Social Psychology: How Other People Shape Judgment and Action

You disagree with a proposal but hear everyone else endorse it. You see a person stumble, notice that nobody reacts, and hesitate. Social psychology asks how other people, group expectations, and the structure of a situation influence what we think and do.

Learning goals and Feynman practice

Explain attribution, distinguish conformity from compliance and obedience, interpret classic experiments, and identify mechanisms that can inhibit or support helping. Separate the observed result from its proposed explanation and from the popular story told about it.

Read a section, close it, and teach it to a beginner. Give a fresh example and one alternative explanation. Repair the missing step before opening the hidden answers. Do not rehearse the ethically problematic classic experiments on other people.

1. Social psychology: the situation includes other minds

Social psychology scientifically investigates how thoughts, feelings, and behaviour are influenced by the actual, imagined, or implied presence of others. An audience can matter even when no one is physically present: imagining a future evaluation may change how you prepare.

The discipline examines both individuals and their contexts. Showing that a situation influences behaviour does not establish that personality is irrelevant or remove responsibility for harmful actions.

Attribution: what caused the behaviour?

A dispositional attribution emphasizes a person’s tendencies or characteristics. A situational attribution emphasizes circumstances, constraints, incentives, or roles. Someone’s late arrival might reflect poor planning, a cancelled train, or both.

The fundamental attribution error, often discussed alongside correspondence bias, concerns overinterpreting behaviour as evidence of a corresponding disposition while insufficiently accounting for context. Related terms do not capture every identical process, and the pattern varies across tasks, cultures, and available information.

In the quiz-master study, randomly assigned questioners could generate questions from their own knowledge. Contestants lacked that selection advantage. Ratings nevertheless favoured the questioners’ apparent knowledge. The lesson is that performance can reflect control of the task as well as ability; a fair comparison must account for both.

One influential processing account proposes an initial trait inference followed by effortful situational correction. It is a useful model, not a rule that every trait judgment is automatic and every contextual inference slow. Goals, salience, learning, and cognitive demands affect processing.

Teach it back: Why would randomly assigning the quiz-master role fail to make observed quiz performance an unbiased intelligence test?

Self-check: distinguish assignment from equal opportunity

Random assignment helps separate pre-existing characteristics from the role. But the role itself gives control over which questions are asked, creating unequal opportunities to display knowledge. The performance difference cannot be interpreted as a pure ability difference.

2. Three kinds of social influence

Form	Main feature	Example
Conformity	Adjusting to a group norm or perceived consensus.	Giving the group’s answer despite initial disagreement.
Compliance	Responding to a request.	Agreeing when a neighbour asks for help.
Obedience	Following an instruction from perceived authority.	Carrying out a supervisor’s directive.

These categories can overlap in real settings. Agreement does not establish private belief: someone may publicly comply while privately dissenting.

3. Conformity: information and belonging

Informational influence involves using others as evidence about what is correct. Normative influence involves fitting in, maintaining relationships, or avoiding rejection. Both can operate at once.

Sherif: constructing a norm under ambiguity

In the autokinetic task, a stationary light in darkness can appear to move. With little external reference, estimates are uncertain. Group judgments can converge, illustrating how other people’s answers become information and help establish a norm.

Convergence alone does not mean everyone knowingly gives a false answer. The ambiguity is part of why a group estimate can be persuasive.

Asch: consensus against a clear comparison

In classic line-judgment conditions, a participant heard confederates unanimously choose an incorrect comparison. Many participants conformed on some critical trials, but many responses remained independent.

Figures around three quarters conforming at least once and roughly one third of critical-trial answers conforming describe different denominators. They are not universal rates, and “at least once” does not mean consistently agreeing with the group. Procedures and samples matter.

Moderator	Typical lesson	Qualification
Group size	Increasing a small unanimous majority can increase pressure.	The pattern often levels off; there is no fixed threshold for every task.
Dissenter or ally	Breaking unanimity can support independence.	Agreement with the participant is not the only possible way to break consensus.
Private response	Can reduce public conformity.	Privacy does not remove informational influence.
Cultural setting	Norms and interdependence can affect responses.	A country or cultural label does not determine an individual’s behaviour.

Teach it back: Why could private responding reduce conformity without proving that the original answers reflected only fear of rejection?

Self-check: one manipulation can affect multiple processes

Privacy changes evaluation and accountability, but may also alter attention, expectations, or how independently participants consider the evidence. A reduction supports a role for public response conditions without uniquely isolating every motive.

4. Compliance: the foot-in-the-door effect

In Freedman and Fraser’s classic 1966 research, a small initial request could increase later agreement to a larger one. The road-safety sign example has several conditions, including differences in the initial activity. Do not combine a petition condition with a small-sign condition and treat one percentage as the result for every version.

Possible mechanisms include self-perception, commitment, consistency, and changes in how the request is interpreted. The finding does not establish an irresistible need for consistency or guarantee success whenever requests escalate.

A person may infer, “I support local safety initiatives,” after agreeing to a small action. A later request can then fit that self-description. But size, cost, topic, timing, and perceived manipulation affect the outcome.

Teach it back: Why would more agreement after a small request not by itself identify consistency as the only cause?

5. Obedience: Milgram and the limits of extrapolation

Milgram’s studies asked participants acting as teachers to administer apparently increasing shocks when a learner made errors. The learner was a confederate and the shocks were not actually delivered, although participants encountered a highly stressful deceptive situation.

In a prominent condition, 65% continued to the maximum labelled level. This is a result from one procedure and sample, not a population constant or a count of people proven to enjoy harming others. Participants differed in distress, interpretation, trust, and willingness to continue, and some refused.

Variations changed proximity, institutional context, experimenter presence, and peer behaviour, producing different patterns. The findings support investigating authority, gradual commitment, perceived legitimacy, and responsibility rather than claiming a single irresistible obedience reflex. A synthesis across Milgram conditions examines this variability. [1]

Burger’s partial replication

Burger’s 2009 study used screening and additional safeguards and stopped at the 150-volt decision point. About 70% in its base condition were willing to continue beyond that threshold. [2] It did not observe 70% delivering the maximum 450-volt setting.

No statistically detected sex difference in a small study does not prove the sexes are identical under every condition. Ethical safeguards and procedural changes also limit direct comparisons with earlier samples.

Teach-back check: why cannot a 150-volt stopping point establish the percentage who would reach 450 volts?

Continuation at an earlier point is a different measured outcome. Predicting later behaviour requires assumptions, and the modified procedure may affect the decisions. An unobserved endpoint should not be reported as an observed result.

6. The Stanford Prison Experiment: a lesson in evaluating evidence

The 1971 mock-prison study assigned male volunteers to guard and prisoner roles and ended after six days rather than the planned two weeks. Some participants displayed abusive behaviour or substantial distress. The familiar account claims that roles alone rapidly transformed everyone.

Archival analyses challenge that inference. Researcher involvement, guard instructions, demand characteristics, participant selection, uneven behaviour, and the absence of adequate comparison conditions limit what the study establishes. Le Texier’s critique and research on the guard orientation document important problems. [3, 4]

It is misleading to describe all guards as spontaneously sadistic or all participants as uniformly absorbed by their roles. The volunteers were not simply a representative sample of Stanford students, and random role assignment did not remove the researchers’ influence.

The study remains important for research ethics and the history of social psychology. It does not cleanly prove that roles automatically overwhelm personality, or establish the primary cause of abuses in a different real-world institution. Investigating harmful institutions requires evidence about leadership, norms, accountability, incentives, and individual actions.

Teach it back: Why does random assignment fail to isolate the role’s effect if the investigators also coach the role’s behaviour?

7. Bystander intervention: several obstacles can look like indifference

The bystander effect concerns reduced individual helping under some conditions when other people are present. It is a probabilistic, context-sensitive pattern, not a law that groups never help. Meta-analytic evidence identifies important moderators, including danger and the role other bystanders can play. [5]

Mechanism	Question the observer is implicitly answering	Possible barrier
Pluralistic ignorance	Is this actually an emergency?	Everyone uses others’ outward calm as evidence despite private concern.
Diffusion of responsibility	Am I the person who should act?	Each assumes another person will take responsibility.
Evaluation apprehension	What if I misunderstand and look foolish?	Concern about social judgment inhibits action.

These mechanisms differ. They cannot be inferred solely from the absence of help, and danger, skill, access, and physical constraints also matter. An individual rate of helping is also different from the probability that at least one person in a larger group helps.

The Genovese account needs correction

The murder of Kitty Genovese became associated with a widely repeated story of 38 witnesses watching and doing nothing. Historical analysis challenges that simplified account. [6] It should not be taught as an uncontested description of events. Laboratory evidence about helping should stand on its own methods, not depend on a dramatic historical narrative.

The intercom study

Darley and Latane manipulated how many other participants a listener believed were present during an apparent seizure communicated through an intercom. Believing others were available could delay reporting the emergency or reduce intervention.

The crucial variable was perceived group presence, not simply a crowd physically blocking the participant. The result supports examining responsibility allocation, without proving that every nonhelper lacked empathy.

Teach it back: How could everyone privately feel alarmed while each interprets the others’ hesitation as evidence that nothing is wrong?

8. The Good Samaritan study: time pressure can alter helping

Darley and Batson studied seminary students travelling to give a talk, with different levels of induced hurry. Some talks concerned the Good Samaritan; not all did. Students encountered a person apparently in need along the route.

Reported helping was about 63% in the low-hurry condition versus 10% in the high-hurry condition. These figures describe the study’s conditions and small sample, not an individual’s fixed probability of helping. Time pressure mattered more clearly than the assigned talk topic under that procedure.

The result does not establish that values never influence helping or that rushing removes responsibility. It shows why assessing the situation can improve explanation beyond declaring helpers good people and nonhelpers bad people.

9. Worked example: a silent project meeting

Everyone appears to endorse a risky plan. One explanation is agreement; another is normative pressure. Members may also assume that others know more, or privately disagree while interpreting silence as consensus.

Collecting independent judgments before discussion can make the distribution visible. Inviting a contrary view may reduce unanimity pressure. Explicitly assigning responsibility for checks can reduce ambiguity about who should act.

These are applications of the mechanisms, not guaranteed interventions. Compare actual outcomes and consider whether the group needs better information, safer disagreement, clearer authority limits, or a different decision process.

Retrieval workshop and spaced review

From memory, distinguish attribution, conformity, compliance, and obedience. Reconstruct the three helping barriers and explain why they predict similar behaviour for different reasons. For each classic study, state the comparison, outcome, and one limitation.

Repair: situations cause behaviour, so personality and responsibility do not matter

Situational effects show that context contributes to behaviour. People still differ, interpret circumstances, resist, and make choices. Explaining a contribution is not equivalent to excusing harm or denying agency.

Repair: most people conformed at least once, so most responses were conformist

Participant-level and trial-level proportions have different denominators. Someone can conform once while remaining independent on many other trials. Report which measure the percentage describes.

Apply: how would you distinguish informational from normative influence?

Compare conditions changing ambiguity, public visibility, and group expertise while measuring judgments and private beliefs. No single manipulation perfectly isolates motives, so use converging evidence and control alternative explanations.

Review tomorrow and in a week. Score each explanation: 0 for a label, 1 for a definition, 2 for a mechanism, example, and limitation. Repair the weakest explanation first.

One-minute explanation: Other people shape interpretation, agreement, requests, authority, and helping. Context can produce behaviour that observers mistakenly attribute entirely to character. Social influence is powerful but variable, and different mechanisms can produce similar outcomes. Classic studies need their procedures, denominators, ethical issues, and later critiques before their findings are generalized.

Sources and further reading

The supplied citations lacked a bibliography. These research papers support the qualifications and later evaluations; meeting examples and self-checks are teaching aids.

1. Meta-Milgram: an empirical synthesis of the obedience experiments.
2. Burger, J. M. (2009). Replicating Milgram: would people still obey today?.
3. Le Texier, T. (2019). Debunking the Stanford Prison Experiment.
4. Revisiting the Stanford prison experiment, again: examining demand characteristics in the guard orientation.
5. Fischer, P., et al. (2011). The bystander-effect: a meta-analytic review.
6. Manning, R., Levine, M., and Collins, A. (2007). The Kitty Genovese murder and the social psychology of helping.