Persuasion, Ethics and Law: What Moves People, What Does Not, and Where the Lines Are

The first eight lessons of this course are about getting attention and keeping it. This one is about what attention can accomplish once you have it, which turns out to be less than most campaign plans assume and more than cynics claim, and about the legal and ethical lines a page crosses most easily when it is winning. The two subjects belong together. Pages drift toward the ethical edge when they believe the next outrageous post will change minds. The evidence says it will not, which removes the temptation.

What the field experiments found

Persuasion research moved from the lab to the field over the past twenty years, and the field results humbled nearly everyone. The studies below are randomized experiments with real voters, not surveys or focus groups, and they report effects in units you can picture.

Study	Design and sample	Effect	What it tells an advocacy page
Kalla and Broockman, APSR 2018	Meta-analysis of 40 field experiments plus 9 original ones, 49 in all, on campaign contact and advertising in U.S. elections	Average persuasive effect on candidate choice in general elections: zero. Contact within two months of the vote: 0.58 percentage points, 95 percent CI -0.50 to 1.66. Effects found only when candidates took unusually unpopular positions and campaigns targeted persuadables, or very early with decay. Primary and ballot measure campaigns showed meaningful effects; one 2015 mayoral primary canvass raised support about 11 points	Do not plan to flip general election voters at scale. Ballot measures and primaries, which is most Idaho local advocacy, are different
Coppock, Hill and Vavreck, Science Advances 2020	59 experiments with 49 real ads from the 2016 presidential campaign, about 34,000 YouGov panelists, fielded weekly March to November	Candidate favorability: about 0.05 points on a 1 to 5 scale (0.049), significant only because the sample was huge. Vote choice: about 0.7 points, not significant. No meaningful difference by party, battleground status, attack versus promotional, or PAC versus campaign sponsor	The format of an ad matters far less than people believe. Small, uniform effects
Broockman and Kalla, Science 2016	501 voters, 56 canvassers, door to door in South Florida; a single roughly 10-minute conversation encouraging perspective taking	Reduced transphobia by more than Americans' average decrease in homophobia from 1998 to 2012; effects lasted at least three months and survived counterarguments; increased support for a nondiscrimination law	Durable attitude change comes from long, listening conversations, which a page can organize but cannot replace
Gerber, Green and Larimer, APSR 2008	180,002 households, Michigan August 2006 primary, mailers randomized	Turnout against a 29.7 percent control: civic duty mailer +1.8 points; told they were being studied +2.5; shown their own voting record +4.9; shown neighbors' records +8.1	Mobilization of people who already lean your way is where large effects live, and social accountability drives it

Read the first row carefully, because it is often misquoted. Kalla and Broockman did not find that persuasion is impossible. They found it is close to impossible in general elections, where party identity does the work, and that it is real in primaries and on ballot measures, where voters lack a party cue. Their meta-analysis notes that nearly all the primary and ballot measure elections it covered were competitive. For an Idaho advocacy network that lives on levies, bonds, initiatives, county commission primaries and planning decisions, that distinction is the whole game. Your audience often has no party shortcut for the question in front of them, and that is when information and contact can move them.

Coppock, Hill and Vavreck add a second humbling point. Their 59 experiments looked for the heterogeneity that campaign lore predicts, the idea that the right message to the right person in the right place has big effects even if the average is small. They did not find it. Attack and promotional ads worked about equally. Battleground and non-battleground residents differed by 0.008 scale points. Democrats, independents and Republicans responded with differences smaller than their standard errors. The small average was not hiding large offsetting effects. It was small all the way down.

Against that, the Broockman and Kalla canvass is a different kind of result. The conversation was long, non-judgmental, and built around asking the voter to recall a time they were judged for being different. It changed a measured attitude by more than a decade and a half of national drift, and the change was still there three months later when the researchers tried to argue people out of it. Nothing a page posts will do that. What a page can do is recruit, train and schedule the people who have those conversations, and that is a mobilization objective.

What this means for an advocacy page

Put the evidence together and a page's realistic jobs come into focus. Rank them.

1. Mobilization. Getting people who already agree to show up, sign, call, testify and vote. The effects here are measured in whole percentage points, as in Gerber, Green and Larimer's 8.1-point neighbors mailer, and social accountability is the mechanism: people act when others they know will see whether they acted. A page that names the hearing date, shows who is coming, and thanks people publicly afterward is using that mechanism.
2. Agenda setting. Deciding what people are thinking about, rather than what they conclude. McCombs and Shaw's 1972 study in Public Opinion Quarterly argued that by reflecting what candidates say, media "may well determine the important issues" and that readers learn how much importance to attach to an issue from the amount and position of coverage. A local advocacy page is now part of local media. Getting the water rate structure onto the commission's agenda and into the newspaper's follow-up questions is a win the persuasion literature supports.
3. Relationship building. The slow accumulation of trust that makes an audience open your emails and believe your documents when the contested moment arrives. The experiments do not measure this directly, but the deep canvassing result points at it: change happened through a listening relationship, however brief.
4. Information for the undecided on measures and primaries. The one place direct persuasion shows real effects. Here the content science of lesson 05, the documents, the narrative, the plain arithmetic, earns its keep.
5. Conversion of opponents in general elections. Last, and nearly hopeless by the evidence. Budget nothing for it.

Misinformation science: prebunking, nudging and correcting

An advocacy page meets false claims weekly: a rumor about what the bond funds, a doctored chart, a quote the mayor never said. Three lines of research tell you what to do, and a fourth tells you what not to fear.

Prebunking

Roozenbeek, van der Linden, Goldberg, Rathje and Lewandowsky tested short videos that each taught viewers to spot one manipulation technique: emotional language, incoherence, false dichotomies, scapegoating and ad hominem attacks. Across six randomized lab studies with 6,464 participants and a YouTube field study with 22,632 participants, the videos improved recognition of the techniques, improved confidence in spotting them, improved judgments of trustworthy versus untrustworthy content, and improved sharing decisions, with effects holding across the political spectrum. Lab effect sizes for technique recognition ran from Cohen's d of 0.28 (scapegoating) to 0.68 (false dichotomies). The YouTube field effect was smaller, about 5 percent improvement in recognition, Cohen's h of 0.09, which is what you would expect from a 90-second ad watched once. The paper is in Science Advances, 2022. The lesson for a page is to teach your audience the tricks before the campaign season, when nobody is defensive, in the form of short explainers: "Here is how a false dichotomy about the bond will look when it arrives."

Accuracy nudges

Pennycook, Epstein, Mosleh and colleagues found in Nature in 2021 that people's sharing decisions barely tracked whether a headline was true, even though their accuracy judgments did, and that most said sharing accurate news mattered to them. Simply prompting people to think about accuracy closed part of the gap. In survey experiments with samples from 710 to 1,268, sharing discernment was 2.0 to 2.4 times larger in the prompted groups. In a Twitter field experiment with 5,379 users who had previously shared low-quality sources, a single message asking them to rate the accuracy of an unrelated headline raised the average quality of the news they then shared by 4.8 percent on days they tweeted. The effect is modest and the mechanism is inattention, not malice. For a page, this means your own posting template can include an accuracy beat ("Here is the document; check it") that primes your audience the same way, and that your comment moderation can ask "is this accurate?" rather than "this is false."

Corrections, and the backfire that mostly is not there

Nyhan and Reifler's 2010 experiments in Political Behavior documented several instances where a correction increased misperceptions among the targeted ideological group, and the finding spread far beyond its evidence. Wood and Porter went looking for it at scale: five experiments, more than 10,100 subjects, 52 issues chosen for their potential to backfire, corrections of politicians' misstatements. They found "no corrections capable of triggering backfire." People accepted factual corrections even when those corrections cut against their partisan commitments. The paper, The Elusive Backfire Effect, appeared in Political Behavior in 2019. The 22 scholars who wrote The Debunking Handbook 2020 reached the same conclusion: backfire effects occur occasionally, cannot be reliably predicted, and the fear of them should not stop anyone from correcting misinformation. The handbook's recommended structure is to lead with the fact if it can be stated simply, warn that a myth is coming and state it once, explain how the myth misleads, and close by restating the fact. Give a causal alternative that fills the gap the myth filled. The social posts course covers the mechanics in Rebutting Without Spreading.

Law: defamation

An advocacy page criticizes officials, developers, boards and sometimes neighbors, and the law of defamation is the first thing an angry subject reaches for. The constitutional framework is clear and protective of criticism of public conduct. In New York Times Co. v. Sullivan (1964), the Supreme Court held that a public official cannot recover damages for a defamatory falsehood about official conduct unless the statement was made with "actual malice," meaning "with knowledge that it was false or with reckless disregard of whether it was false or not." Later cases extended that standard to public figures. In Gertz v. Robert Welch, Inc. (1974), the Court held that states may set their own standard for private individuals "so long as they do not impose liability without fault," and that presumed or punitive damages require knowledge of falsity or reckless disregard even for private plaintiffs.

Who is criticized	What they must prove	Source	Practical rule for a page
Public official, about official conduct	Falsity plus actual malice (knowledge of falsity or reckless disregard)	New York Times v. Sullivan (1964)	Criticize conduct hard, cite the record, correct errors fast; a documented good-faith mistake is not actual malice
Public figure (a prominent developer, a well-known activist)	Falsity plus actual malice	Cases following Sullivan	Same, and be careful calling someone a public figure who only became known because of your coverage
Private person	Falsity plus at least negligence (fault), under state law; actual malice for presumed or punitive damages	Gertz v. Robert Welch (1974)	Highest care. Verify before naming a private person; usually describe the conduct and omit the name
Any plaintiff in Idaho	Must sue within two years	Idaho Code 5-219(5)	Keep your sourcing files at least that long

Idaho's Idaho Code 5-219 lists libel and slander among the actions that must be brought "within two (2) years." Idaho also still has a criminal libel chapter on the books: Idaho Code 18-4801 defines libel as "a malicious defamation, expressed either by writing, printing, or by signs or pictures," and 18-4802 sets the punishment at a fine up to $5,000 or up to six months in county jail. Criminal libel statutes are rarely used and face constitutional limits of their own after Sullivan, but the chapter exists and a page should know it does.

Three habits keep a page on the right side of all of this. Say what the record shows and link the record. Separate fact from opinion in your wording ("the minutes show the vote was 3 to 2" versus "this was a bad vote"). And correct errors visibly and quickly, with the original post edited and the correction noted, because a prompt correction is strong evidence against reckless disregard. This course does not advise anyone on a specific dispute; it describes the standards so a page can build habits that stay well inside them.

Law: Idaho's synthetic media election statute

In 2024 Idaho enacted House Bill 664, signed March 25, 2024 as Session Law Chapter 172 and effective the same day, which added Idaho Code 67-6628A, titled the "Freedom From AI-Rigged (FAIR) Elections Act." The statute defines synthetic media as an audio or video recording of a person's speech or conduct, created with generative adversarial networks or other digital tools, that appears real but is false: to a reasonable viewer it depicts an event, action or speech that did not occur, and it gives what the statute calls a fundamentally different impression than the unaltered original would. It applies to electioneering communications, which Idaho Code 67-6602(6) defines as communications that unambiguously refer to a candidate, are distributed within 30 days of a primary or 60 days of a general election, and reach an audience that includes that candidate's electorate, excluding news, editorials, commentary and letters in periodicals or broadcasts not controlled by a candidate or committee, communications in the regular course of business, and membership communications sent only to members.

The statute is not a flat ban. A candidate whose action or speech is deceptively depicted may seek an injunction to stop publication and may sue the information content provider (the person responsible for creating the content) for general and special damages. The plaintiff must prove the use of synthetic media by clear and convincing evidence, courts are encouraged to act quickly, and the prevailing party may be awarded fees and costs. The medium that distributes the communication is generally not liable unless it strips the disclosure or alters the content into synthetic media, and federally licensed broadcasters transmitting candidate communications under 47 U.S.C. 315 are excluded.

The disclosure that provides an affirmative defense is spelled out. The communication must include the text "This (video/audio) has been manipulated." For video, the text must be prominent, easily readable by an average viewer, and on screen for the full duration. For audio, it must be clearly spoken at the beginning and end, and for audio longer than two minutes, repeated at intervals of no more than two minutes. The statute contains no satire or parody exemption; the disclosure is the route for anything that would otherwise qualify. The section shows no amendments since 2024 as of October 2026, and we found no reported Idaho decision applying it.

For an advocacy page the operational rule is simple. If you alter audio or video of a candidate in any way that changes what they appear to have said or done, during the 30 or 60 day window, the exact disclosure goes on it, in the exact placement, or the content does not go out. Images and text are outside the statute's definition, but the platform rules below reach them. And every paid electioneering communication still carries the "paid for by" identification the course on literature, signs and ads covers under Idaho Code 67-6614A, which requires that the person responsible for express advocacy "be clearly indicated on such communication."

Law: the FTC Endorsement Guides and partner posts

The FTC's Guides Concerning the Use of Endorsements and Testimonials, 16 CFR Part 255, were revised on July 26, 2023. They define an endorsement as any advertising, marketing or promotional message that consumers are likely to believe reflects the opinions or experiences of a party other than the sponsor, and they say that tags, likes and the use of a person's or organization's name can count. Section 255.5 requires disclosure of a material connection between an endorser and the sponsor that might affect the weight or credibility of the endorsement and that the audience would not expect: payment, free or discounted products, employment, and business, family or personal relationships. The disclosure must be clear and conspicuous, which section 255.0(f) defines as "difficult to miss (i.e., easily noticeable) and easily understandable by ordinary consumers," and in social media "the disclosure should be unavoidable."

The Guides are written for commercial advertising, and whether the FTC would pursue a nonprofit advocacy campaign is a separate question this lesson does not answer. The practice is still the right one, for two reasons. Paid promotion of an advocacy message by a page that looks independent is exactly the "front group" conduct the PRSA Code of Ethics lists as improper, and it is the first thing an opponent will expose. The FTC's Disclosures 101 for Social Media Influencers gives the workable rules: put the disclosure with the endorsement, not on a profile page or behind a "more" link; do not bury it in hashtags; superimpose it on images long enough to read; put it in the video itself as audio and text; repeat it in live streams; use plain terms like "paid partnership" or "sponsored" and avoid "sp," "spon" and "collab"; and do not assume a platform's built-in disclosure tool is enough on its own. M+R's 2026 benchmarks found 58 percent of participating nonprofits worked with influencers at least once in 2025, 24 percent only paid, so this is now ordinary advocacy practice, and the disclosure norm should travel with it.

Law and policy: manipulated media on the platforms

Each platform has its own rule on altered and synthetic media, and the rules changed substantially in 2024 and 2025. The summaries below are from each platform's published policy as checked in October 2026.

Platform	Rule on altered or AI-generated media	Enforcement	Source
Meta (Facebook, Instagram, Threads)	After an Oversight Board recommendation, Meta stopped removing content solely under its manipulated video policy in July 2024 and moved to labeling. "AI info" labels appear on content detected as fully AI generated or disclosed by the poster; for content merely edited with AI the label moved to the post menu in September 2024. Content rated False or Altered by fact-checkers is ranked lower and overlaid. Removal still applies under other policies, such as voter interference, harassment and incitement	Labels, downranking, removal under other policies	Meta Newsroom, April 2024
X	Prohibits sharing inauthentic media that may deceive or harm: media "significantly and deceptively altered, manipulated, or fabricated" and shared as authentic, and unaltered media shared with false context about source, time, place or people. Applies where it may cause widespread confusion on public issues, affect public safety or cause serious harm. Sharing in non-deceptive ways is the stated exception	Reach restriction, labels, post deletion, locks, suspension	X policy, April 2025 version
YouTube	Creators must disclose realistic AI-generated or altered content: a real person appearing to say or do something they did not, altered footage of a real event or place, a realistic scene that did not occur. Obvious fantasy, animation, color filters, captions and script help need no disclosure. YouTube may label automatically, including C2PA-tagged content	Label on player or in description; repeated failure to disclose can bring removal or Partner Program suspension	YouTube Help
TikTok	Realistic AI or edited media must be clearly labeled (a sticker or caption such as "synthetic," "fake," "altered"). Synthetic likenesses of private adults and young people are barred; public figures' likenesses may not be used for political or commercial endorsements. Misleadingly spliced or edited video is barred. Election misinformation about how to vote, eligibility and results is prohibited regardless of intent, and paid political promotion is not allowed at all	Removal; ineligibility for the For You feed	TikTok Community Guidelines
Bluesky	Prohibits content intended to misrepresent, defraud or deceive, false information that could cause immediate harm, coordinated inauthentic campaigns, false voting information, and sexual content using a real likeness; parody and satire accounts must say so in display name and bio	Labels and removal under the September 19, 2025 guidelines	Bluesky Community Guidelines

The common denominator is disclosure. On every platform, an obviously labeled edit or an AI image that announces itself is treated differently from the same content passed off as real. Idaho's statute reaches the same conclusion for candidates in the election window. A network policy that says every AI-generated or altered image, clip or voice carries a visible label, every time, satisfies all five platforms and the state, and it costs nothing but a caption.

Law: copyright and fair use

Advocacy pages share screenshots of news articles, clips of council meetings, charts from reports and photos found online. Most of that is somebody's copyrighted work. 17 U.S.C. 107 lets some unlicensed uses stand as fair use, for purposes including "criticism, comment, news reporting, teaching ... scholarship, or research," weighed on four factors: the purpose and character of the use, including whether it is commercial; the nature of the copyrighted work; the amount and substantiality of the portion used; and the effect on the market for the original. The U.S. Copyright Office stresses that fair use is decided case by case and that there is no formula, no fixed number of words, seconds or pages that is automatically safe. Its Fair Use Index collects the court decisions by category.

The practical pattern that fits the factors: use the smallest piece that makes your point, add your own commentary so the use is transformative rather than a substitute for the original, prefer factual works (a government report) over creative ones (a photographer's portfolio), and never republish a whole article or a full video. Government works produced by federal employees are generally free of copyright; works of Idaho state and local governments are a separate question, and recordings of public meetings posted by a city are usually fine to clip with attribution. When a platform receives a copyright notice under 17 U.S.C. 512, it takes the material down and tells you; you may file a counter notification stating under penalty of perjury that the removal was a mistake, after which the platform restores the material "not less than 10, nor more than 14, business days" later unless the claimant goes to court. Section 512(f) makes anyone who knowingly misrepresents either a claim or a counter claim liable for damages. Know the process before you need it, and keep the notes on where every image came from.

Law: Section 230 in two paragraphs

47 U.S.C. 230(c)(1) says that no provider or user of an interactive computer service "shall be treated as the publisher or speaker" of information provided by another information content provider. For a page, that cuts two ways. The platform is not liable for what you post, so the platform's rules, not the law, decide what stays up. And you, as the page, are generally not liable for what commenters post on your page, because you did not provide that content. Section 230(c)(2) protects good-faith moderation, so deleting or hiding comments does not make you the publisher of the ones you keep.

The protection does not cover your own posts, which you did author, and the exceptions in 230(e) leave federal criminal law, intellectual property and sex trafficking claims untouched. Idaho's synthetic media statute borrows the same vocabulary, naming the "information content provider" as the party who can be sued and treating the interactive computer service as not the publisher unless it alters the content or removes the disclosure. The division of responsibility is consistent: the creator answers for the creation.

Private information, doxxing and harassment

Advocacy fights get personal, and a page that posts a developer's home address or a commissioner's cell number has crossed a line every platform draws. X's private information policy prohibits sharing home addresses and location information, government ID numbers, non-public phone numbers and emails, financial details, medical information and the identity of an anonymous user without permission, as well as threatening to expose any of it or offering a bounty for it; it treats names, ages, workplaces, business locations and public records differently, though home addresses can be actioned even when public elsewhere. Bluesky's guidelines bar home addresses, personal contact details, financial, medical and identity documents, real-time location, and unauthorized private messages unless sharing serves the public interest, and they bar persistent targeting, coordinated humiliation campaigns and bad-faith mass reporting. TikTok bars synthetic likenesses of private individuals outright. Meta's policy pages on privacy violations and harassment did not render for verification in this session, so they are not summarized here; their substance is similar and the Transparency Center hosts them.

The distinction every one of these policies draws is between the public conduct of people in public roles and the private lives of everyone. A commissioner's vote, statements, campaign filings and office contact information are fair subjects. Where she lives, what her children do and her personal phone are not. Comment sections are where this breaks down, and Comments, Moderation and Community covers the mechanics of removing doxxing posted by supporters before it becomes your page's problem.

An ethics code for advocacy pages

The PRSA Code of Ethics is the nearest professional standard. Its core values are advocacy, honesty, expertise, independence, loyalty and fairness, and its disclosure provision, resting on the principle that "open communication fosters informed decision making in a democratic society," names front groups and covert grassroots campaigns on behalf of undisclosed interests as improper conduct. An advocacy page is doing public relations for a cause, and the same standard fits. The code below is written to be copied into a page's About section. Edit it, but publish it, because a published standard is the only kind an audience can hold you to.

Who we are. This page is run by [organization], a [status, such as 501(c)(4) or unincorporated association], and funded by [sources in general terms]. The people who post here are [named roles]. We disclose when a post is paid for and who paid.

What we claim. We link the primary document for every factual claim we make about a public body, a budget or a vote. When we are giving an opinion, we say so. We do not say "sources say."

How we correct. When we get a fact wrong we edit the original post, mark the correction, and post the correction in the same format and place as the error, within 24 hours of confirming it.

What we do not alter. We do not edit audio or video of any person in a way that changes what they appear to have said or done. Any image, clip or voice we generate or alter with software carries a visible label saying so, on every platform, every time.

Who we name. We criticize the public conduct of people in public roles. We do not publish home addresses, personal phone numbers, family details or private information about anyone, and we remove it from our comments when others post it.

How we treat opponents. We quote them accurately and in context, link to their full statement, and do not attribute motives we cannot document. We do not run or fund accounts that hide their connection to us.

What we do with your data. We collect email addresses only with your consent, say how often we will write, and honor every unsubscribe.

Where to complain. [Contact]. We answer within [time].

Questions and answers

If persuasion barely works, why post at all?

Because the things that do work run through the page: mobilizing supporters with social accountability (turnout effects of 1.8 to 8.1 points in Gerber, Green and Larimer), setting the local agenda, recruiting people into the long conversations that change attitudes, and informing the undecided on measures and primaries where Kalla and Broockman found real effects. The evidence redirects the page's effort; it does not retire the page.

Our issue is a county levy. Does the "zero effect" finding apply?

No. Kalla and Broockman's zero is for candidate choice in general elections, where party identity anchors voters. Their meta-analysis found meaningful persuasive effects in primary and ballot measure campaigns, nearly all of them competitive. A levy has no party label on the ballot, which is why your documents, arithmetic and personal stories can move people who have not decided.

Is a long phone conversation as good as a door knock?

The durable result in Broockman and Kalla came from in-person canvassing. Later work by the same authors and others has tested phone versions of deep canvassing with encouraging results, but that work is not among the sources verified for this lesson, so treat it as promising rather than established. The mechanism, a non-judgmental exchange that invites perspective taking, is what matters, and a page's role is the same either way: recruit and schedule the callers.

Can we share a clip from a TV news story about our issue?

A short clip with your own commentary, used to comment on or criticize the coverage or the official in it, fits the fair use factors in 17 U.S.C. 107 better than the whole segment reposted without comment. Keep it short, add your analysis, link to the original, and expect that a broadcaster may still send a takedown; you then have the 512(g) counter-notice route and the 10 to 14 business day restoration window if you believe the use was fair.

We want to run a parody video of a commissioner in October. Does 67-6628A apply?

If the video uses altered audio or video of the commissioner so that a reasonable viewer would think they said or did something they did not, and the commissioner is a candidate within the 60-day window before the general election, yes. The statute has no parody exemption. The disclosure "This video has been manipulated," on screen and readable for the full duration, is the affirmative defense. Platform labels are additional, not substitutes.

A volunteer posted a developer's home address in our comments. What now?

Remove it. Every platform policy summarized above treats a home address as private information regardless of whether it appears in a public record, and X's policy says addresses can be actioned even when public elsewhere. Your ethics code should say you remove such posts, and the moderation lesson covers the bins and the receipts to keep.

Do we have to disclose that a board member's spouse runs the partner page that shared our post?

Under 16 CFR 255.5, family relationships are material connections when the audience would not expect them and they could affect the weight given to the endorsement. The FTC's influencer guidance says to disclose family relationships with the sponsor. A one-line note on the partner's post ("my spouse serves on the board of this group") meets the standard and costs nothing.

Is quoting an opponent's post in our rebuttal defamation?

Accurately quoting what someone said is not a false statement about them. The risk is in your characterization of it. "She lied" asserts knowledge of her state of mind; "her post says X, and the budget document says Y" asserts facts you can prove. Write the second kind of sentence.

Hard questions

Isn't this whole course just teaching manipulation with better footnotes?

The persuasion evidence is the answer. Manipulation, in the sense of changing what people believe through tricks, barely works at scale; 59 ad experiments moved favorability 0.05 points on a five-point scale and the heterogeneity people hoped for was not there. What works is mobilizing people who already agree and giving undecided people on measures accurate information and real stories, and this lesson's ethics code commits a page to disclosure, documents, labels on altered media and visible corrections. A page that follows it is doing in public, with its name attached, what a civic group has always done. The techniques in Roozenbeek and colleagues' prebunking videos are the ones this course teaches readers to recognize, not deploy.

Doesn't engagement optimization reward outrage no matter what your code says?

The ranking research in lesson 01 says moral-emotional language and arousal do spread farther, so the pull is real. Two things cut against it. The persuasion literature says the outrage does not convert anyone, so a page optimizing for it is buying reach that does not serve any objective in the table above. And the misinformation research says the audience you want, the one that shares accurately, responds to accuracy prompts and recognizes manipulation techniques once taught. A page can choose the smaller, steadier audience that acts. The 12-month plan in lesson 10 is built on that choice.

Why follow an FTC rule written for product influencers when we are a nonprofit?

Partly because the Guides' own definition of endorsement is broad enough to reach a paid post for a cause, and the FTC has not said advocacy is exempt. Mostly because the disclosure norm is the same one the PRSA code imposes and the same one your opponents will apply to you. An undisclosed paid post, once exposed, discredits the message and the page, and M+R's benchmarks show paid influencer work is now mainstream in advocacy, so the exposure risk is real. Disclosure is cheap and the alternative is expensive.

Doesn't Idaho's synthetic media law chill legitimate satire?

It might, and the absence of a parody exemption is the strongest criticism of 67-6628A. The statute's answer is the disclosure defense: labeled manipulation is protected, unlabeled is not, and only the depicted candidate can sue, with a clear and convincing evidence burden. Whether that is the right balance is a fair argument for the Legislature and eventually the courts. As of October 2026 we found no reported decision applying the statute, so the chilling effect, if any, is untested. A page that labels its edits is not the test case.

Sullivan protects us. Why be so careful about private people?

Because Sullivan protects criticism of public officials and public figures, and Gertz lets Idaho hold you to an ordinary fault standard for everyone else. The neighbor who spoke at the hearing, the small landlord, the mid-level employee at the agency are private people, and a false statement about one of them made carelessly can be sued over for two years under Idaho Code 5-219. The public-conduct rule in the ethics code keeps the page where the law protects it most.

If backfire is rare, why do corrections so often seem to fail?

Because failure and backfire are different. Wood and Porter found corrections did not make beliefs worse; they did not find that one correction erases a belief. Corrections compete with repetition, trusted messengers and motivated reasoning, and they often reach a different audience than the myth did. The Debunking Handbook's advice, lead with the fact, name the myth once, explain the technique, restate the fact, and the practice of correcting in the same venue and format, are about raising the success rate. Nothing in the evidence says to stop.