12% of the world

Meta-Analysis

Analyzing average effects across different studies to see if there is an overall effect

- Increases generalizability of results

Field Experiments

Conducting experiments in the real world

- Increases generalizability of results

- Cheap car vs expensive car driver aggression (Doob, Gross, 1968)