Effective strategies for changing public opinion: A literature review

by Jamie_Harris54 min read9th Nov 20212 comments


Sentience InstitutePolicyMoral advocacyMoral circle expansionBallot initiativeFarmed animal welfare

Many thanks to Janet Pauketat, Courtney Dillard, Jacy Reese Anthis, and Ilana Rudaizky for reviewing and providing feedback.


Social movements often try to affect public opinion as a lever for legislation and other social change. This report provides a broad overview of research relating to (1) whether advocates can successfully influence public opinion and how they can do so most effectively, and (2) what the other causes of public opinion change are and how advocates can harness them. Key findings include that direct persuasion attempts to change attitudes (rather than behavior) tend to have “small” or “very small” short-term effects, but advocates may be able to have lasting indirect effects on public opinion via policy change or reframing the issues. We list factors that affect how persuasive messages are and make tentative suggestions for how advocates can cost-effectively leverage external influences on public opinion such as the media, celebrities, politicians, and policy.


Public opinion — “the preferences of the adult population on matters of relevance to government”[1] — is important for social change: While historical causation is always difficult to assess, our previous research projects indicate a significant effect of public opinion on legislative outcomes.[2] Political scientist Alan Monroe (1998) found that in 70% of policy decisions between 1980 and 1993 in which the public favored the status quo, the US government maintained it.[3] Sentience Institute’s “Summary of Evidence for Foundational Questions in Effective Animal Advocacy” lists a number of other instances where public opinion could be important (citations omitted):

  • “Public opinion could play an important role in affecting whether legislation is preserved or subsequently overturned…
  • The decision-making of the Supreme Court of the United States seems to be substantially influenced by public opinion.
  • When pre-decision public opinion is more closely aligned with a Supreme Court decision, the risk of legislative backlash is lower and the effects of the ruling on public opinion seem likely to be more positive…
  • Corporate welfare campaigns have been partly dependent upon mobilizing the public to express dissatisfaction with a particular practice in animal farming, such as the caging of layer hens. It may be crucial that the public is already opposed to a practice for such campaigns to be successful.”[4]

Hence, although much animal advocacy research has treated behavior change, especially individual diet change, as the main outcome of interest,[5] understanding the causes of public opinion change also seems useful for cost-effectively encouraging social change.

The research reviewed here is relevant for evaluating the usefulness of moral circle expansion (i.e., advocacy to increase the number of beings given moral consideration, such as legal protection) and other forms of values spreading as strategies for influencing the long-term future. The impact of these strategies rests on fundamental questions, such as[6]:

  • Does advocating for a value system increase or reduce the number of people with these values?
  • Do persuaded individuals take action in support of their new values, such as further advocacy?
  • Do the actions of persuaded individuals tend to align with the intentions of the original advocates?
  • Will the values that are spread be durable or will they revert to an equilibrium?


Below are a number of strategic claims supported by the evidence in this review:

  • Direct advocacy and persuasive messages focused on attitudes seem likely to have effect sizes conventionally interpreted as “small” or “very small.” Though not the norm, persuasive messages can sometimes cause attitudes to change in the opposite direction to intended (“boomerang” attitude change), e.g. when the difference between the message’s position and the receiver’s position is very large.
  • There are a number of communicator, message, and receiver factors that have small or moderate effects on how persuasive messages are. For example:
  • More communicator credibility (expertise, trustworthiness, and caring), likeability, attractiveness, communicator-receiver similarity, and authority usually enhance persuasiveness. These factors tend to be more influential when the receiver is not thinking carefully about the issue, e.g. if it is of low relevance to them.
  • Advocates can increase their effectiveness if they make emotional appeals to fear, guilt, anger, or disgust. However, these tactics can reduce message effectiveness or lead to boomerang effects if they are used incorrectly or excessively.
  • Describing but refuting opposing arguments can increase the effectiveness of persuasive messages and make audiences more resistant to opposing persuasion attempts.
  • Persuasive messages tend to be more effective if the audience is less familiar with or opinionated about an issue. This suggests that advocates can influence public opinion towards less salient attributes or sub-topics (e.g. a specific farming practice) if they are able to shape the initial media coverage or share persuasive messages widely. It also suggests that tactics that are aimed at increasing salience may be counterproductive (by making subsequent attitude change more difficult) on issues where public opinion is currently unfavorable.
  • Framing variations can influence attitudes through different mechanisms to direct persuasion attempts by encouraging audiences to place more weight on some considerations than others.
  • Protests and social movement events can influence public opinion as well as the public’s perceptions of the importance of certain issues.
  • The media can cause public opinion change through persuasive messaging or one-sided coverage. Additionally, if it highlights certain attributes of an issue over others, the media can influence overall public opinion without persuasion, similarly to how advocates can influence attitudes through framing manipulations. If advocates can focus media attention on more favorable attributes of an issue, they may be able to indirectly influence public opinion.
  • High media coverage tends to make the public think that an issue is more important.
  • Politicians and celebrities can act as advocates, influencing the attitudes of their audience, though they are not always more persuasive than alternative spokespeople. They can also influence media coverage and public perceptions of the importance of certain topics or attributes. This suggests that if advocates can encourage supportive public comments from these individuals, this could be worthwhile, but efforts would be counterproductive if they encouraged hostile comments.
  • If advocates can secure policy change, public opinion will tend to move towards support for those policies.
  • Intergroup contact and several other prejudice reduction strategies can encourage moral consideration of outgroups, though it may be difficult for advocates to cost-effectively utilize these strategies on a large scale.
  • The effects on attitudes from exposure to single messages — whether from persuasive messaging, framing manipulations, or media coverage — are most pronounced immediately after exposure to the message. The effects diminish with time, though at least some effects have been found weeks afterwards. For advocates to cost-effectively cause lasting public opinion change, they therefore probably need to encourage some sort of self-perpetuating mechanism such as new legislation, social norms, or framings used by the media.
  • A number of indirect or long-term factors can influence public opinion, such as demographic changes and major events beyond the control of advocates.


Search terms were inputted into Google Scholar, seeking to identify meta-analyses, textbooks, and summary articles relevant to the causes of public opinion change, especially those that might be relevant for advocacy strategy. The citations of and by some of the most important and relevant items were also searched. There were no strict inclusion or exclusion criteria. For example, although the focus was usually on meta-analyses and textbooks, if these were not available for topics of interest, individual studies were sometimes sought. Items were included if they seemed useful and relevant in some way to the topic, even if they did not meet a certain definition of the term “public opinion” (of which there are several[7]). For example, this review does not usually distinguish between attitudes and public opinion, so includes research on both.[8] Research on the causes of changes in less relevant outcomes (e.g. behavior or knowledge) was usually excluded, but sometimes discussed briefly for comparison to research on public opinion.

This topic is too broad to be suited to a single systematic review and the outcomes used are too diverse for an overview of reviews to be appropriate either.[9] This report does not attempt to comprehensively review the literature on any specific topic, but rather to identify and summarize a variety of research findings of interest to advocates and researchers of social change. The reviewed research comes from a number of disciplines, including communication studies, political science, psychology, and sociology.

When reporting on the effect sizes from included meta-analyses, the metrics provided by the original reviewers are used, most commonly standardized mean difference (SMD) / Cohen’s d and Pearson’s r correlation coefficients. Guidelines exist for interpreting some of these different forms of outcome measures, most notably Jacob Cohen’s definitions of r = 0.1 or d = 0.2 as a “small effect size,” r = 0.3 or d = 0.5 as a “medium effect size,” and r = 0.5 or d = 0.8 as a “large effect size.”[10]



This section focuses on tactics that advocates can use to influence public opinion fairly directly, e.g. via direct advocacy and persuasion; opportunities for more indirect influence are reviewed in the following section.

Direct advocacy and persuasion

Although findings are mixed, a number of studies suggest that social movements can influence public opinion in the direction that they intend through public campaigns and persuasive messaging.[11] For example, several studies on anti-abortion and anti-death penalty attitudes suggest that educational interventions can have positive effects, but the effects may be small and short-term.[12]

Rains et al. (2018) review and summarize “149 meta-analyses exploring human communication phenomena” (not all focusing on attitude change) which have a mean effect size of r = .21, a small effect size.[13] Marketing scholar Jacob Hornik and colleagues (2016) combined 2,276 different effect sizes from advertising studies in meta-analysis and found a mean weighted effect size on persuasion outcomes (including “attitude toward the product, attitude toward the brand, purchase intention, and product choice”) of r = .19.[14] Although neither of these reviews focused exclusively on attitude outcomes, meta-analyses with a narrower focus on attitude change tend to find small or very small effects.[15] For example:

  • A meta-analysis of 36 experiments found that “communicating evidence of a policy’s effectiveness increased support for the policy (SMD = 0.11, 95% CI [0.07, 0.15], p < 0.0001)” and “[c]ommunicating evidence of ineffectiveness decreased policy support (SMD = −0.14, 95% CI [−0.22, −0.06], p < 0.001).”[16]
  • A meta-analysis found significant effects on attitudes from 10 experiments and quasi-experiments of descriptive social norms manipulations, which focus on “the perceived prevalence of a behavior” (d = 0.17, 95% CI [0.06, 0.27]) and 4 studies of injunctive norms manipulations, which focus on “the social (dis)approval of one’s actions” (d = 0.34, 95% CI [0.17, 0.50]).[17]
  • A meta-analysis of 30 studies found that “[f]act-checking has a significantly positive overall influence on political beliefs (d = 0.29).”[18]
  • A meta-analysis of 49 field experiments of political campaign advertising and outreach actually found an average effect of zero on US voting choices.[19] 

Many of these effects come from studies using only short messages in artificial contexts, so they may not be very informative about the effects we should expect from real-world advocacy contexts. More extensive interactions (e.g. a lengthy conversation, a documentary, a book) or repeated exposure to similar arguments (e.g. via shifts in media coverage, discussed below) may have larger effects.[20]

However, there is evidence from a number of meta-analyses on health behavior (e.g. persuading people to eat more healthily, exercise more, or smoke less) that education or information only interventions, social norms interventions, mass media campaigns, social marketing, and advertising also tend to have small or very small effect sizes. These studies tend to focus on more realistic interventions, such as lessons conducted over a number of weeks in schools or large-scale mass media campaigns visible to the general public.[21] Attitude and behavior change are correlated,[22] so this provides a rough guide for expected effect sizes of changes in attitudes from targeted individuals in comparable interventions focused on public opinion. Some reviews of mass media campaigns find changes in attitudes that are larger than the changes in behaviors, though they are still sometimes very small.[23] Even if advocates are able to provide repeated persuasive messages to their audience, encouraging attitude change is difficult.


If advocates design and deliver messages in the most effective manner possible, the effects on attitudes may still be small.[24] Nevertheless, there are a number of factors that make small or moderate differences to the effectiveness of persuasive messages.

Theories of persuasion moderators

In chapters 2 to 8 of Persuasion: Theory and Research (2015), communication scholar Daniel J. O’Keefe outlines various theories that help to explain outcomes from persuasive messages and the factors that moderate how effective they are.[25] For example:

  • Chapter 2 describes “social judgement theory,” where, “the effect of a persuasive communication depends upon the way in which the receiver evaluates the position it advocates.”[26] A key prediction of this theory, supported by numerous studies, is that the relationship between message discrepancy (“the difference between the message’s position and the receiver’s position”) and attitude change “is suggested to be something like an inverted-U-shaped curve.”[27] In other words, advocates need to carefully evaluate how radical to make their messages in order to achieve maximum effect — too conservative and the attitude change produced will be in the intended direction but small, too radical and the attitude change produced could be small, non-existent, or even in the opposite direction to intended (“boomerang” attitude change).[28] The ideal message will vary depending on the audience and their current view. 
  • Social judgement theory also predicts that when someone has less extreme views on a topic, e.g. because it has little personal relevance to them, a persuader might be able to successfully advocate for a more discrepant position.[29] This suggests that persuasion attempts relating to less salient topics will be more tractable, as will those relating to public policy and other institutional changes rather than, say, a recipient’s dietary choices. However, a number of strategies may facilitate effective persuasion for audiences with strong attitudes.[30]
  • Chapter 3 describes “functional approaches to attitude,” where, “[t]he basic idea is that attitudes may serve various functions.”[31] This is relevant because “substantial evidence suggests that persuasive appeals that are matched to the receiver’s attitude function will be more persuasive than mismatched appeals.”[32] This implies that it is useful to understand why your audience holds certain attitudes and tailor messages towards their attitude function where possible.
  • Chapter 5 summarizes and evaluates cognitive dissonance theory which entails that attitudes can change because “persons seek to maximize the internal psychological consistency of their cognitions (beliefs, attitudes, etc.).”[33] The theory has many implications for persuasion. One key point is that people may resolve cognitive dissonance — discomfort from conflicting thoughts — in a number of ways, e.g. by changing their attitudes to justify their new behavior (if you manage to change behavior without changing attitudes first), or changing their behavior to bring it in line with their new attitudes (if you manage to persuade them of something).[34] This provides an explanation for why attitude change relating to personal behaviors is so hard — people will want their attitudes to match their behaviors, so may reject arguments that you present to them.
  • Chapter 8 analyzes the Elaboration Likelihood Model (ELM), which “suggests that important variations in the nature of persuasion are a function of the likelihood that receivers will engage in elaboration of (that is, thinking about) information relevant to the persuasive issue.”[35] There is evidence that, “attitudes shaped under conditions of high elaboration will (compared with attitudes shaped under conditions of low elaboration) display greater temporal persistence, be more predictive of intentions and subsequent behavior, and be more resistant to counterpersuasion.”[36] All else being equal, it is clearly preferable to encourage attitude change under conditions of high elaboration, though this may be more difficult. There is also evidence supporting ELM’s implication that, under conditions of low elaboration, such as when a topic has little personal relevance to the receiver, heuristics such as the communicator’s perceived credibility (discussed below) become especially important as determinants of attitude change, relative to argument quality.[37]

Communicator factors

There are a number of “communicator factors” that make persuasive efforts more or less effective. These have been identified in numerous studies and summarized in chapter 10 in O’Keefe’s Persuasion (2015) and chapter 8 in The Dynamics of Persuasion (2017), another textbook on persuasion by communications scholar Richard Perloff[38]:

  • Although high credibility communicators tend to be more persuasive, surprisingly, “at least sometimes low-credibility communicators are significantly more effective than high-credibility communicators.”[39] One notable finding is that, “[w]ith a counterattitudinal message [i.e. one which opposes the receiver’s current view], the high-credibility communicator will tend to have a persuasive advantage over the low-credibility source; with a proattitudinal message, however, the low-credibility communicator appears to enjoy greater persuasive success than the high-credibility source.”[40] The perceived “credibility” of the communicator is influenced by their perceived “expertise,” e.g. education, occupation, and experience; citation of evidence sources (rather than providing vague claims); and “nonfluencies in delivery” (like saying “uh” a lot).[41] Credibility is also influenced by perceived trustworthiness[42] and perceived caring (goodwill), though the latter is less consistently identified as important.[43] Credibility is especially important under conditions of low elaboration and if the communicator’s identity is clear before a message is delivered.[44]
  • As with credibility, communicators who are more liked by their audience are usually more persuasive, especially under conditions that prompt low elaboration. Again, however, sometimes disliked communicators are more effective, typically when the audience has “freely chosen to listen to the message.”[45] There is also evidence that, “[t]he effects of liking can apparently be overridden by credibility.”[46]
  • Physical attractiveness and communicator-receiver similarity can both affect persuasiveness, but operate indirectly such as through effects on liking and credibility.[47]
  • Authority can induce compliance (i.e. outward conformity), but not necessarily private acceptance.[48] It is unclear whether compliance will be sufficient for advocacy purposes.

There are many complexities affecting when being evaluated positively on each of these factors is more or less useful. As a general rule, however, more credibility (expertise, trustworthiness, and caring), likeability, attractiveness, communicator-receiver similarity, and authority usually enhance persuasiveness.[49]

Message factors

Scholars have also identified a number of “message factors” that affect persuasiveness:

  • Narratives can sometimes be more persuasive than nonnarrative messages, especially if the audience identifies with the characters or is “transported by” (“caught up in, or carried away by”) the story.[50] 
  • The use of evidence also enhances persuasion, either by improving the strength of the arguments or by acting as a cue of the communicator’s credibility.[51] Meta-analyses examining whether statistical evidence or narratives are more effective have come to conflicting conclusions.[52] Statistical evidence may be most effective in specific contexts, e.g. under conditions of high elaboration.[53]
  • More broadly, one meta-analysis has found that “vividness” of persuasive appeals has small positive effects on attitudes[54] and another has found that emotional appeals tend to be more persuasive than rational appeals in advertising.[55] 
  • A meta-analysis of 38 studies found that, overall, metaphors have a very small persuasive advantage over literal messages (r = .07), which rose to a moderate effect size (r = .42) “under optimal conditions, when a single, nonextended metaphor was novel, had a familiar target, and was used early in a message.”[56]
  • Intense language may be effective at enhancing enthusiasm among supporters, but ineffective at persuading those with contrasting views.[57]
  • There are usually stronger effects for messages that explicitly state their recommendation, rather than omitting it.[58] Relatedly, “messages with more specific descriptions of the recommended action are more persuasive than those providing general, nonspecific recommendations.”[59]
  • Fear appeals can be effective — they tend to have small positive effects on attitudes, intentions, and behaviors.[60] However, to have the intended effects, they need to first successfully “convince message recipients that they are susceptible to negative outcomes,” that “the recommended response will alleviate the threat,” and that the recipients are capable of achieving the recommended response.[61] They also must not induce so much fear that recipient becomes incapacitated.[62]
  • Guilt appeals may be even more effective — one meta-analysis found “a strong positive overall effect of guilt (r = .49, 95% CI 0.31–0.64)” on “health-related attitudes and intentions.”[63] Like fear appeals, they may be effective “only if certain conditions are met. Research suggests that guilt appeals can work if: (a) the message induces empathy, (b) instills a sense of social, normative responsibility to help, and (c) convinces individuals that the recommended behavior will reduce guilt or repair the problem. However, if the message goes too far, eliciting reactance [a perception of threat] and making people angry, it can backfire.”[64] There is evidence that encouraging shame may also be persuasive, though perhaps more likely to backfire than encouraging guilt.[65]
  • A meta-analysis of 36 studies found no significant overall effects of using intentional evocations of anger in persuasive messaging focused on attitudes. However, there were small positive effects when the anger evoked was relevant to the message and when the message was combined with strong arguments or an appeal that reassures the individuals that the recommended behavior will address the problem.[66]
  • Some studies (including two focused on opposition to animal exploitation) suggest that disgust may enhance persuasion.[67] However, as with guilt and fear appeals, other studies suggest negligible or counterproductive effects.[68]
  • Although persuasive messages that explicitly encourage positive emotions such as pride and joy are less well studied than those that encourage negative emotions like fear and guilt, such messages may also be effective.[69]
  • Hornik et al.’s (2016) meta-analysis of advertising studies found that humor had the second largest persuasive effect size of the seven tested appeal types (r = .35), above fear appeals and various more rational persuasion types.[70] A more recent meta-analysis focusing on a broader range of persuasion studies (not limited to advertising) found that the use of humor in persuasion has significant effects on knowledge (r = .23, k = 29), attitudes (r = .12, k = 58), and behavioral intent (r = .09, k = 29). However, the effects were insignificant and close to zero for each of these outcomes for both “political topics (k = 21), such as gun control and social security” and “health topics (k = 27), such as cervical cancer and mouth hygiene,” suggesting that humor is not likely to make messages relating to serious ethical issues either consistently more or less persuasive.[71] That said, if humor helps to increase public attention to a persuasive message (even if it does not increase its persuasiveness per se), then it could sometimes still be useful as long as it does not trivialize the movement or have other indirect negative effects.[72]
  • Similarly, Hornik et al. (2016) found sex appeals to have the largest persuasive effect size of all tested advertising appeal types (r = .46),[73] but communications scholar Ronald D. Smith (2017) notes that a “consistent finding from persuasion research is that sex appeal should not be used simply for shock value. It is far more effective when the sexual theme has a legitimate association with the product (such as lingerie, perfume, or condoms) or with the cause (such as birth control or responsible sexual behavior).”[74] A similar effect may occur in social advocacy. Indeed, a study testing the effects of PETA advertisements suggests that sexual ads will backfire for animal advocates.[75]
  • Fast speech rate may enhance persuasion by acting as a cue of credibility under conditions of low elaboration and may capture attention. But it may reduce the communicator’s apparent goodwill (so could be counterproductive for sensitive issues)[76] and comprehension by the audience.[77]
  • There are a number of factors that can make imagery persuasive,[78] though a meta-analysis of 12 studies found that, overall, adding “visual images to verbal texts had no significant effect on persuasion.”[79]
  • “Refutational two-sided messages” (those which describe but refute opposing arguments) “are dependably more persuasive than one-sided messages; nonrefutational two-sided messages, on the other hand, are slightly less persuasive than their one-sided counterparts.”[80] There are a number of caveats and exceptions where one-sided messages can be more effective.[81]

Relatedly, “showing receivers refutations of weak opposing arguments makes receivers more resistant to persuasion (by subsequent attack messages) than they otherwise would have been.”[82] This “inoculation” strategy is one of several possible strategies for preventing unwanted attitude change. A meta-analysis suggests that, “providing people with arguments and information supporting their current views” can confer resistance to persuasion, but that these “supportive” treatments are less effective than “inoculation.”[83] Warning a receiver that they are about to hear a message intended to persuade them also stimulates resistance to persuasion.[84]

Although not easily employed to alter public opinion, there are a number of techniques that have been found to increase the effectiveness of interpersonal persuasion efforts aimed at changing behaviors, such as the “foot-in-the-door,” “door-in-the-face,” “that’s-not-all,” “low-balling,” “fear-then-relief,” “but-you-are-free,” and “disrupt-then-reframe” techniques.[85] Techniques focused on persuasion within groups could potentially be applied at the level of whole organizations or even social movements, such as the finding that a member of a group with a minority view can sometimes persuade the majority to change their view by either “conforming with the group and then deviating” or “consistently disagreeing with the group.”[86] There is also evidence that certain advertising techniques — such as increased exposure to a product and association between the product and certain images or attributes — can positively influence attitudes toward the product. The success of these techniques is influenced by a variety of factors, but it seems plausible that they could sometimes be employed to influence public opinion towards policies or social issues.[87]

Other factors

Persuasion can also be affected by “receiver factors” such as demographic factors or the receiver’s mood,[88] which will often be difficult for advocates to account for but suggest that optimal persuasive messages will be tailored to their audiences. Some of this research has more generalizable advocacy-relevant implications. For example:

  • It may sometimes be possible to encourage receptivity to persuasive messages before sharing the message itself, such as by asking people to reflect on their values.[89] 
  • “Reactance” is a state created by a perception of threat to the receiver’s freedom, which makes the receiver more likely to reject a view advocated to them. It can be reduced by avoiding forceful claims (e.g. “it is impossible to deny all the evidence”) and emphasizing the receiver’s freedom of choice.[90]
  • It is harder to persuade someone if they are already knowledgeable about the topic.[91] This suggests that persuasive interventions focusing on professionals with relevant expertise are less likely to result in attitude change than interventions focusing on the general population. It also suggests that first impressions matter; advocates should ensure that awareness-raising tactics are persuasive to their audience, since it will be harder to change attitudes once knowledge has increased.

Psychologists Stuart Oskamp and P. Wesley Schultz (2005) summarize research findings as demonstrating that, “[p]rint media (books, magazines, and newspapers)... produce better comprehension and retention of complex material [as well as higher attitude change] than other media, but that this advantage does not hold for simple material… [P]eople’s knowledge of current affairs is more closely related to their use of print media.”[92] They also summarize that, “[t]here is general agreement that personal communication usually has a stronger influence on people’s attitudes and behavior than does mass communication.”[93] Of course, this does not necessarily mean that efforts to alter public opinion through personal communication will be more cost-effective, since they may also be more resource-intensive.

There is a substantial amount of research demonstrating a “third-person effect,” where people believe that a persuasive message will have a stronger influence on others than on themselves. For example, Paul et al.’s (2000) meta-analysis found an effect size of r = 0.50, i.e. substantially “greater perceived effects on others than on oneself,”[94] though Sun et al. (2008) found a smaller effect size (d = 0.65).[95] In some contexts, it may be sufficient for advocates to encourage the perception that an issue is important or an attitude is widely held;[96] the third-person effect suggests that doing so is tractable.


When advocates successfully encourage attitude change, a substantial proportion of the change may endure for at least several weeks, as Oskamp and Schultz (2005) summarize:

After 4 to 6 weeks, the amount of attitude change retained may be from one-third to two-thirds of the initial change, which of course may have been small to begin with. In a study of five different TV documentaries shown to college students, Fitzsimmons and Osburn (1968) found that only one retained a significant attitudinal effect after 4 weeks. However, many experiments have found attitude changes lasting as long as 6 months, and a very impressive classroom study by Rokeach (1971) showed significant attitude changes lasting well over one year.[97]

This is promising, given that, “all of these findings stem from studies in which the persuasive message was delivered only once” and that, “[r]esearch has shown that repeated re-exposures to a persuasive message will strengthen and prolong any prior opinion change.”[98] More recently, political scientist Seth J. Hill and colleagues have noted that, “[s]cholars do not usually test for the duration of the effects of mass communication.”[99] Nevertheless, a “handful of recent studies [have] found that persuasion effects can be quite shortlived, decaying in a few weeks or even a few days.”[100] A review of nine brief interventions intended to reduce implicit racial preferences found that “all nine interventions immediately reduced implicit preferences” but “none were effective after a delay of several hours to several days.”[101] Two experimental studies of a documentary “that presents the health, environmental, and animal welfare motivations for reducing consumption of meat and animal products” found that it “did not meaningfully affect any of the… exploratory attitude outcomes” at follow-up after two weeks.[102] Hill et al.’s own study (which uses political advertising data rather than an experimental design) and a subsequent experiment suggest that over half of the attitude change observed from persuasive interventions decayed within weeks, but that non-negligible proportions of the effects lasted for longer periods.[103] Regression across numerous studies of social norms manipulations found no significant effects of “[t]he time in days between message exposure and final assessment” on attitudes outcomes.[104]

A natural experiment from Germany suggests that one-sided persuasion campaigns may have effects on social norms, beliefs, and behaviors that last for years.[105] An observational study found that viewers of an advertisement designed to increase opposition to Canadian seal hunts still had significantly higher opposition two months after exposure than beforehand, though it had levelled off relative to their opposition immediately after viewing.[106] This provides some evidence that persuasive interventions may cause long-term attitude change, but it also seems clear that most of the attitude change they cause will be temporary. For advocates to cost-effectively cause long-term public opinion change, they therefore probably need to encourage some sort of self-perpetuating mechanism such as new legislation, social norms, or framings used by the media.

Prejudice reduction strategies

Research on advocacy and persuasion efforts that focus specifically on reducing prejudice is especially relevant to social movements targeting moral circle expansion. Psychologist Levy Paluck and colleagues (2021) reviewed and meta-analyzed “418 experiments reported in 309 manuscripts from 2007 to 2019 to assess which approaches work best and why.”[107] Table 1 is a summary of their findings.


Table 1: Effects of prejudice reduction strategies, summarized from Paluck et al. (2021)[108]

Entertainment“[E]ntertainment interventions have tested interactive narratives that allow individuals to participate in the construction of stories about outgroups, films made by and for Black audiences, pro-integration music lyrics, and entertainment education that incorporates educational messages about prejudice into an entertaining storyline of a soap opera or film.”0.430.270.5912
Value consistency and self-worth

“These interventions include reminders of

individuals’ or their group’s egalitarian preferences or history in order to inspire consistency with that history in the present moment, remind people of moral exemplars, and provoke introspection about one’s existing beliefs and prejudices.”

Extended and imaginary contact“[T]he majority of studies testing the extended contact hypothesis used fictional friends or characters in books or movies that belong to the same ingroup as the audience member to test whether the fictional character’s contact with an outgroup member would reduce prejudice.”0.370.30.44137
Social categorizationThese interventions “encourage participants to rethink group boundaries or to prioritize common identities shared with specific outgroups.”0.370.270.4659
Overall 0.360.310.4416
Cognitive and emotional training“Interventions categorized as cognitive and emotional training share the idea that individuals can be trained to use thinking and emotion regulation strategies to fight off their personal implicit or explicit prejudices.”0.340.250.43104
Multicultural, antibias, moral education“Antibias education and multicultural education draw variously on theories addressing the socialization of prejudice, cognitive and moral development, and learning. The form of these interventions also ranges widely.”0.300.180.4220
Diversity trainings“The notion of diversity training encompasses a wide category of interventions that are ‘designed to attack bias’ among managers and workers.”0.30-0.120.716
Interpersonal contactContact between members of groups. E.g. included studies “randomized criminology students to have contact with individuals incarcerated for serious crimes” or “randomly assigned Jewish and Arab Israelis to meet one another on peace encounters.”
Peer influence, discussion / dialogue“This category of interventions is united by the idea that people who share important identities, peers, or ingroup members have a powerful influence over one another’s impression of the attitudes and behaviors that are typical, desirable, and correct. The interventions in this category wield peer influence in various ways to reduce prejudice.”

LL = 95% confidence interval lower limit; UL = 95% confidence interval upper limit; n = number of outcomes.


They found similar sized effects for “explicit attitudes or beliefs” (d = 0.35, 95% CI [0.3, 0.39], n = 335), behavior (d = 0.42, 95% CI [0.29, 0.55], n = 50), and other outcome types, though Table 1 combines all the outcome types they measured.[109]

They find evidence of substantial publication bias, noting that, “the average effect size drops 48%, to 0.187, when we focus solely on the top quintile of sample sizes” and that their analysis suggests that “a study large enough to generate a standard error of approximately zero would, on average, produce no change in prejudice at all.”[110]

The outcome measures tend to focus on prejudice itself, which is arguably less useful for advocates than measures of support for policies that affect marginalized groups.[111] However, we might expect many interventions to affect both types of outcomes. For example, there is evidence that intergroup contact affects both.[112]

Some of the reviewed intervention types, such as intergroup contact and various training types, seem likely to be very expensive. It therefore remains unclear whether these mechanisms could be employed to cost-effectively encourage widespread attitude change.[113] Paluck et al. note that 76% of their included studies evaluate “treatments that are easy to implement, brief (under 10 minutes), inexpensive, and thought to have lasting effects,” but highlight that there is little evidence about the long-term effects of these interventions.[114]


“Framing” variations might influence public opinion through different psychological mechanisms to direct persuasion efforts. Political scientist Thomas E. Nelson and colleagues (1997) explain that, by highlighting certain aspects of a topic over others, “[f]rames may supply no new information” and have no effect on the recipient’s beliefs about the topic, “yet their influence on our opinions may be decisive through their effect on the perceived relevance [“weight”] of alternative considerations.”[115] In contrast, traditional persuasion influences attitudes by providing new information and altering beliefs without influencing the weight of those beliefs.

For example, when presented with a choice of options that involve risk, the option that people are most likely to select varies substantially according to whether the positives (e.g. “lives saved”) or negatives (e.g. “lives lost”) are emphasised by the question, even when the options presented are logically identical. A meta-analysis of 136 empirical papers found a small effect (d = 0.31) from such framing variations, with respondents more likely to avoid risk when the positives are emphasized.[116]

Various other small differences in survey question wording (or perhaps ballot language[117]) that highlight certain aspects of a topic can elicit substantially different levels of support. A common survey question on the death penalty is Gallup’s “Are you in favor of the death penalty for persons convicted of murder?” — this tends to receive majority approval in the US, but support for the death penalty can fall by 30% or more when the question instead asks whether the respondents support the death penalty or life without parole for convicted murderers.[118] Political scientists Dennis Chong and James N. Druckman (2007) note that variations in framing have been demonstrated to affect policy preferences in “experiments, surveys, and case studies across a range of issues.”[119] 

Such framing effects could occur at the level of societal discourse of a topic. As discussed in the section on “The media’s agenda-setting effects” below, advocates could influence such discourse via the media. Encouraging widespread adoption of a particular frame could take a long time but have substantial effects on policy preferences.[120] It therefore seems useful for advocates to identify the frames that are most persuasive to their audiences and then to apply these in their messages.[121]


Chong and Druckman (2007) cite evidence that framing efforts are more effective if:

  • The audience perceives the connection being made between the frame and the issue as valid,
  • The audience does not have strong values that contradict the frame,
  • The frame is “delivered by credible sources,” or
  • The frame “invoke[s] longstanding cultural values.”[122]

Given that framing can operate through different mechanisms to persuasion, it might have different moderators. For example, Nelson et al.’s (1997) experiment found evidence that framing variations have stronger effects for people with high “domain-specific knowledge about the arguments surrounding an issue,”[123] whereas other studies have found that the opposite tends to be the case for persuasion attempts.[124] So even if persuasion attempts fail, reframing the issue might still successfully shift overall attitudes, and vice versa.[125]


As with research focused on the duration of the persuasive messages, some studies provide evidence that at least some attitude change caused by a single positively or negatively framed article can persist for several weeks, but that the size of the effect will diminish during that time.[126] Presumably then, for a new frame to have substantial, long-lasting effects on attitudes, advocates would need to engage in long-term campaigns with repeated messaging using the new frame (which could be very expensive), or successfully encourage the media, politicians, or other influencers to adopt the frame as well. Indeed, a review of 16 longitudinal, experimental studies testing how long news framing effects last found that “[s]tudies focusing on repetitive framing are [not] conclusive but suggest that repetitive news frame exposure strengthens the framing effect to some extent.”[127] They also found that:

  • “Studies focusing on competitive framing often conclude that news framing effects persist beyond initial exposure, but are relatively easily altered, sometimes only one day later, by competitive exposure.”[128]
  • “[S]tudies using non-salient issues are able to detect longer-lasting framing effects, probably because there is less exposure to issue-relevant news in the interim period between initial and delayed post-tests.”[129]
  • “[N]egative news frames are likely to have stronger and therefore longer-lasting effects on opinions than positive news frames”[130]

Protest and social movement events

Protest could presumably influence public opinion either by spreading persuasive arguments or affecting how social issues are framed and reported in the media. A recent observational study found that activism by the women’s movement has had substantial cumulative effects on gender attitudes such as support for female presidents.[131] However, other studies suggest that the cumulative effect of social movement protest and mobilization on public opinion is not always significant and positive, as sociologists Edwin Amenta and Francesca Polletta (2019) summarize:

Research indicates that protest may not budge public opinion, as was the case with the Occupy movement and anti–Vietnam War protests. A movement’s impact may be canceled out by the impact of a counter movement, as was the case with environmentalists on climate change. Or movement action may backfire, leading to more negative views of the group or issue, as was the case for nuclear freeze proponents. Movements’ influence on public opinion may depend on their being endorsed by more mainstream political elites. Anti–Vietnam War sentiment was limited until political leaders and reporters began to criticize the war, and public support for the Equal Rights Amendment in Oklahoma declined after legislators rejected it. At a cross-national level, public opinion on issues of importance to a movement’s constituency may be driven more by political regime type, demographics, religion, and economic development than by the existence of a movement.[132] 

Some studies find that protests can improve attitudes towards movements’ intended beneficiaries[133] and attitudes towards the movements themselves,[134] though violent protest can have negative effects on such attitudes.[135]

Activism can be widely covered in the media,[136] which can in turn influence the public’s assessment of how important affected issues are.[137] Relatedly, a number of studies have found that social movement activism can have significant effects on the political agenda, encouraging hearings and the introduction of legislation relevant to the targeted issues.[138]


This section does not attempt to review all possible causes of public opinion change; the focus is on factors that seem especially well-studied in the academic literature or especially relevant to social movements targeting moral circle expansion.

The media’s persuasive effects

Psychologists Stuart Oskamp and P. Wesley Schultz (2005) summarize that the scholarly consensus on whether and when the media tends to affect public opinion or not has changed:

The first, a powerful effects model, was dominant from the 1920s through the 1940s, as illustrated in the deep fear of the possibly irresistible effects of propaganda on a defenseless public… It was followed by the minimal effects model, articulated by Klapper (1960), based on the many empirical studies which found no effects or very limited effects of the media in changing people’s beliefs, attitudes, and behavior. More recently, a model of powerful effects under limiting conditions has gained more adherents. It denies the early all-powerful view of the media, but stresses that they have important effects in particular circumstances and with particular individuals. Thus current research is apt to focus on the interacting variables and contingent conditions under which media effects will emerge most clearly—for instance, under conditions of heavy viewing and weak prior predispositions. Furthermore, current conceptions include a wide range of media effects—not just changing attitudes, but also forming new attitudes, beliefs, or behaviors, reinforcing already existing ones, and crystallizing previously vague or unstated beliefs or attitudes.[139]

They summarize that there is evidence that, “mass communication usually serves to reinforce existing attitudes and opinions” and “[w]hen mass communication does produce attitude change, minor change in the extremity or intensity of the attitude is much more common than is ‘conversion’ from one side of an issue to the other side.”[140]

This does not preclude the possibility that advocates could encourage substantial public opinion changes in usual circumstances, such as if they manage to encourage sustained shifts in the overall tone of media coverage. Indeed, there is evidence that where “coverage of a public issue is unbalanced in a pro or con direction, public opinion is likely to shift in that direction subsequently.”[141] This may be tractable if advocates can strategically raise the salience of certain attributes (see “Attribute salience and effects on public opinion” below) or create newsworthy persuasive materials such as documentaries or exposes that shift the balance of what news is being covered.[142]

There are several theories about how the media influences public opinion. These could be relevant to advocates, but tend to make broad claims and have mixed results in empirical tests.[143] For example, in the “two-step flow” theory, “a small minority of ‘opinion leaders’... act as intermediaries between the mass media and the majority of society,”[144] suggesting that it may be most cost-effective to focus resources on changing the opinions of influential audience members. However, some research suggests that, at least in some circumstances, it could be more cost-effective to ignore supposed opinion leaders and focus on whoever can be persuaded most easily.[145]


Oskamp and Schultz (2005) note that, “[m]ass communication can be quite effective in changing attitudes in areas where people’s existing opinions are weak” and “can be quite effective in creating opinions on new issues where there are no existing predispositions to reinforce.”[146] Presumably then, advocates could influence public opinion towards less well-known attributes or sub-topics (e.g. a specific farming practice) if they are able to shape the initial media coverage. It also suggests that tactics that are aimed at increasing awareness without necessarily being persuasive may be counterproductive for issues where public opinion is currently unfavorable because they may make subsequent attitude change more difficult.

Many of the other moderators described in the section above on “Direct advocacy and persuasion” likely hold in the context of media messaging.


Many of the studies of the duration of persuasive effects described in the section above on “Direct advocacy and persuasion” were tested using media articles or data; the effects from a single exposure to persuasive media content will diminish with time, though there may be some lasting effects, and repeated exposure to similar content may encourage longer-term attitude change.

The media’s agenda-setting effects

Agenda-setting research finds evidence that, when the news media covers certain issues, the public tends to increase its evaluation of how important those issues are, i.e. that the media “can have strong, direct effects in the short term by influencing not what people think, but what they think about.”[147] This effect has been identified using a wide variety of methodologies, issue foci, geographical foci, and media types, but the prototypical design is that the media’s agenda is assessed through content analysis, the public’s agenda is assessed through survey questions, and the correlation between the two is then estimated.[148] Luo et al.’s (2019) meta-analysis found that the mean correlation from the 67 included studies was 0.49.[149] Maxwell McCombs and Sebastian Valenzuela, prominent scholars of the media’s agenda-setting effects, summarize in their book Setting the Agenda (2021) evidence that these correlations are usually mostly explained by the media’s effect on the public, rather than by the public’s effect on the media or some other factor:

  • Experiments have also demonstrated the media’s agenda-setting effects,
  • Comparisons between the media and public agenda at different time points tend to suggest that the media changes focus first, with the public following afterwards, and
  • Studies have found correlations between the media and public agendas even when controlling for plausible lurking variables.[150]

There is also evidence that the media has various other effects related to its public agenda-setting function: it can influence the political agenda,[151] the public’s knowledge and perceptions of reality,[152] and the public’s views about which criteria should be used to evaluate politicians.[153]

The agenda-setting hypothesis suggests that advocates can probably influence the public agenda if they are able to increase media attention to an issue.[154] The logical next questions are therefore: can social movements influence the media agenda, and if so, how? These questions are not the focus of this review, though McCombs and Valenzuela summarize some evidence that public relations professionals have a substantial influence on the media agenda[155] and some studies have found that social movement protests have influenced media coverage and the public agenda.[156]


McCombs and Valenzuela provide evidence that the media tends to have weaker effects if the public has high personal experience of an issue or is otherwise already relatively certain about an issue’s importance.[157] For example, studies have found lower correlations between media coverage and the public’s perceived importance of issues like crime and the cost of living — which affect the public very directly — than between media coverage and the public’s perceived importance of issues like pollution, drug abuse, and energy.[158] One implication of these findings for social movements is that successfully attracting media coverage of institutional campaigns may have a stronger agenda-setting effect than successfully attracting media coverage of individual diet change topics.

Relatedly, McCombs and Valenzuela summarize one study which seems to suggest that media coverage may have lower agenda-setting effects if the public already has higher awareness about a topic.[159] This suggests that if certain topics have been on the agenda for some time already, further efforts to raise their salience may be less effective.

Demographic variables tend to have little or no moderating effect on agenda setting. There is evidence from several studies that people with more years of education more closely mirror the media agenda, though the difference is very small.[160]


McCombs and Valenzuela summarize that, “the point of decay of agenda-setting effects, defined as the point in time where significant correlations between the media agenda and the public agenda disappear, ranges from eight to twenty-six weeks.”[161] This suggests that advocates should usually not attempt to increase the salience of certain topics via the media unless they suspect that they will be able to sustain high media attention, since, if they are only successful in temporarily raising media attention, the effects on the public’s perceptions of its importance will also be short-lived.

Advocates might also seek to increase salience if they have specific, strategic goals in mind for why higher salience will be useful at a particular time, e.g. to support a legislative campaign on an issue with high public support. McCombs and Valenzuela note, however, that, “the span of time involved in the transfer of issue salience from the media to the public agenda is generally in the range of four to eight weeks,”[162] so advocates need to account for this time lag in their planning.

Attribute salience and effects on public opinion

The “first level” of agenda setting research focuses on assessing whether and how media attention influences the salience and perceived importance of certain “objects” (i.e. topics) among the public, while the “second level” assesses whether and how media attention influences the salience and perceived importance of more specific attributes of those objects.[163] Luo et al. (2019) found that the correlations tended to be larger in studies of first level agenda-setting than studies of second-level agenda-setting — this difference was significant at p < .10 (β = 0.53).[164] McCombs and Valenzuela also note that the media only seems to have much of a second level agenda-setting effect where “both the political system and the news media are reasonably open and free.”[165]

There is some evidence that increases in the salience of certain attributes can have disproportionately large effects on the salience of the broader object.[166] More importantly, there is also evidence that certain attributes tend to be covered in a more positive tone in the media than other attributes; if those attributes (and their corresponding more or less positive coverage[167]) become more salient, then this can affect public opinion on the topic as a whole.[168] This suggests that advocates can potentially alter both the public’s perceived importance of a topic and public opinion on that topic by identifying specific attributes that tend to be seen more or less positively, then strategically raising the salience of certain attributes while avoiding raising the salience of others.

Politicians and celebrities

Politicians and celebrities can advocate particular ideas — presumably their persuasion efforts may be successful and are subject to the same moderators as other communicators.[169] A number of studies have directly evaluated the persuasive and agenda-setting effects of politicians and celebrities too.

Several studies suggest that politicians can increase approval for specific policies by framing them as directed towards broad goals that the public support,[170] and numerous others suggest that presidents can intentionally shape public opinion through persuasive messaging.[171] Benoit et al.’s (2003) meta-analysis found that viewing US presidential debates had an effect size of .14 (95% CI .03–.32, k = 4) on “preference for one candidate’s issue positions over another’s.”[172] So presidents can influence public opinion, though the effects may be small.

There is also evidence that political parties’ positions influence the attitudes of partisan supporters.[173] This influence may be stronger when party positions are polarized[174] and when the public is less informed about an issue.[175] Cumulatively, repeated cues from political elites pushing in a certain direction could have substantial effects on public opinion, whereas competing partisan cues in polarized debates could roughly balance each other out and potentially even drown out the effects of other persuasion efforts.[176]

In addition to altering attitudes through persuasion or reframing, politicians can affect the public’s perceptions of the importance of issues via comments, actions, and press releases that are reported in the media.[177] Benoit et al.’s (2003) meta-analysis found that US presidential debates had a mean weighted agenda-setting effect size of .29 (95% CI .22–.44, k = 3).[178]

There is evidence that celebrities can have substantial effects on attitudes, though they are not always more effective than alternative spokespeople.[179] Some advertising studies have found support for the “matchup hypothesis,” where “[c]elebrities or other endorsers are particularly apt to enhance consumer attitudes when their characteristics ‘match up with’ or are relevant to the product being promoted,”[180] and we might expect a similar effect to hold for promotion of policies.

Even if celebrities or politicians are not seen as credible to speak on a particular issue, their involvement could be beneficial if it helps to bring attention to certain issues and arguments, either through a second-level agenda-setting effect[181] or because it increases the prevalence of persuasive arguments pushing in a certain direction.[182] But their involvement could backfire for the same reasons, so it seems more important to focus on getting these public figures promoting the right messages than on maximizing their general discussion of an issue.

Policy change

Oskamp and Schultz (2005) summarize several studies showing that public opinion often follows US foreign policy quite closely.[183] There is evidence from observational analyses and social movement case studies that public opinion changes can occur from policies affecting social issues and the breadth of the moral circle, too. For example, numerous countries have seen public support for the death penalty decline since it was abolished.[184] Some studies suggest that international policies and policies in neighboring jurisdictions can also affect public opinion.[185] When the Supreme Court makes a decision, this tends to cause public opinion to move towards the opinion implied by that decision, though this does not always happen.[186] This all suggests that if advocates can encourage policy change, public opinion will tend to move towards support for those policies. Nevertheless, political scientist James Stimson (2015) presents evidence that public preferences regarding the general direction of further government action sometimes shift in the opposite direction to trends in government policy itself.[187]

Indirect or long-term factors

Reliably changing public opinion is less tractable if the main causes of fluctuations are slow-moving and long-term, or only have an indirect relationship to change and are therefore difficult to predict and control.[188]

Referring to the impact of events on public opinion about foreign policy, Oskamp and Schultz (2005) summarize that:

[E]ven spectacular events have no effect on attitudes, or they may cause only a brief fluctuation followed by a return to the preexisting attitude… When rapid attitude shifts occur, they are usually related to major events in international affairs or in the economy, such as the improvement in U.S.-Soviet relations that took place under Gorbachev, and particularly the fall of the Berlin Wall, which signaled the end of the Cold War. As a contrasting example, after the Chinese army's massacre of student protesters in Tiananmen Square in 1989, the percentage of Americans who expressed favorable opinions of China plunged briefly from 72% to 31%. However, even the most dramatic changes in political alignments usually involve attitude changes by only 20% to 30% of the population, and such changes almost always involve a combination of spectacular events and cumulative events. Either type of event alone is apt to produce attitude changes of no more than 10%.[189]

External events have been found to affect a range of other social and political attitudes, such as hurricanes affecting support for pro-environmental politicians, financial crisis affecting support for conservative economic policies, and nuclear disaster decreasing support for the use of nuclear power.[190] However, as with the events affecting foreign policy opinions, their effects are often small or temporary.[191]

Public opinion surveys usually find that some demographic variables are statistically significant predictors. For example, Sentience Institute’s US surveys find that women, younger people, more liberal people, Democrats, black and Hispanic people, people from the Northeast of the US, and vegetarians and vegans have significantly higher Animal Farming Opposition than other respondents.[192] We should expect, then, that demographic trends might lead to changes in overall public opinion,[193] though such changes might be slow and difficult to detect.[194]

Of course, even if advocates cannot easily influence indirect or long-term factors, they may be able to take advantage of them. For example, it might be possible to implement institutional changes during a temporary spike in public support for certain policies.


  • Where I have cited textbooks in this review, I have not tended to look up the individual citations for the claims made. Often, the textbooks will provide a general characterization of certain streams of research and then provide a small number of illustrative examples, rather than rigorously reviewing the strength of evidence for various claims. In this sense, it is possible that some of the characterizations here do not actually have much supporting evidence.[195]
  • Reviewed research items often discussed the strengths and weaknesses of different research approaches (e.g. field studies vs. laboratory experiments) and whether the key assumptions of various theories had been tested directly or indirectly, but rarely assessed the risk of bias in included studies.
  • I have attempted to avoid selective reporting of findings, but I have not attempted to make the coverage of particular topics in this review systematic or comprehensive. It is possible that some of the findings cited here have been critiqued and challenged by other research that I have not seen.
  • Given the many different streams of research relating to the causes of public opinion change, it is likely that some have accidentally been missed. Others have been intentionally excluded in the interests of brevity (e.g. if they had unclear implications for advocacy).
  • This review mostly focuses on strategies or factors influencing public opinion that are well-studied in the academic literature. It may therefore miss effective but less well-studied strategies.


  • This review has focused on the causes of public opinion change under the assumption that public opinion is important for at least some institutional and social movement outcomes. This assumption has been evaluated by sociologists and political scientists — a review of relevant research on the effects of public opinion change could be informative both for testing the assumption and assessing how to most effectively encourage the intended outcomes. It could also shed further light on the questions examined in this review, since in asking how advocacy and public opinion affect policy, researchers sometimes implicitly evaluate how advocacy affects public opinion.[196]
  • This review has covered a wide range of topics fairly briefly; more systematic, in-depth reviews of some of these topics could be valuable. For example, there does not currently seem to be a meta-analysis or detailed review of the effects of protests and activism on public opinion, even though a number of studies exist on this topic. Additionally, a more comprehensive comparison of the effect sizes of various intervention types on public opinion could be valuable, albeit time consuming.[197]
  • Experiments could be conducted to more directly assess the relative effectiveness of different influences on public opinion. For example, which matters more: the frames used in media coverage (e.g. animal protection vs. environmental vs. human health focus), or the tone (positive or negative)? Which has a larger effect: a story about a favorable speech by a politician or a story about a well-attended public protest?
  • Experiments could be conducted that test out findings from persuasion research specifically in the context of social movements focusing on moral circle expansion. For example, are refutational two-sided messages more effective than one-sided messages at building support for institutional reforms that affect animals, like bans on factory farming? Which sources are seen as more or less credible (and hence more or less persuasive in certain contexts) when encouraging moral concern for artificial sentience: AI researchers, philosophers, cognitive scientists, robots, politicians, or members of the public?
  • American public opinion has changed substantially on a number of issues, such as the right to gay marriage, legalisation of marijuana, desegregation, and whether the public would vote for qualified women or African Americans for president.[198] Case studies of these issues could reveal insight into the causes of public opinion change.
  • The research on framing and second-level agenda setting suggests that media coverage and persuasive messages will vary substantially in their persuasiveness depending on their content. There are a wide variety of research types that could be conducted to shed light on which frames and attributes specific social movements should prioritize, including content analyses, focus groups, surveys, experiments, and more informal exploration.[199]
  • It might theoretically be possible to influence attitude formation through repeated positive initial exposure to certain ideas or shaping which attitudes parents pass on to their children (e.g. via educational or supportive services).[200] In general, such factors shaping initial attitude formation seem likely to be difficult to influence, so this review has mostly ignored them, but a review of relevant research could be useful.[201]
  • A review of studies with outcomes different from but related to attitudes — such as knowledge, attention, memory, emotion, identity and so on — could be helpful, e.g. for social marketing purposes.


See the version on Sentience Institute's website.


2 comments, sorted by Highlighting new comments since Today at 3:36 AM
New Comment

Thanks for writing this up! One minor question: when you say

Marketing scholar Jacob Hornik and colleagues (2016) combined 2,276 different effect sizes from advertising studies in meta-analysis and found a mean weighted effect size on persuasion outcomes (including “attitude toward the product, attitude toward the brand, purchase intention, and product choice”) of r = .19.

I think that's referencing this from the paper:

The overall reliabilitycorrected and sample-weighted correlations between appeals and response behavior are .22 and .19 (the uncorrected correlations are .23 and .20) for Aad and persuasion, respectively (with a 95% confidence interval from .13 to .27)

I'm kind of confused what we are measuring the correlation between. I think this is the correlation between how much someone says that they like an advertisement and how likely they are to actually purchase the product, right?

It could be that this correlation is very low yet advertising is extremely effective, or vice versa, right?

(This isn't really an important point, I was just confused why people were using r, and figured other people might have the same confusion.)

Yeah I'm also a little confused about why they're using r without digging back into it in detail. But if I read it correctly, then their correlation coefficient there somehow pools together pretty weak proxies for behaviour ("attitude toward the product, attitude toward the brand," potentially also "purchase intention") with actual behaviour ("product choice").

I definitely don't think that we should pay too much attention to the findings of that particular meta-analysis when thinking about how to change attitudes or behaviour in the context of the farmed animal movement or other EA-adjacent cause areas. But it is still weakly relevant evidence and it would have been disingenuous of me not to include it, I think. (My prior was that using humour and sex appeals are both usually pretty bad ideas for serious social movements, especially the latter.)