Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Persuasion and Attitude Change

Abstract. Persuasion and attitude change comprise the message-mediated processes through which evaluation, beliefs, intentions, and behavior are altered, resisted, or sustained. This fundamental research monograph traces the historical development of experimental persuasion research, analyzes attitudes as multidimensional evaluative constructs, and compares elaboration-likelihood, heuristic–systematic, metacognitive, narrative, inoculation, and reactance accounts. It critically synthesizes experiments on source credibility, argument processing, affect, repetition, counterarguing, and motivated interpretation; randomized interpersonal and electoral field interventions; meta-analyses of fear appeals and misinformation correction; and recent research on artificial communicators. Methods are assessed for causal identification, construct validity, heterogeneous treatment effects, behavioral transfer, temporal durability, reproducibility, and generalizability. The synthesis shows that persuasive outcomes depend on message properties, prior commitments, audience processing, source attribution, and the operational definition and timing of observed change. It distinguishes persuasive efficacy from factual accuracy and identifies unresolved problems in cross-cultural transportability, mechanism identification, and human–artificial communication. The method is a source-based critical narrative synthesis, not a newly performed experiment, preregistered systematic review, or original meta-analysis.

Keywords: persuasion; attitude change; social psychology; communication; source credibility; elaboration likelihood model; heuristic–systematic processing; metacognition; inoculation theory; psychological reactance; narrative transportation; motivated reasoning; fear appeals; behavioral change; misinformation correction; campaign field experiments; conversational artificial intelligence

The Scientific Object and Research Questions

Persuasion is a communication-dependent process through which exposure to messages, their sources, or their communicative context modifies an evaluation, belief, intention, or behavioral disposition. Attitude change concerns the magnitude, direction, organization, accessibility, or durability of evaluation toward an attitude object. A persuasive message can change what a person believes without changing how the person feels; it may shift an intention while leaving action unchanged, or modify behavior without a durable change of evaluative commitment. These distinctions define the scientific object of this dossier. Its central unit is a specified communicative intervention followed by an outcome assessed relative to a defensible counterfactual, rather than a generic claim that one person influenced another Wood (2000) Eagly & Chaiken (1993).

The research asks how individuals receive a communication, construct its meaning, assess the source and evidence, generate arguments and counterarguments, regulate confidence in their own thoughts, and update or defend attitudes. It further asks which changes survive delayed measurement, competing messages, social interaction, and situations offering opportunities for action. Such questions join social psychology, experimental communication research, consumer psychology, political communication, and media studies. They call for separate definitions of message characteristics, prior attitudes, exposure, mental processing, observed response, and context. The historical and contemporary literature does not support a context-free ranking of persuasive techniques: the same factor can operate differently when attention, motivation, perceived autonomy, or initial commitments change Petty & Cacioppo (1986) Petty & Briñol (2008).

This monograph is a critical narrative synthesis of original experiments, theoretical analyses, published field studies, and research reviews. It does not claim a registered systematic review, newly collected data, or a pooled effect estimate constructed for this work. Scientific conclusions are organized according to causal designs, construct validity, replication, and evidence transfer. The boundaries are equally important: social norms, obedience, interpersonal attachment, and public opinion institutions are neighboring research objects when they possess their own mechanisms. Here they enter only when communication experimentally engages them. The same message-driven process may be studied in classrooms, households, campaigns, advertising, health communication, and human–artificial encounters without multiplying canonical scientific identities Cialdini & Goldstein (2004) DellaVigna (2009).

Attitudes as Evaluative Structures

An attitude is an evaluative orientation toward a specified entity, person, policy, action, practice, or proposition. Attitudes may have cognitive content, affective components, behavioral associations, and varying degrees of certainty, accessibility, importance, and ambivalence. A response to an item such as support for a policy is one indicator of evaluation; the theoretical construct includes patterns of response across contexts and time. An attitude object must be named with sufficient precision to establish whether researchers measure the same thing. Support for an institutional rule, perceived truth of an empirical claim, and willingness to enact a practice may move together yet remain distinct dependent variables Eagly & Chaiken (1993) Fishbein & Ajzen (2010).

Persuasive communication can influence an attitude by changing the perceived desirability of attributes, altering beliefs about their likelihood, evoking an emotional association, shifting which considerations are salient, or changing confidence in thoughts already held. The observable direction of movement is determined partly by the pre-message state. A respondent already at the top of a rating scale cannot report further agreement on that scale, even if confidence intensifies. Conversely, apparent moderation may reflect uncertainty, social desirability, or temporary recall of an alternative rather than stable conversion. Research therefore benefits from multiple outcome measures, baseline assessments, and procedures that avoid ceiling effects and demand characteristics Wood (2000) Briñol & Petty (2009).

The connection between attitudes and behavior depends on correspondence of target, action, context, and time. The reasoned action approach describes behavior through attitudes toward the act, perceived norms, perceived control, and intentions, while acknowledging constraints on action. A persuasion experiment that increases agreement with an educational statement has not thereby demonstrated increased enrollment, compliance, voting, or purchasing. Such transfers require observation or credible behavioral indicators, preferably after delay and under the conditions relevant to the claimed application. Distinguishing belief correction from evaluative attitude change is particularly important when messages present new factual evidence without directly advocating a choice Fishbein & Ajzen (2010) Ecker et al. (2022).

Origins of Experimental Persuasion Research

Modern experimental persuasion research took shape through programs that separated characteristics of communicators, message construction, audiences, and temporal consequences. Within the Yale communication tradition, Hovland and colleagues treated source credibility as a manipulable variable rather than a fixed quality attached to a public figure. Hovland and Weiss's 1951 investigation established an influential experimental problem: how apparent trustworthiness affects immediate acceptance and the later relation between remembered content and remembered source. Their observations became central to discussion of the sleeper effect, although the effect's robustness and qualifying conditions must be evaluated across later independent designs rather than inferred from a single early study Hovland & Weiss (1951) Wood (2000).

Subsequent research broadened the object from communication inputs to recipients' active processing. Early distinctions between source influence and substantive argument processing became more precise when researchers experimentally varied personal relevance, argument characteristics, and communicator expertise. The 1980 experiment by Shelly Chaiken compared high- and low-involvement settings while manipulating message arguments and source likability; the study provided a basis for separating heuristic and systematic processing hypotheses. The 1981 study by Petty, Cacioppo, and Goldman directly crossed relevance, expertise, and argument quality to investigate when participants' reported evaluations followed message content rather than a source cue Chaiken (1980) Petty et al. (1981).

The history of the field also includes research on repeated exposure, resistance, consistency, narratives, group influence, and social identity. Zajonc's 1968 synthesis and experiments placed familiarity-related evaluation on the research agenda, while McGuire and Papageorgis investigated whether prior defenses could reduce vulnerability to subsequent counterarguments. Later theoretical integrations showed that a single source or message variable can perform several different causal functions. Historical progress is best understood as increasingly specified questions about process and conditions rather than as replacement of one universal rule by another Zajonc (1968) McGuire & Papageorgis (1961) Petty & Briñol (2008).

Exposure, Reception, and the Construction of Meaning

Exposure is the necessary first stage of many persuasive sequences, but an exposure count alone does not establish that a message was attended to, understood, believed, or retained. A user may scroll past a headline, hear a fragment while occupied, encounter a repeated slogan, or actively analyze a complete argument. A study should distinguish technical opportunity for exposure from actual perceptual contact and from subsequent mental processing. When message visibility or duration varies experimentally, researchers can identify the effect of that assigned exposure regimen; they still need independent evidence to explain the intervening psychological process DellaVigna (2009) Petty & Cacioppo (1986).

Message comprehension depends on language, prior knowledge, attention, working memory, and situational expectations. A persuasive argument typically includes premises, inferential connections, proposed consequences, and an explicit or implicit evaluative conclusion. The same lexical content can be interpreted as advice, coercion, humor, information, deception, or identity signaling depending on attributed source and audience relationship. Manipulation checks should establish whether participants recognized the relevant claim and cue without turning those checks into additional persuasive treatment. Measures of recall and recognition contribute useful evidence, although remembered messages need not be accepted and accepted messages need not be recalled in full detail Wood (2000) Green & Brock (2000).

Selection mechanisms are especially consequential in real-world exposure. Individuals choose media, avoid threatening claims, seek congenial information, and encounter material ranked by social or technical systems. In observational settings, a relationship between message exposure and later evaluation may arise because prior evaluation influenced exposure in the first place. Randomized delivery identifies an average effect of an assigned communication but may sacrifice the choice and competitive attention of a natural information environment. A complete research program therefore pairs controlled message experiments with realistic exposure records and carefully defined causal comparisons Lord et al. (1979) Kalla & Broockman (2018).

Communicator Credibility, Expertise, and Trust

Communicator credibility comprises perceptions of competence, knowledge, reliability, honesty, and relevant intention. Perceived expertise concerns the likelihood that a source can evaluate the issue accurately; perceived trustworthiness concerns whether the source communicates according to that knowledge and with what incentives. Experimental variations must avoid confusing expertise with prestige, agreement, social status, identity, and familiarity. The same apparently credible source may improve acceptance under low scrutiny, guide selection of arguments under high scrutiny, increase elaboration when diagnostic information is scarce, or alter the confidence with which recipients use their thoughts Hovland & Weiss (1951) Briñol & Petty (2009).

Source labels also affect how evidence is interpreted. A recipient who expects persuasion from a financially interested communicator may discount the claim, while a recipient who treats an expert as accountable may use that expertise as an evidential shortcut. Such judgments can be adaptive when no one can independently verify all messages. Yet a credible source does not make a false claim true, and a distrusted source does not make a valid argument false. Credibility research should therefore include accuracy of claims and quality of evidence as separate dimensions. A factorial design can cross source features with argument strength and observe whether reactions are additive or interactive Petty et al. (1981) Horcajo et al. (2010).

An important advance is the self-validation analysis of source effects. Briñol and Petty reviewed how a source can influence not only the thoughts a person generates, but also confidence in those thoughts. A credible communicator can amplify persuasion when recipients have favorable thoughts about a message, while increasing confidence in negative reactions can undermine persuasion. This contingency explains why apparently strong source cues sometimes generate counterintuitive outcomes. It also illustrates a general lesson for causal interpretation: a source manipulation cannot by itself identify the psychological path through which the resulting attitude shifted Briñol & Petty (2009) Petty & Briñol (2008).

The Elaboration Likelihood Model

The elaboration likelihood model organizes persuasion around variation in the amount and nature of issue-relevant thinking. At relatively high elaboration, recipients examine arguments, compare them with prior information, and generate responses that may support or oppose the advocated position. At relatively low elaboration, simple cues or affective associations can be comparatively influential. The distinction names regions of a processing continuum and the routes by which change is produced; it should not be interpreted as a fixed division of persons into rational and irrational types. Motivation, ability, personal relevance, distraction, and prior knowledge shape where a recipient lies on that continuum in a particular episode Petty & Cacioppo (1986) Petty et al. (1981).

The model's strongest contribution is its treatment of persuasion variables as functionally flexible. An attractive or apparently expert source may operate as a simple cue under low elaboration, increase thought about a message when source information signals the need for careful consideration, influence the positivity of thoughts, or validate the thoughts already generated. Argument strength is likewise consequential only through the processing actually undertaken and the operational meaning of strength. The model thus proposes interactions rather than a universal main effect for every cue. Experimental tests must manipulate the theoretically relevant inputs and measure resulting cognition without treating self-reported effort as a perfect observation of mental process Petty & Cacioppo (1986) Briñol & Petty (2009).

The often-cited prediction that highly elaborated attitudes persist longer or better resist counterpersuasion is a conditional consequence, not an automatic property of thinking more. The content of thought may be weak, biased, or based on false premises; durable conviction can preserve error. Haugtvedt and Petty's 1992 experiments linked differences in need for cognition to persistence and resistance for attitudes formed after persuasion, supporting investigation of processing depth as one contributor to durability. A modern design should also compare delayed change, belief accuracy, behavioral stability, and contextual transfer, because persistence alone is not an adequate measure of successful epistemic persuasion Haugtvedt & Petty (1992) Lord et al. (1979).

Heuristic and Systematic Information Processing

Heuristic–systematic accounts examine how recipients combine relatively economical decision rules with detailed analysis of message content. Heuristics such as presumed expert reliability or social consensus can reduce demands on attention and memory, particularly when the audience has limited motivation or capacity for exhaustive analysis. Systematic processing involves more extensive evaluation of what claims imply and whether the evidence supports them. These modes can coexist: a recipient may use an expertise cue to decide how much scrutiny to allocate, then scrutinize the specific arguments. The theoretically important quantity is the causal contribution of each processing operation in the setting being studied Chaiken (1980) Wood (2000).

Chaiken's 1980 experiments manipulated involvement, message argument quantity, and source likability. The reported pattern supported different weights on message and source factors as involvement changed. The historical result is important because it used an experimental interaction to test a process account rather than simply collecting ratings of a persuasive speaker. At the same time, the original student samples, specific topics, and particular manipulations do not establish universal magnitudes for all messages. Modern tests should vary argument diagnosticity as well as argument count, measure processing at appropriate points, and predefine competing explanations for observed source effects Chaiken (1980) Petty et al. (1981).

An important unresolved question concerns how individuals allocate scarce attention across many simultaneous claims. A heuristic may be an efficient response to reliable social information, while systematic processing may be costly and can still generate motivated reasoning. Neither route guarantees accuracy without information about the evidence and inferential environment. Research must separate the normatively desirable result from the hypothesized cognitive mechanism and ask whether participants improve calibration, revise confidence appropriately, and respond to contrary data. Such distinctions are essential in digital environments where numerous purported experts and source cues compete for immediate attention Ecker et al. (2022) Briñol & Petty (2009).

Thought Generation, Counterarguing, and Metacognition

Persuasive effects depend partly on the thoughts that recipients spontaneously generate while processing a message. Favorable arguments can elicit agreement, but the same arguments may prompt counterexamples in a knowledgeable or defensive audience. Thought-listing procedures seek to capture this internal dialogue by asking participants to report responses after exposure. Their validity depends on timing, reporting demand, coding reliability, and whether the task itself changes processing. A high count of listed thoughts cannot establish deep reasoning, and positive thought valence alone does not determine whether the recipient trusts or uses those thoughts Petty & Cacioppo (1986) Petty & Briñol (2008).

Metacognition adds a distinct mechanism. Self-validation theory proposes that people may evaluate their own cognitive responses for reliability and applicability, giving greater weight to thoughts held with confidence. Horcajo, Petty, and Briñol examined majority and minority source status as conditions affecting confidence in message-related thoughts. Such experiments show why an apparent social cue may alter persuasion after thought generation rather than through simple imitation. The temporal position of a cue matters: information presented before processing can shape what is thought, whereas information presented after processing can affect the weight accorded to those thoughts Horcajo et al. (2010) Briñol & Petty (2009).

The analysis of thought confidence makes persuasive outcomes more diagnostic. A message may improve attitudes through genuinely stronger arguments, reduced attention to counterevidence, or an increase in confidence attached to prior favorable cognitions. These are empirically distinguishable pathways if designs cross thought valence, source information, timing, and independent measures of confidence. Mediation estimates are valuable but depend on assumptions that may be violated when processing and reporting share unobserved causes. Replication should therefore target observable intervention contrasts as well as the assumed psychological mediator. Confidence, correctness, and persuasion are separate quantities throughout this literature Petty & Briñol (2008) Ecker et al. (2022).

Argument Quality, Evidence, and Message Organization

Argument quality names the degree to which a message provides relevant, coherent, credible support for the conclusion it advocates. Experimental persuasion studies frequently operationalize quality through strong and weak argument sets developed in pilot testing. This can be scientifically useful when investigators specify what makes the arguments strong and verify that independent judges assess their diagnostic value. It can also become circular when strength is defined only by producing favorable thoughts in the same participants later used to test the theory. Strong designs separate content validity, plausibility, evidential support, novelty, and rhetorical appeal before estimating their persuasive contributions Petty et al. (1981) Petty & Cacioppo (1986).

Message sidedness and counterargument acknowledgment introduce additional structural choices. A communication may present one conclusion as uncontested, acknowledge opposing considerations without rebuttal, or explain why specific objections are insufficient. The relative effect depends on audience knowledge, prior position, trust, and whether the rebuttal provides genuinely new evidence. An apparently balanced presentation can build credibility with skeptical readers or inadvertently amplify unsupported counterclaims. Claims about superior message formats therefore need comparison groups matched on length, complexity, exposure, and informational content, and ideally should examine what participants remember and how their confidence changes Wood (2000) McGuire & Papageorgis (1961).

Repeated presentations may increase familiarity or perceived plausibility, but repetition also risks fatigue, boredom, and active resistance. Zajonc's mere-exposure tradition concerns changes in evaluation with stimulus recurrence under specified conditions, while message repetition in a complex communication environment involves additional learning, source attribution, and strategic inference. Researchers should not treat familiarity-induced preference as proof that a claim was logically evaluated. A useful factorial design independently varies repetition, source credibility, argument diagnosticity, and exposure interval, then examines immediate judgments and delayed knowledge. Such designs help separate learning, processing fluency, and motivational effects Zajonc (1968) Ecker et al. (2022).

Emotion, Threat, and Risk Communication

Emotion influences the attention assigned to messages, the accessibility of memories, the perceived importance of consequences, and the value attached to possible responses. Persuasion research examines fear, hope, anger, guilt, pride, amusement, and empathy under distinct theoretical mechanisms. An emotion can be part of the message's represented consequences, an affective reaction to presentation, or contextual information a person uses in judging an issue. The causal interpretation requires measuring which role was manipulated and separating affect from factual content. Inducing arousal and then observing a reported attitude does not by itself identify the path from emotion to evaluation Wood (2000) Tannenbaum et al. (2015).

Threat-based messages are an instructive case because their performance depends on severity and susceptibility beliefs as well as perceived response efficacy and capacity to act. The meta-analysis by Tannenbaum and colleagues reviewed 127 articles comprising 248 independent samples and reported an average positive effect of fear appeals on a composite of attitudes, intentions, and behaviors, with random-effects d = 0.29. This estimate summarizes heterogeneous designs and outcomes; it does not justify treating the same effect as expected for any particular population or intervention. An ethically responsible interpretation attends to message accuracy and the practical feasibility of the recommended response Tannenbaum et al. (2015) Witte & Allen (2000).

Threatening language may also generate reactance when a recipient perceives an attempt to constrain freedom. A recent meta-analysis covering studies published between 2005 and 2024 found that higher freedom-threatening language was associated with more anger, negative cognitions, and measured reactance relative to lower-threat wording, while some framing contrasts showed little difference. Its effect estimates apply to the analyzed communication manipulations and should not be collapsed into a claim that emotional appeals always backfire. Persuasive safety and effectiveness are best evaluated together, including whether readers understand the risk, preserve autonomous choice, and retain actionable knowledge Li & Shi (2026) Fransen et al. (2015).

Prior Commitments, Motivated Reasoning, and Polarization

Recipients do not approach persuasive information with identical beliefs or equal incentives to revise them. Prior attitudes determine which claims seem surprising, what background evidence is accessible, and whether an advocated conclusion threatens valued commitments. Lord, Ross, and Lepper's 1979 experiment used materials about capital punishment to study biased assimilation and attitude polarization among participants holding opposing prior views. Its significance lies in the interaction between prior commitment and evaluation of nominally relevant evidence, rather than in a claim that all contradictory evidence invariably polarizes every audience Lord et al. (1979).

Selective interpretation can arise from differential standards of evidential scrutiny, identity-related motivation, perceived source interests, or genuinely different background information. A recipient may accept an apparently favorable study after little inspection while identifying design problems in an unfavorable study; another may revise the confidence attached to the initial position without crossing the response scale's neutral point. Experiments should distinguish these processes by presenting balanced materials, recording preexisting beliefs, varying source attribution, and evaluating perceived quality against external criteria. Simple before-and-after polarization measures can otherwise confound extremity, certainty, scale constraints, and response demand Lord et al. (1979) Ecker et al. (2022).

Persuasion within disagreement consequently depends on the quality of the argument and on the social meaning assigned to reconsideration. Interventions may prompt curiosity, curiosity may expose the recipient to additional evidence, and the additional evidence may either revise or strengthen an attitude. The mechanism must be tested rather than inferred from the eventual direction of movement. The wider persuasion literature supports a conditional model in which processing can generate favorable thoughts, resistance, or selective updating, and the persistence of the outcome depends on more than the quantity of elaboration Petty & Cacioppo (1986) Wood (2000).

Inoculation, Resistance, and Counterpersuasion

Resistance to persuasion is an active scientific object because an unchanged post-message rating can conceal many processes. A recipient may fail to notice an appeal, understand and reject it, retrieve a previously rehearsed rebuttal, distrust its source, or accept the claim while refusing the requested action. Inoculation theory proposes that exposure to weakened attacks and opportunities for refutation can prepare individuals to respond to subsequent persuasive challenges. McGuire and Papageorgis established an early experimental line comparing forms of prior belief defense as means of generating resistance to later communications McGuire & Papageorgis (1961).

Banas and Rains's meta-analysis synthesized experimental inoculation research and found evidence of an overall resistance advantage under the studies and operationalizations included. This supports the general proposition that defensive preparation can alter later message reception, while leaving consequential questions about how long the protection lasts, which attitudes are affected, and whether effects extend to novel misinformation tactics. An intervention that improves detection of a specific argument should not automatically be characterized as broad immunity to propaganda. Transfer requires new test stimuli, delayed assessments, and outcome criteria beyond recognition of familiar attack patterns Banas & Rains (2010) Ecker et al. (2022).

Recipients may also resist by avoidance, counterarguing, asserting autonomy, or selectively discounting the persuasion agent. Fransen, Smit, and Verlegh integrated motives and resistance strategies in a framework that clarifies why a failed persuasive attempt is informative only when the relevant defensive response has been measured. In experimental research, threatening wording should be varied independently of argument validity; otherwise, an increase in resistance cannot be attributed clearly to perceived pressure. The contemporary reactance meta-analyses by Li and Shi give additional evidence that threatening language evokes anger and negative cognitions under studied conditions Fransen et al. (2015) Li & Shi (2026).

Affect, Narrative Transportation, and Identification

Stories can present an advocated evaluation through characters, events, emotional consequences, and causal sequences rather than through explicitly listed reasons. Narrative transportation refers to a state of engagement with the story world in which attention, imagery, and affect become organized around the narrative. Green and Brock's 2000 experiments studied transportation and story-consistent judgments, helping establish a distinct experimental tradition in narrative persuasion. The measurable effects concern particular stories and assessments; a transportation score is an indicator of engagement, not a direct measure of factual accuracy or durable belief Green & Brock (2000).

Narrative influence may involve identification with a character, mental simulation of an outcome, reduced attention to external objections, or enhanced availability of examples. These mechanisms overlap but yield distinguishable predictions. Manipulating narrative perspective while holding informational content constant can help test identification; contrasting coherent and disrupted stories can test narrative organization; delayed questions can assess whether a favorable judgment survives after narrative emotion subsides. Because stories jointly vary many cues, broad claims that narratives outperform statistics should be restricted to evidence from direct comparisons of matched outcomes and populations Green & Brock (2000) Wood (2000).

The social and moral context of narrative persuasion deserves special attention. A compelling personal account may expand the consideration given to an unfamiliar person's experience, but it can also increase acceptance of atypical or inaccurate exemplars. Researchers need a distinction between empathy, attitude change toward a group, willingness to perform a specific behavior, and acceptance of factual assertions. In research on communication across social differences, outcomes should include durable measures of interpersonal evaluation and opportunities to act, not merely immediate emotional resonance. This makes narrative inquiry compatible with the broader framework of cognitive elaboration, affective learning, and source appraisal Broockman & Kalla (2016) Kalla & Broockman (2020).

Persistence, Decay, and Temporal Dynamics

Persuasive change is a trajectory rather than a single post-treatment score. Immediate evaluations can reflect salient message fragments, task instructions, or transient emotional activation. Delayed tests reveal whether the relevant belief or evaluation remains accessible after new experiences and competing messages intervene. Duration should be defined operationally: hours, days, weeks, and months test different persistence hypotheses. A design needs comparable measures at multiple time points and adequate tracking of attrition, since losing participants with particular initial attitudes can mimic attitude stability or decay Haugtvedt & Petty (1992) Kalla & Broockman (2018).

Persistence is influenced by the cognitive and social processes that establish an attitude. Attitudes founded on elaborated argument processing can have distinctive resistance profiles, but well-rehearsed misleading claims can also remain confidently held. Repetition and social endorsement may preserve accessibility independent of strong evidence. In campaign settings, the timing of exposure relative to an election creates a stringent test of durability: effects measured immediately after unusual early contact may diminish before the consequential behavior occurs. Kalla and Broockman's field analysis foregrounded this temporal limitation Petty & Cacioppo (1986) Kalla & Broockman (2018).

Measurement itself may affect the process under observation. Repeated attitude questionnaires invite retrieval, rehearsing reasons, and inference about the investigator's preferred answer. Control groups should therefore be considered for testing effects, and preanalysis plans should specify whether the estimand concerns immediate change, a change sustained until a later date, or the integrated effect across follow-up occasions. If the treatment raises intentions at one week but not behavioral outcomes at three months, that is evidence about an outcome- and time-specific intervention, not a logical contradiction. Strong synthesis retains this outcome structure instead of assigning the message a single lifetime effectiveness label Fishbein & Ajzen (2010) Wood (2000).

From Evaluation to Choice and Behavior

Changing what a person reports about a subject does not entail an equivalent change in action. Behavioral outcomes depend on opportunities, incentives, skills, competing goals, social expectations, and perceived control. Fishbein and Ajzen's reasoned action approach distinguishes attitudes toward performing a behavior from wider evaluations of an object and embeds behavioral intentions among additional determinants. For example, a favorable attitude toward public transportation, a positive evaluation of using it tomorrow, an intention to travel, and an observed transit trip are different measurements. A persuasive message may influence any subset of them Fishbein & Ajzen (2010).

Studies that claim action change should specify the target behavior, observation window, opportunity set, and denominator. A charitable appeal may alter donation choices in a controlled experiment, while a consumer advertisement may affect brand recognition without affecting purchases. The absence of a behavioral effect could reflect low exposure, an unchanged attitude, weak intention, practical obstacles, or insufficient statistical precision. Researchers should separate these explanations using designs with credible objective outcomes and conditions that approximate the settings for which a claim is made DellaVigna (2009) Fishbein & Ajzen (2010).

Behavioral follow-up can uncover displacement and spillover as well as the requested action. A campaign encouraging one choice may redirect behavior from an adjacent alternative while leaving total participation unchanged; a communication correcting an incorrect belief may alter reported knowledge without shifting policy preference. Investigators should specify which outcomes are primary and whether adverse consequences are plausible. Intention-to-treat estimates establish the effect of assigning the message regimen; exposure-adjusted or mediation claims need further assumptions. This evidential discipline keeps the scientific object tied to measurable persuasion rather than a general narrative of social change Kalla & Broockman (2018) Ecker et al. (2022).

Persuasion in Political Campaigns

Campaign communication is a valuable test of persuasive mechanisms because audiences have prior partisan commitments, multiple competing sources, and consequential behavioral choices. The study by Kalla and Broockman synthesized 40 existing field experiments and added nine original experiments, examining campaign contact in United States general elections. Their conclusion was that average persuasive effects on candidate choice were approximately zero in the tested general-election settings, with important qualifications for atypical circumstances and rapidly decaying early effects. These data constrain theories that assume frequent campaign exposure necessarily produces substantial conversion Kalla & Broockman (2018).

A near-zero average effect in this context is compatible with substantial persuasion in another context or for another outcome. Mobilization, candidate choice, policy beliefs, affective polarization, and campaign donation are different estimands; the same message may change the salience of an issue while leaving vote choice constant. An experimental null finding also has an uncertainty interval and a specific treatment contrast. Researchers should record the communication medium, participants' preexisting preferences, timing, and comparison condition before comparing studies. The field evidence helps make the laboratory literature accountable to behavior under competitive information exposure Kalla & Broockman (2018) DellaVigna (2009).

Political communication illustrates how source identity, narrative, and accuracy interact without changing the underlying scientific definition of persuasion. When a message concerns a deeply held value, source affiliation can shape whether the audience processes the claim as information or as a group signal. Randomized message and source experiments can estimate the causal role of these features under controlled conditions, while electoral field studies examine durability under realistic selection and competition. Contemporary artificial communicators add new delivery mechanisms and require dedicated validation of source disclosure, information accuracy, and consent Kalla & Broockman (2020) Hackenburg et al. (2025).

Interpersonal Dialogue and Prejudice Reduction

Interpersonal conversations offer a research setting in which a message can respond to the recipient's statements, prompt narratives, and reveal conversational intentions. Broockman and Kalla's 2016 door-to-door field experiment investigated an approximately ten-minute perspective-taking conversation and measured prejudice toward transgender people over subsequent months. Its empirical importance is the randomized real-world intervention with delayed outcomes, which provides stronger evidence for durability in that setting than an immediate self-report survey alone. It does not establish that any short conversation will produce the same change on every social topic Broockman & Kalla (2016).

Kalla and Broockman's 2020 investigation reported three preregistered field experiments across seven United States locations, involving 230 canvassers and 6,869 voters. Their first experiment contrasted conversation based on arguments with a related condition that included a nonjudgmental exchange of narratives. Argument-only conversations showed no detectable effect on the targeted exclusionary outcomes in that design, whereas narrative exchange produced a small, sustained average shift; additional studies extended tests to attitudes toward transgender people. The size and reliability of a particular intervention must remain linked to its design, population, treatment, and follow-up window Kalla & Broockman (2020).

The strongest interpretation joins informational and relational mechanisms without conflating them. Attentive exchange may change willingness to consider another person's experience, reduce defensive counterarguing, and supply concrete examples that reorganize evaluation. Yet the experiments do not isolate every proposed psychological mediator, and a self-reported attitude is not itself a measure of discrimination in everyday situations. Follow-up work can randomize narrative components, examine conversation fidelity, code conversational turns, and collect appropriately consented behavioral outcomes. Such studies could distinguish active listening, perspective-taking, exposure dosage, and the informational quality of reasons Kalla & Broockman (2020) Green & Brock (2000).

Consumer Communication and Advertising

Commercial persuasion introduces incentives and outcome measures that differ from those used in classical laboratory studies. A message may attempt to change perceived product quality, price fairness, brand trust, purchase intention, or actual buying. These outcomes are related but cannot be combined into a single category of success. The seller's commercial interests also affect perceived source credibility and expectations about selective information. Consumer studies therefore require careful specification of whether the intervention manipulates information about attributes, source identity, repetition, packaging, social proof, or access to the product DellaVigna (2009) Wood (2000).

Controlled message experiments offer high internal validity for the effect of an advertisement on a measured evaluation when exposure and comparison conditions are matched. Field designs may observe sales or website choices but face difficulties separating exposure from self-selection and the effects of concurrent pricing, distribution, and competing campaigns. Brand favorability may improve while actual purchases remain unchanged because of costs or habitual behavior. The reasoned action framework helps identify intermediate outcomes without claiming they all refer to the same attitude object. Credibility and resistance analyses remain useful when commercial sponsorship is disclosed Fishbein & Ajzen (2010) Briñol & Petty (2009).

Advertisements that rely on narratives, emotional risk, repeated visual cues, and implied consensus can recruit several mechanisms simultaneously. A campaign may be memorable because its story is engaging, persuasive because the product claim is diagnostic, or influential because repeated exposure makes the brand familiar. Experimental factorial designs can estimate component effects, but natural campaigns often lack the independence needed for unambiguous causal decomposition. Ethical and scientific appraisal should distinguish truthful claims from persuasive appeal and should measure unintended distortions in perceived product evidence. Consumer persuasion is thus a domain of application of general mechanisms, not a separate copy of the canonical persuasion research object Green & Brock (2000) Zajonc (1968).

Digital Media, Misinformation, and Correction

Digital media make persuasion continuous, scalable, and selectively distributed. A user may encounter a persuasive recommendation embedded in a feed, a search result, a video, a message from a social contact, or an interactive assistant. Platform ranking and self-selection jointly determine which messages are encountered, and these mechanisms complicate causal attribution. A valid experiment must specify whether the intervention changes the content, number, identity, order, timing, or mode of exposure. Otherwise a reported platform effect may combine delivery and psychological responses that require different theories and policy responses DellaVigna (2009) Ecker et al. (2022).

Misinformation correction is a special case of evidence-based attitude and belief change. Ecker and colleagues reviewed psychological drivers of misinformation acceptance and resistance to correction, including familiarity, social signals, prior belief, and the persistence of causal explanations after a claim is withdrawn. Walter and Murphy's meta-analysis synthesized 65 correction effects and found that correction magnitude varied substantially with issue domain and whether the misinformation was real-world or constructed. These studies support the feasibility of correcting false beliefs while underscoring that accuracy outcomes, policy attitudes, and subsequent behavior need distinct measurements Ecker et al. (2022) Walter & Murphy (2018).

Correction quality depends on what information is replaced and whether recipients acquire a usable alternative explanation. Merely stating that a claim is false may leave its original explanatory function intact. An evidence-grounded correction can provide a coherent replacement, identify reliable sources, and avoid unnecessarily reinforcing the original false claim. The empirical task is to compare such components under randomized exposure and assess both immediate belief revision and later reliance on the correction. A message can successfully correct a factual claim and leave a broader preference unchanged because evaluative commitments draw on multiple considerations. This distinction is indispensable when assessing digital public communication Ecker et al. (2022) Walter & Murphy (2018).

Cultural Variation, Audience Differences, and Development

Persuasion effects arise within culturally learned conventions about expertise, deference, interpersonal obligation, disagreement, and acceptable communication. These conventions influence what is perceived as a credible source, a respectful request, a compelling reason, or an unwelcome intrusion. Cross-cultural hypotheses should identify the contextual process and the specific comparison, rather than treating national origin as a psychological mechanism. Researchers should examine whether translated materials carry equivalent meanings and whether rating scales have comparable response characteristics before interpreting group differences as differences in persuasion susceptibility Wood (2000) Eagly & Chaiken (1993).

Audience variation also includes age, knowledge, prior exposure, personal relevance, cognitive resources, and familiarity with the message format. The elaboration literature suggests that motivation and ability change the nature of information processing; this predicts within-person as well as between-person variability. Need for cognition is one individual-difference construct relevant to some experimental outcomes, but it is neither a universal persuasion score nor a proxy for intelligence or epistemic accuracy. Longitudinal evidence must distinguish changes attributable to development from cohort differences, education, social environment, and platform exposure Haugtvedt & Petty (1992) Petty & Cacioppo (1986).

Intervention transportability is an empirical problem. Evidence from college students in a controlled laboratory does not determine effects among other populations or in multilingual, commercially saturated, and politically polarized environments. A persuasive message can also affect trust, autonomy, and perceived social belonging in addition to the targeted evaluation. Comparative studies should preregister moderators where possible, test measurement invariance, recruit more varied samples, and examine whether local adaptation alters the treatment being evaluated. These principles allow an international psychology of persuasion to grow from verifiable findings rather than generic claims about cultural groups Kalla & Broockman (2020) Tannenbaum et al. (2015).

Experimental Methods and Measurement

Scientific measurement of persuasion begins with a stated estimand: the contrast between a specified communication intervention and a defined alternative for a named outcome at a named time. Researchers may randomize participants to messages differing in argument quality, source credibility, format, repetition, or perceived pressure. A successful experiment specifies the sampling frame, assignment method, intervention content, comparison condition, manipulation checks, outcome scales, follow-up interval, and exclusions. If participants interact with one another, clustering and interference require explicit analysis. When messages themselves are sampled, item variability matters alongside participant variability, because effects may be driven by a narrow set of unusually effective texts Petty et al. (1981) Chaiken (1980).

Attitudes are often assessed using semantic differential measures, agreement ratings, feeling thermometers, implicit tasks, stated intentions, or repeated judgments in response to different contexts. Each operationalization observes a particular property and may introduce response bias. Before interpreting score change as persuasion, investigators should consider internal consistency, test–retest reliability, scale direction, floor and ceiling effects, item wording, and whether respondents understand the target in the same way. For politically or morally sensitive topics, anonymous measurement, indirect questioning, and behavioral tests can sometimes clarify whether the intervention changed evaluation or simply changed willingness to disclose an opinion Eagly & Chaiken (1993) Fishbein & Ajzen (2010).

Mechanism-focused research can combine randomized factorial manipulations with process measures: thought listings, attention traces, recall, source attribution, perceived credibility, emotional response, and confidence judgments. Measures of the proposed mediator should be collected at times justified by theory and should avoid interrupting the message in a way that changes the treatment. Statistical mediation is informative when its assumptions are justified, but experimental manipulation of an intermediate process supplies stronger causal leverage where feasible. Factorial interactions should be interpreted with attention to the metric, power, and preregistered contrasts. The ultimate aim is to discriminate explanations through predictions that would differ under alternative causal models Briñol & Petty (2009) Petty & Briñol (2008).

Evidence Appraisal, Reproducibility, and Causal Inference

Laboratory studies and field experiments answer complementary questions. Laboratory randomization can isolate a source cue or argument feature under controlled exposure; field randomization can estimate the effect of a delivery strategy amid natural competing demands. Neither method alone establishes the psychological mechanism across all settings. Evidence appraisal should compare manipulation fidelity, attrition, baseline imbalance, statistical precision, multiple comparisons, outcome preregistration, and the credibility of measurement. A result supported by a large sample may be statistically precise yet trivial for the proposed practical consequence. Conversely, an underpowered null result does not prove the absence of all persuasion Kalla & Broockman (2018) DellaVigna (2009).

Meta-analyses can estimate the average contrast represented in eligible studies while revealing variation in designs and populations. Tannenbaum and colleagues' fear-appeal synthesis summarized 248 independent samples and reported a positive mean standardized effect on combined outcomes, alongside substantial between-study heterogeneity. Such estimates support conditional conclusions about message designs within a literature, not universal effect sizes for untested contexts. The later reactance syntheses by Li and Shi reviewed 33 studies from 28 articles, showing how specific features of freedom-threatening language relate to anger and negative cognitions. The units of evidence and the definitions of outcomes must be preserved whenever summaries are compared Tannenbaum et al. (2015) Li & Shi (2026).

Reproducibility requires accessible materials, documented analysis choices, appropriate blinding, and independent replication that varies laboratories, message sets, and population characteristics. A historically influential demonstration should be evaluated as one observation within a research program, particularly where results depend on few stimuli or immediate questionnaires. The strongest claims combine replicated causal contrasts, evidence about intervening cognitive processes, and outcome-specific follow-up. Conflicts between studies can arise from distinct target attitudes or competing counterfactuals rather than irreconcilable theories. A critical synthesis should report both favorable findings and conditions of failure, preserving uncertainty where designs do not distinguish the explanations under discussion Wood (2000) Kalla & Broockman (2020).

Artificial Communicators and the Artificial Era

Communications produced by artificial systems introduce new experimental variables while retaining the core persuasion questions: what message was presented, by what attributed source, to whom, and with what consequences? An artificial communicator may generate a fixed statement, choose among candidate messages, or adapt interactively to the reader's responses. Source disclosure, personalization, response timing, factual accuracy, and conversational memory are distinct factors that should be tested independently. The Artificial Era therefore expands the range of communicative systems that can be investigated without assuming that persuasive success demonstrates subjective experience or a human-equivalent psychological organization Bai et al. (2025) Hackenburg et al. (2025).

Bai and colleagues reported three preregistered experiments involving 4,829 participants. In the tested policy-attitude tasks, exposure to large-language-model-generated messages shifted reported support relative to neutral-message controls; the generated messages were approximately as effective as messages written by lay humans. Hackenburg and colleagues' 2025 research combined three large-scale experiments involving 76,977 participants with an assessment of factual accuracy across model-generated claims. They reported that post-training and prompting choices affected persuasion more substantially than several other tested design dimensions, while approaches that improved measured persuasive impact could reduce factual accuracy. These are particularly important warnings against equating effectiveness with evidential soundness Bai et al. (2025) Hackenburg et al. (2025).

The scientific next step is to treat artificial persuasion as an identifiable causal system. Experimental protocols should preserve conversation transcripts where consent and privacy rules permit, record model configuration, document retrieval sources and message revisions, randomize disclosure and adaptive response features, and measure delayed belief and behavior. Accuracy must be independently assessed, and human responses to attributed artificial agency must be distinguished from system capabilities demonstrated in controlled tests. Human–artificial communication can be investigated across education, consumer guidance, political messaging, and health information without declaring a universal beneficial or harmful effect. The resulting research links persuasion psychology to computational social science, human–computer interaction, and the psychology of human–artificial relations Hackenburg et al. (2025) Ecker et al. (2022).

Integrative Scientific Synthesis and Open Problems

The evidence supports a multistage account: a message must reach an audience, be interpreted in context, enter attention and evaluation, generate or alter thoughts, and interact with source judgments, emotions, perceived freedom, and prior commitments. Changes can then be expressed as belief revision, reported attitude, confidence, intention, or observable conduct. These stages may overlap and feed back on one another. No single model specifies every causal path for every setting; the elaboration likelihood and heuristic–systematic traditions, self-validation research, narrative theory, and studies of reactance offer partly complementary descriptions of selected mechanisms. Their empirical value lies in discriminating predictions and operational specificity Petty & Cacioppo (1986) Chaiken (1980) Briñol & Petty (2009).

Four recurrent evidential problems organize future work. First, investigators must explain when short-term reported evaluation predicts durable behavior. Second, they must characterize heterogeneous responses without converting differences in participants, messages, and baseline attitudes into unsupported personality types. Third, they must separate informational accuracy from the rhetorical impact of communication and the credibility of its messenger. Fourth, they must establish external validity across media, languages, cultures, and artificial conversational systems. Progress requires coordinated behavioral follow-up, multiple message samples, transparent analyses, calibrated measures, and sufficiently varied populations Kalla & Broockman (2018) Walter & Murphy (2018) Hackenburg et al. (2025).

Ethical evaluation can be empirically informed by tracking factual validity, manipulation transparency, autonomy, foreseeable harms, and disparities in outcomes, while recognizing that no single attitude score captures all these properties. A study designed only to maximize expressed agreement may miss coerced compliance, uncertainty suppression, durable misinformation, or behavioral displacement. Future experiments should make the factual status of contested claims explicit, measure informed decision quality where possible, and preregister adverse outcomes alongside target effects. The most powerful scientific synthesis concerns the conditions under which particular communications cause particular changes, how those changes are sustained, and what explanatory models survive independent tests Li & Shi (2026) Ecker et al. (2022) Fishbein & Ajzen (2010).

Conclusion

Persuasion and attitude change form a central research domain of social psychology concerned with message-mediated modifications of evaluation, belief, confidence, intention, and action. Its history runs from experimental manipulation of communicator credibility through theoretical accounts of elaboration and heuristic processing to field interventions, correction studies, and interactive artificial communication. Original experiments establish the importance of source appraisal, recipient engagement, argument processing, and resistance, while modern evidence shows that effects depend on the intervention, the audience, the outcome, and the time of observation. The domain is best organized around explicitly defined causal comparisons and valid psychological measurements Hovland & Weiss (1951) Petty et al. (1981) Wood (2000).

Two findings anchor the synthesis. First, a communication may alter what an audience reports immediately without producing a lasting behavioral difference; the political field-experimental literature makes this limitation especially visible. Second, some specific conversation and correction interventions demonstrate measurable effects under controlled or field conditions, with differing durability and evidentiary quality. Meta-analyses also show measurable responses to fear appeals and freedom-threatening language while documenting the heterogeneity of those treatments. Generalization should proceed by testing mechanisms and moderators, not by transferring an isolated effect size between contexts Kalla & Broockman (2018) Kalla & Broockman (2020) Tannenbaum et al. (2015) Li & Shi (2026).

Within the Artificial Era, human evaluation is increasingly shaped by messages selected, composed, or exchanged through artificial systems. Emerging experiments make artificial communicators an empirically tractable part of persuasion psychology and raise urgent questions about disclosure, accuracy, trust, autonomy, and longitudinal effects. The established scientific task remains to determine what changed, why it changed, for whom, and for how long. A rigorous account of persuasion integrates theories of cognition and social influence with experimental identification, behavioral observation, transparent evidence appraisal, and the explicit separation of persuasive efficacy from truth Bai et al. (2025) Hackenburg et al. (2025) Ecker et al. (2022).

References

  • Bai, H., Voelkel, J. G., Muldowney, S., Eichstaedt, J. C., & Willer, R. (2025). LLM-generated messages can persuade humans on policy issues. Nature Communications, 16, Article 6037. https://doi.org/10.1038/s41467-025-61345-5
  • Banas, J. A., & Rains, S. A. (2010). A meta-analysis of research on inoculation theory. Communication Monographs, 77(3), 281–311. https://doi.org/10.1080/03637751003758193
  • Briñol, P., & Petty, R. E. (2009). Source factors in persuasion: A self-validation approach. European Review of Social Psychology, 20(1), 49–96. https://doi.org/10.1080/10463280802643640
  • Broockman, D., & Kalla, J. (2016). Durably reducing transphobia: A field experiment on door-to-door canvassing. Science, 352(6282), 220–224. https://doi.org/10.1126/science.aad9713
  • Chaiken, S. (1980). Heuristic versus systematic information processing and the use of source versus message cues in persuasion. Journal of Personality and Social Psychology, 39(5), 752–766. https://doi.org/10.1037/0022-3514.39.5.752
  • Cialdini, R. B., & Goldstein, N. J. (2004). Social influence: Compliance and conformity. Annual Review of Psychology, 55, 591–621. https://doi.org/10.1146/annurev.psych.55.090902.142015
  • DellaVigna, S. (2009). Psychology and economics: Evidence from the field. Journal of Economic Literature, 47(2), 315–372. https://doi.org/10.1257/jel.47.2.315
  • Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Harcourt Brace Jovanovich College Publishers. https://discover.library.unt.edu/catalog/b1755332
  • Ecker, U. K. H., Lewandowsky, S., Cook, J., Schmid, P., Fazio, L. K., Brashier, N., Kendeou, P., Vraga, E. K., & Amazeen, M. A. (2022). The psychological drivers of misinformation belief and its resistance to correction. Nature Reviews Psychology, 1, 13–29. https://doi.org/10.1038/s44159-021-00006-y
  • Fishbein, M., & Ajzen, I. (2010). Predicting and changing behavior: The reasoned action approach. Psychology Press. https://www.routledge.com/Predicting-and-Changing-Behavior-The-Reasoned-Action-Approach/Fishbein-Ajzen/p/book/9780805859249
  • Fransen, M. L., Smit, E. G., & Verlegh, P. W. J. (2015). Strategies and motives for resistance to persuasion: An integrative framework. Frontiers in Psychology, 6, Article 1201. https://doi.org/10.3389/fpsyg.2015.01201
  • Green, M. C., & Brock, T. C. (2000). The role of transportation in the persuasiveness of public narratives. Journal of Personality and Social Psychology, 79(5), 701–721. https://doi.org/10.1037/0022-3514.79.5.701
  • Hackenburg, K., Tappin, B. M., Hewitt, L., Saunders, E., Black, S., Lin, H., Fist, C., Margetts, H., Rand, D. G., & Summerfield, C. (2025). The levers of political persuasion with conversational artificial intelligence. Science, 390(6777), Article eaea3884. https://doi.org/10.1126/science.aea3884
  • Haugtvedt, C. P., & Petty, R. E. (1992). Personality and persuasion: Need for cognition moderates the persistence and resistance of attitude changes. Journal of Personality and Social Psychology, 63(2), 308–319. https://doi.org/10.1037/0022-3514.63.2.308
  • Horcajo, J., Petty, R. E., & Briñol, P. (2010). The effects of majority versus minority source status on persuasion: A self-validation analysis. Journal of Personality and Social Psychology, 99(3), 498–512. https://doi.org/10.1037/a0018626
  • Hovland, C. I., & Weiss, W. (1951). The influence of source credibility on communication effectiveness. Public Opinion Quarterly, 15(4), 635–650. https://doi.org/10.1086/266350
  • Kalla, J. L., & Broockman, D. E. (2018). The minimal persuasive effects of campaign contact in general elections: Evidence from 49 field experiments. American Political Science Review, 112(1), 148–166. https://doi.org/10.1017/S0003055417000363
  • Kalla, J. L., & Broockman, D. E. (2020). Reducing exclusionary attitudes through interpersonal conversation: Evidence from three field experiments. American Political Science Review, 114(2), 410–425. https://doi.org/10.1017/S0003055419000923
  • Li, Z., & Shi, J. (2026). Message effects on psychological reactance: Meta-analyses. Human Communication Research, 52(1), 38–52. Published online June 30, 2025. https://doi.org/10.1093/hcr/hqaf016
  • Lord, C. G., Ross, L., & Lepper, M. R. (1979). Biased assimilation and attitude polarization: The effects of prior theories on subsequently considered evidence. Journal of Personality and Social Psychology, 37(11), 2098–2109. https://doi.org/10.1037/0022-3514.37.11.2098
  • McGuire, W. J., & Papageorgis, D. (1961). The relative efficacy of various types of prior belief-defense in producing immunity against persuasion. Journal of Abnormal and Social Psychology, 62(2), 327–337. https://doi.org/10.1037/h0042026
  • Petty, R. E., & Briñol, P. (2008). Persuasion: From single to multiple to metacognitive processes. Perspectives on Psychological Science, 3(2), 137–147. https://doi.org/10.1111/j.1745-6916.2008.00071.x
  • Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 19, pp. 123–205). Academic Press. https://doi.org/10.1016/S0065-2601(08)60214-2
  • Petty, R. E., Cacioppo, J. T., & Goldman, R. (1981). Personal involvement as a determinant of argument-based persuasion. Journal of Personality and Social Psychology, 41(5), 847–855. https://doi.org/10.1037/0022-3514.41.5.847
  • Tannenbaum, M. B., Hepler, J., Zimmerman, R. S., Saul, L., Jacobs, S., Wilson, K., & Albarracín, D. (2015). Appealing to fear: A meta-analysis of fear appeal effectiveness and theories. Psychological Bulletin, 141(6), 1178–1204. https://doi.org/10.1037/a0039729
  • Walter, N., & Murphy, S. T. (2018). How to unring the bell: A meta-analytic approach to correction of misinformation. Communication Monographs, 85(3), 423–441. https://doi.org/10.1080/03637751.2018.1467564
  • Witte, K., & Allen, M. (2000). A meta-analysis of fear appeals: Implications for effective public health campaigns. Health Education & Behavior, 27(5), 591–615. https://doi.org/10.1177/109019810002700506
  • Wood, W. (2000). Attitude change: Persuasion and social influence. Annual Review of Psychology, 51, 539–570. https://doi.org/10.1146/annurev.psych.51.1.539
  • Zajonc, R. B. (1968). Attitudinal effects of mere exposure. Journal of Personality and Social Psychology, 9(2, Pt. 2), 1–27. https://doi.org/10.1037/h0025848