Meta tags:

Headings (most frequently used words):

share, this, the, logic, of, science, is, follow, anti, vaccers, are, fundamentally, wrong, about, placebo, controlled, trials, understanding, analogies, in, logical, arguments, stop, blaming, china, and, india, for, climate, change, big, pharma, not, buying, favorable, peer, reviews, backfire, effect, real, on, facebook, via, email, archives, tags,

Text of the page (most frequently used words):
the (501), and (228), that (223), are (120), for (103), this (91), not (81), they (71), effect (63), but (59), with (59), have (57), studies (54), there (52), #backfire (51), vaccines (49), was (45), more (43), you (43), other (38), study (37), effects (36), control (36), emissions (36), can (32), people (32), argument (31), were (31), china (31), about (30), only (29), even (29), from (28), when (28), one (28), their (28), autism (27), like (27), than (27), actually (27), all (26), group (26), then (26), paper (26), because (24), science (23), trials (23), rate (23), side (23), 2019 (22), how (22), average (22), has (22), very (22), get (22), case (22), reviewers (22), funding (21), usa (21), evidence (20), results (20), sample (20), being (20), controlled (20), had (20), size (20), peer (19), new (19), misinformation (19), 100 (19), don (19), see (19), correlation (19), review (19), way (19), 2015 (18), 2017 (18), should (18), often (18), really (18), who (18), treatment (18), also (18), would (18), india (18), cohort (18), placebo (18), vaccine (17), research (17), does (17), false (17), real (17), rare (17), these (17), share (16), time (16), here (16), power (16), 2020 (15), fact (15), where (15), those (15), did (15), them (15), will (15), some (15), total (15), background (15), randomized (15), logic (14), change (14), arguments (14), anti (14), may (14), facebook (14), which (14), first (14), any (14), into (14), let (14), many (14), aren (14), world (14), detect (14), interest (13), cause (13), shown (13), each (13), big (13), note (13), need (13), population (13), countries (13), 2016 (12), 2018 (12), political (12), used (12), same (12), again (12), going (12), just (12), point (12), rather (12), per (12), causation (12), safety (11), medicine (11), 2024 (11), 2026 (11), posted (11), information (11), been (11), factors (11), example (11), words (11), look (11), things (11), authors (11), two (11), over (11), however (11), simply (11), still (11), medical (11), outcome (11), causes (11), fallacies (10), global (10), climate (10), 2021 (10), opens (10), window (10), journal (10), its (10), further (10), low (10), specific (10), true (10), most (10), important (10), using (10), lack (10), several (10), doesn (10), received (10), while (10), greenhouse (10), rates (10), reviewed (9), january (9), february (9), july (9), effective (9), nyhan (9), corrections (9), large (9), based (9), data (9), show (9), find (9), what (9), found (9), statement (9), test (9), lot (9), problem (9), different (9), papers (9), conflicts (9), generally (9), payments (9), journals (9), country (9), capita (9), analogies (9), post (8), bad (8), nature (8), june (8), follow (8), why (8), correcting (8), ecker (8), therefore (8), good (8), out (8), statements (8), correlated (8), thus (8), second (8), use (8), between (8), know (8), someone (8), clear (8), pharma (8), industry (8), pharmaceutical (8), trial (8), completely (8), vaccers (8), gas (8), figure (8), say (8), march (7), april (7), august (7), september (7), posts (7), swire (7), taking (7), thing (7), never (7), quality (7), conclusion (7), make (7), understand (7), individuals (7), positive (7), looking (7), whether (7), participants (7), groups (7), topics (7), topic (7), scientists (7), causing (7), fallacy (7), much (7), anything (7), companies (7), mean (7), funded (7), company (7), before (7), researcher (7), billion (7), such (7), claim (7), america (7), above (7), under (7), below (7), considerations (7), common (7), comic (7), relationship (7), account (6), blog (6), logical (6), 2022 (6), comments (6), role (6), 2013 (6), reason (6), well (6), failed (6), views (6), reliability (6), related (6), level (6), few (6), approach (6), significant (6), finally (6), after (6), though (6), responses (6), negative (6), indeed (6), design (6), once (6), larger (6), occurs (6), talking (6), against (6), wrong (6), issue (6), man (6), reviews (6), work (6), without (6), reviewer (6), own (6), works (6), proves (6), blame (6), increasing (6), take (6), our (6), premise (6), 2004 (6), powered (6), third (6), now (5), rules (5), evaluating (5), october (5), december (5), 2025 (5), tagged (5), loading (5), reasoning (5), weeks (5), thompson (5), checks (5), testing (5), exposure (5), facts (5), improvement (5), questions (5), key (5), update (5), trying (5), saying (5), researchers (5), scores (5), start (5), able (5), consistently (5), less (5), factor (5), correct (5), tests (5), single (5), extremely (5), huge (5), times (5), tested (5), presented (5), claims (5), buying (5), receives (5), potential (5), paid (5), grant (5), payment (5), pfizer (5), etc (5), numbers (5), totally (5), another (5), investing (5), renewable (5), energy (5), details (5), 1970 (5), compared (5), irrelevant (5), doing (5), want (5), strawman (5), structure (5), useful (5), right (5), children (5), simulation (5), designs (5), powerful (5), sizes (5), within (5), wordpress (4), com (4), alternative (4), november (4), email (4), wood (4), porter (4), garrett (4), 2014 (4), checking (4), three (4), short (4), showed (4), subgroups (4), accuracy (4), positives (4), said (4), interesting (4), measure (4), later (4), great (4), consistent (4), serious (4), lower (4), better (4), included (4), number (4), directly (4), original (4), result (4), chance (4), always (4), apply (4), years (4), sometimes (4), around (4), best (4), weak (4), discussion (4), made (4), favorable (4), else (4), paying (4), down (4), cases (4), previously (4), usually (4), goes (4), consider (4), percent (4), something (4), smaller (4), far (4), particularly (4), drug (4), relevant (4), long (4), mind (4), outsized (4), obviously (4), contributed (4), stated (4), produced (4), since (4), meanwhile (4), line (4), hand (4), both (4), together (4), illustrate (4), analogy (4), effectiveness (4), placebos (4), measles (4), mumps (4), rubella (4), vaccination (4), covid (4), destefano (4), association (4), cannot (4), thousands (4), 000 (4), high (4), sign (3), already (3), consensus (3), hoc (3), gmos (3), evolution (3), appeal (3), address (3), your (3), receive (3), older (3), behavior (3), during (3), misperceptions (3), source (3), seriously (3), beliefs (3), fail (3), reifler (3), 2010 (3), guess (3), corrective (3), media (3), spread (3), knowledge (3), literature (3), ever (3), reporting (3), conducted (3), opposite (3), reported (3), likely (3), individual (3), currently (3), become (3), convinced (3), dig (3), belief (3), necessarily (3), give (3), subsequent (3), given (3), typically (3), described (3), truth (3), consistency (3), massive (3), updated (3), statistically (3), ran (3), majority (3), slightly (3), statistical (3), produce (3), largest (3), response (3), makes (3), document (3), corrected (3), highly (3), looked (3), previous (3), examples (3), publication (3), arguing (3), comes (3), widespread (3), anytime (3), general (3), misleading (3), pretty (3), current (3), analysis (3), nothing (3), dangerous (3), wildly (3), unrelated (3), substantially (3), small (3), back (3), institution (3), scientist (3), million (3), university (3), takes (3), reasons (3), top (3), speaking (3), simple (3), england (3), admittedly (3), resulting (3), basic (3), points (3), deniers (3), claiming (3), investment (3), keeping (3), dishonest (3), americans (3), picture (3), lags (3), behind (3), united (3), european (3), combined (3), increase (3), biggest (3), focus (3), sense (3), decrease (3), action (3), terms (3), play (3), insist (3), comment (3), straw (3), underlying (3), discussed (3), train (3), yourself (3), avoid (3), seen (3), hundred (3), making (3), understanding (3), seat (3), belts (3), needed (3), love (3), associated (3), causal (3), indicates (3), confounding (3), thousand (3), randomization (3), hard (3), technically (3), confidently (3), possible (3), concern (3), left (3), row (3), axis (3), 500 (3), patients (3), ice (3), cream (3), drowning (3), variables (3), view (2), report (2), content (2), log (2), subscribed (2), subscribe (2), join (2), subscribers (2), scientific (2), cherry (2), via (2), elusive (2), factual (2), 163 (2), rumors (2), candidate (2), public (2), moderate (2), partisan (2), bias (2), communication (2), liar (2), phenomenon (2), royal (2), society (2), schmid (2), betsch (2), literally (2), persistence (2), haglin (2), coppock (2), replicate (2), familiarity (2), 2023 (2), cognitive (2), applied (2), patient (2), cameron (2), cited (2), harm (2), extent (2), generalizable (2), peoples (2), portion (2), asked (2), backfires (2), measurement (2), moving (2), promising (2), seems (2), field (2), shifting (2), moon (2), landing (2), fake (2), showing (2), net (2), backfired (2), caveat (2), person (2), contrary (2), willing (2), accurate (2), emerging (2), quote (2), worth (2), discussing (2), closer (2), patterns (2), suggested (2), seeing (2), earlier (2), come (2), rated (2), exactly (2), reliable (2), repeatable (2), stronger (2), error (2), variation (2), explained (2), broader (2), across (2), elicit (2), took (2), recommend (2), reading (2), procedure (2), randomly (2), except (2), twice (2), suggests (2), documenting (2), outcomes (2), tend (2), mentioned (2), trend (2), yet (2), ways (2), read (2), spend (2), biases (2), instead (2), critical (2), reality (2), wrote (2), article (2), reject (2), pro (2), pages (2), difficult (2), state (2), looks (2), misunderstand (2), crude (2), absolutely (2), support (2), documented (2), biased (2), volunteer (2), beyond (2), existing (2), written (2), reviewing (2), getting (2), additional (2), provided (2), brief (2), vast (2), gets (2), dollar (2), expensive (2), little (2), money (2), separate (2), counted (2), place (2), broad (2), four (2), major (2), determined (2), lancet (2), jama (2), nguyen (2), countless (2), situations (2), initial (2), development (2), clinical (2), additionally (2), process (2), input (2), voluntary (2), corruption (2), knowing (2), entirely (2), editor (2), bribe (2), suggest (2), quickly (2), according (2), commenters (2), page (2), through (2), history (2), warming (2), building (2), coal (2), plants (2), representing (2), demand (2), pretend (2), think (2), fossil (2), fuel (2), heavily (2), contribute (2), last (2), production (2), put (2), states (2), unit (2), reference (2), contributing (2), expected (2), isn (2), environmental (2), trends (2), including (2), obvious (2), minute (2), produces (2), gases (2), air (2), issues (2), responsibility (2), actions (2), too (2), excuse (2), fairly (2), substantial (2), others (2), examining (2), ago (2), stop (2), men (2), care (2), try (2), matter (2), addressed (2), carefully (2), trap (2), exact (2), popular (2), guarantee (2), won (2), break (2), syllogism (2), injury (2), particular (2), frequently (2), jump (2), certainly (2), aspect (2), equivalency (2), unstated (2), purpose (2), occur (2), namely (2), stick (2), demonstrate (2), thimerosal (2), risk (2), uno (2), meta (2), taylor (2), mmr (2), developmental (2), disorders (2), smeeth (2), polack (2), madsen (2), 2002 (2), jain (2), hviid (2), age (2), subjects (2), pediatrics (2), anders (2), equal (2), medicines (2), means (2), simplicity (2), developing (2), runs (2), causally (2), inherently (2), fundamental (2), tool (2), definition (2), fundamentally (2), prove (2), truly (2), incredibly (2), hundreds (2), contrast (2), assert (2), possibility (2), simulated (2), models (2), panel (2), increases (2), column (2), decreases (2), log10 (2), scale (2), tick (2), randomize (2), caused (2), quite (2), records (2), conduct (2), end (2), types (2), versions (2), collapse, bar, manage, subscriptions, site, reader, privacy, 253, statistics, ergo, propter, natural, selection, homeopathy, blind, creationism, conflict, picking, cancer, burden, proof, ignorance, emotion, authority, anecdotal, tags, archives, 181, enter, notifications, search, uncategorized, mass, attitudes, steadfast, adherence, 135, electoral, consequences, motivated, vote, choice, 2008, presidential, election, international, opinion, 401, 422, emotions, partisanship, anger, anxiety, susceptibility, 699, 719, might, evaluation, prevalence, psychology, processing, comprehending, trump, open, discussions, human, behaviour, strategies, rebutting, denialism, journalistic, favorability, explain, durability, pnas, 118, e1912440117, 303, 330, polit, behav, limitations, politics, counter, attitudinal, backlash, survey, experiments, british, undermining, contextual, cues, naïve, theory, 617, 637, failure, fear, driven, plos, e0281140, worldview, convenience, philosophical, transactions, 376, 20200145, audiences, principles, implications, reminders, repetition, helping, hindering, retraction, memory, cognition, 185, 192, recall, health, following, myths, message, format, variations, education, counseling, 381, 387, leave, appear, replicated, numerous, weight, shows, due, probably, traits, understood, errors, encountering, away, hypothetical, imagine, believe, check, remain, unchanged, accept, heals, situation, score, whole, benefit, handful, improve, reflection, coverage, finds, revise, mistaken, generates, modest, improvements, originally, explicitly, pointing, presenting, step, whose, correction, backfiring, checked, strong, strongly, suggesting, repeatability, measures, provides, plausible, bunch, cover, neither, elegant, retest, split, incorrect, retested, occurring, underwent, ensure, 388, 532, basically, refers, pointed, agreement, unreliable, lead, relied, opportunity, raises, talk, vary, significance, disappear, least, partially, dubious, reproducible, replication, comprehensive, five, trails, designed, covered, resulted, answers, although, initially, thought, importance, towards, perfectly, essentially, pick, inconsistent, reasonable, degree, published, reproducibility, practice, happen, bringing, act, inevitable, certain, tried, set, record, straight, bemoaned, outlets, mischaracterized, treated, dominant, psychological, reinforces, brought, minds, commentary, ironically, mixed, wide, debunked, add, wild, internet, taken, summary, biasing, duh, shooting, supported, schism, coming, conclusions, existence, assessed, subset, overlap, idea, anonymous, favor, unlikely, exist, distinguished, non, similarly, moderna, itself, arises, entity, lots, competing, examine, regardless, aside, sounds, bucks, bank, chunk, overhead, rest, equipment, assistants, depending, responsible, salary, please, remember, framing, problems, framed, titles, grab, attention, title, crystal, engagement, volunteered, likewise, voluntarily, methods, obtained, lists, names, served, form, dollars, bmj, painting, brush, arrangements, working, phases, stages, especially, market, funders, direct, output, layer, removal, agencies, agency, publications, ends, approving, proposed, sending, submits, puts, shaky, ground, volunteers, ahead, told, invite, suggestions, followed, devastating, whom, active, expertise, studied, staff, submitted, reaches, experts, asks, free, access, limited, period, concepts, detailed, hit, nutshell, bogus, hopelessly, corrupt, close, revealing, literacy, system, absurd, face, blindly, assuming, must, corrupted, whatever, constantly, lately, strategy, emerged, shouldn, trust, gross, misreading, titled, device, manufacturers, assume, leaders, electricity, growth, met, renewables, shining, contributor, steps, curb, seem, fighting, everyone, actual, paint, continues, disproportionate, amount, fingers, invested, 625, year, american, tell, undeniable, playing, relative, silly, products, shipped, overseas, satiate, rampant, consumerism, entire, planet, paints, story, fair, egregious, offender, fortunately, declining, sits, declined, barely, unfortunately, recently, risen, ignoring, identical, laws, regulations, expect, policies, math, totals, complete, levels, contribution, sum, leads, remotely, polluter, minuscule, comparison, emitters, union, dramatic, having, impact, full, odds, peak, 2000s, higher, whereas, specifically, plastic, pollution, cities, smog, crippa, guizzardi, pagani, banja, muntean, schaaf, quadrelli, risquez, martin, taghavi, moharamli, grassi, rossi, melo, oom, branco, suarez, moreno, sedano, san, miguel, manca, pisoni, pekar, ghg, jrc, iea, luxembourg, https, europa, doi, 2760, 9816914, none, absolve, duty, valid, unethical, recent, fewer, worse, outline, core, vilify, plenty, plays, disingenuous, solve, proportion, encounter, various, forms, includes, tiny, utterly, falls, apart, copout, shifted, blaming, attack, reductio, absurdum, part, compare, apples, oranges, personal, holding, position, pesticides, corporate, food, personally, fold, intended, defects, every, respect, mental, jumping, accuse, committing, occasions, called, list, count, bluntly, hear, vaccinated, infected, wearing, helmet, head, debunking, brains, resistant, tricks, ignore, shift, happened, defect, vaccer, ignored, jumped, aware, ourselves, tendency, deep, breath, front, brain, force, thoroughly, analyze, claimed, encompasses, vaccinate, analogous, clearly, replacing, despite, premises, horribly, outside, matters, injuries, evaluate, yesterday, lightly, fun, past, flaws, apparently, filled, equivalencies, unlike, helmets, birth, bags, inside, protected, lawsuits, early, containing, spectrum, disorder, 2511, 2516, elsevier, 3623, 3629, pervasive, 364, 963, 969, efficacy, bnt162b2, mrna, 383, 2603, 2615, 347, 1477, 1482, occurrence, status, among, siblings, 313, 1534, 1540, nationwide, annals, internal, antibody, stimulating, proteins, polysaccharides, ped, 561, 567, school, matched, metropolitan, atlanta, 113, 259, 266, infants, retrospective, kingdom, 114, 584, 591, litterateur, thorough, risks, foods, toxic, chemicals, hierarchy, robust, masks, successes, ivermectin, hydroxychloroquine, modeling, assumed, half, model, stochastic, meaning, accounts, waviness, stochastically, additive, strictest, failing, lag, correctly, nevertheless, remains, applies, allow, assign, grossly, tools, detecting, associations, assigning, indicate, safe, misunderstanding, demonstrated, notion, epidemic, falsified, almost, anyone, uber, impressive, bring, satisfactory, establishing, hopefully, faulty, badly, agree, eliminate, applying, abuse, word, games, demanding, impossible, request, sec, keep, unknown, conclusively, certainty, judge, concerned, undetectable, legitimate, greatly, reduced, occurred, problematic, interpret, weren, interpreting, pause, stress, simplistic, scenario, sex, built, pronounced, differences, bottom, rarer, harder, comparing, text, blue, red, realize, graph, bit, daunting, walk, columns, decreasing, rows, addition, iterations, generating, populations, analyzed, returned, studying, samples, cheap, easy, tens, strength, affect, determine, influences, signal, drawback, dictating, limitation, ethical, consuming, costly, recruit, program, range, dozen, exceptions, mentioning, experimental, regression, measuring, improves, symptoms, infectious, disease, scientifically, confident, driving, dividing, drive, missed, confidence, famous, sales, positively, temperatures, buy, swimming, accidents, could, outset, briefly, discuss, explanation, simplified, sake, brevity, backwards, identify, match, similar, demographically, controls, next, manipulative, randomizing, compares, classic, aka, rct, randomizes, recorded, continuous, regressions, discrete, throughout, thesis, automatically, limits, insistence, talked, visual, vaxxers, suffice, demonstrating, flawed, novel, inert, saline, updates, known, profiles, alt, misc, home, skip,

Text of the page (random words):
utical companies but the study did not examine who was funding which study in other words this paper simply documented any reviewers that had received any industry funding regardless of whether the specific paper being reviewed was funded by the same company that provided the other funding as i have written about before e g here and here there are many studies that have no conflicts of interest and aren t funded by pharmaceutical companies but this paper in no way distinguished those studies so if a researcher received a grant from a pharmaceutical company then reviewed a paper that did not receive funding from a pharmaceutical company the payment to the researcher was still included even though they were reviewing a non industry paper similarly a researcher who received research funding from moderna then reviewed a paper that was funded by pfizer was still included as an industry payment so even beyond the fact that these payments were not actually for peer review the level of existing conflicts of interest is not at all clear the potential conflicts would only exist when a reviewer had received funding from the same company that had funded the paper but those data aren t presented and by the time you consider all of the papers without and conflicts of interest and all of the different pharmaceutical companies out there that number is going to be a very small portion of that 54 in other words the percent of reviewers who reviewed a paper that was funded by the same company that had previously paid them for something is going to by much much smaller than 54 note as well that this paper only documented the existence of unrelated payments to reviewers it in no way assessed whether those payments biased reviewers and even in the subset of cases where a reviewer s funding and a study s funding overlap the idea that an anonymous volunteer peer reviewer is going to be substantially biased in favor of a study because it was funded by a company that had previously given them funding is unlikely on its own that s just not how review usually goes finally even if nothing else that i had said was true and pharmaceutical companies really were directly paying for favorable reviews they aren t that would still only apply to 54 of reviewers based on this study so if for example vaccines actually were dangerous that would mean that the other 46 of reviewers should be shooting down pro vaccine papers the same is true more generally for conflicts of interest if scientists only supported vaccines because of funding from big pharma then there should be a clear schism with papers with and without industry funding coming to wildly different conclusions and that s just not the case in summary this paper did not show that big pharma is buying favorable reviews and it did not show that industry funding is biasing review results it simply showed that reviewers often have work that receives industry funding no duh don t misunderstand me conflicts of interest can bias authors and should be taken seriously but the crude level of analysis used in this paper doesn t really add much of anything to the discussion and absolutely does not support the wild claims being made by the good people on the internet share this share on x opens in new window x share on facebook opens in new window facebook like loading posted in nature of science vaccines alternative medicine tagged anti vaccine arguments bad arguments conflicts of interest debunked evaluating evidence fact checking peer reviewed studies 2 comments is the backfire effect real posted on february 17 2026 by fallacy man the backfire effect is a psychological phenomenon in which correcting misinformation actually reinforces the false view rather than causing someone to reject it nyhan and reifler 2010 this topic comes a lot in the comments on this and other pro science pages sometimes it is presented as an argument that pages like this one actually do more harm than good and other times it is brought up as a general point about how difficult it is to change people s minds most of the commentary i see around this topic is however misleading at best ironically many of the statements i see about the backfire effect are actually misinformation the evidence for the backfire effect is actually pretty weak and mixed and it may not even be a real effect at all if it is a real effect it is pretty clear at this point that it is not a wide spread generalizable effect that occurs anytime misinformation is corrected so in this post let s dig into the literature and see what the current state of evidence actually looks like backfire effects are not widespread the first critical point that many get wrong is that the data have never suggested that backfire effects are a widespread issue that happen anytime misinformation is corrected a lot of times when i see people bringing up this topic they act as if backfire effects are inevitable anytime you correct misinformation on any topic for any group of people in reality the studies documenting backfire effects have always found that they only apply to subgroups and certain topics indeed the first study on backfire effects nyhan and reifler 2010 only found them in 2 of the 5 topics they tested and several years later one of the original authors wrote an article nyhan 2021 in which they tried to set the record straight and bemoaned the fact that many media outlets had mischaracterized backfire effects and treated them as the dominant reason for the persistence of political misinformation to quote one of the original authors nyhan 2021 subsequent research suggests that backfire effects are extremely rare in practice backfire effects have low reproducibility if you read this blog often you know that i spend a lot of time talking about the importance of looking for consistent evidence there are publication biases towards positive results a lot of low quality papers do get published and even for a perfectly conducted study you may get a false positive just by chance as a result on essentially any topic you can find papers arguing for it and papers arguing against it therefore you should not cherry pick studies and should instead look for a consensus of evidence if an effect is real studies should be able to document it with a reasonable degree of consistency when it comes to backfire effects the results are highly inconsistent there is a really interesting review on backfire effects that i highly recommend reading swire thompson et al 2020 they found that although this field looked really promising at first a huge number of studies have failed to find evidence of the backfire effect even when trying to directly replicate previous studies for examples see cameron et al 2013 garrett et al 2013 weeks garrett 2014 weeks 2015 ecker et al 2017 haglin 2017 swire et al 2017 guess coppock 2018 nyhan et al 2019 schmid betsch 2019 swire thompson et al 2019 wood porter 2019 ecker et al 2020 ecker et al 2021 ecker et al 2023 this lack of consistency occurs even in the subgroups topics where researchers initially thought there were real backfire effects also note that the authors of the original paper on backfire effects are authors on several of these papers that failed to find backfire effects in one of the largest and most comprehensive studies of this topic scientists conducted five trails designed to approach the problem from slightly different ways the trials covered a total 52 statements corrections and included a total of 10 100 participants wood porter 2019 they were not able to elicit a backfire effect for a single one of those 52 statements corrections indeed as with many of these studies they found that correcting the misinformation resulted in an average improvement in scores in other words people updated their knowledge with the new information and on average their answers were more correct after the misinformation had been corrected the opposite of a backfire effect all of this makes me extremely dubious about the backfire effect if the effect is real it should be reproducible yet over and over again the largest studies and replication studies fail to document backfire effects so let s look closer at some of the factors that are going on here study quality one of the other things i talk about a lot is that study quality can vary slightly and very often for topics where there is no effect there is a negative correlation between study quality and the statistical significance of the outcomes in other words lower quality studies tend to produce positive results but those effects disappear with larger better controlled studies this at least partially seems to be the case for backfire effects as mentioned earlier some of the largest studies e g wood porter 2019 failed to find a backfire effect but there is also an interesting trend when we look at response reliability basically reliability refers to whether studies are using questions that will give consistent results in the swire thompson et al 2020 review they pointed out that using only a single measure of agreement with a statement is often unreliable and can lead to false positives and 81 of reported backfire effects relied on only a single measure this suggests a large opportunity for false positives and raises serious questions about the majority of studies documenting backfire effects swire thompson et al 2022 took this further with a really elegant study that i recommend reading in short they used a test retest procedure where they randomly split participants into treatment and control groups the treatment groups where shown incorrect statements and asked to rate the truth of the statement then they were shown corrections to the statement then they were retested three weeks later thus if the backfire effect was occurring they should rate the false statements as more true on average on the second test the control group underwent the same procedure except they were not shown the corrections this is a great design because it let the authors use the control group to look at how reliable responses were across the 21 statements they used they also ran the study twice to ensure results were repeatable once with 388 participants and once with 532 participants there are a bunch of interesting results to this study only some of which i have time to cover here first on average the treatment group that was shown corrections updated their views more than the control group that was not shown corrections in other words again the average effect was that people updated their views when shown fact checks the opposite of a backfire effect second in the treatment group only 2 of the 21 statements showed any sign of a backfire effect i e on average people rated the false statement as more true after being shown the correct information but neither was statistically significant in other words out of 21 tests across two trials they did not elicit a single statistically significant backfire effect taking this a step further things get really interesting when we start looking closer at the patterns for the individuals who backfired in other words even though on average there was no backfire effect there were some individuals whose scores suggested that they were more convinced of the false information after seeing the correction this is where those reliability scores from earlier come in using the control group the authors were able to measure the reliability of responses for each of the 21 statements then look at how those reliability scores correlated with the level of backfiring in the treatment group in other words they used the control group which was not shown fact checks to see how consistently people rated the truth of each false statement then they checked whether questions with a low consistency of responses in control group were more likely to have backfires in the treatment group that is in fact exactly what they found there was a strong negative correlation in other words the less reliable repeatable the responses to a statement were the stronger the backfire effect was thus strongly suggesting that measurement error is a big factor in studies reporting backfire effects indeed in the two trials in this study 37 and 53 of the variation in backfire effects was explained by the repeatability measures that s massive and provides a very plausible source of false positives in the broader literature the average effect of correcting misinformation is positive finally it is worth explicitly pointing out that most of the studies i ve been discussing have not only failed to find a backfire effect but they have actually found a net improvement in scores after correcting the misinformation in other words the average effect of presenting fact checks is an improvement in peoples knowledge to again quote one of the researchers who originally described the backfire effect nyhan 2021 there is an emerging consensus that exposure to corrective information typically generates modest but significant improvements in belief accuracy and contrary to media coverage of the backfire effect subsequent research finds that people are often willing to revise mistaken beliefs when given accurate information with that said there is an important caveat saying that the average effect is an improvement in belief accuracy is not the same thing as saying the average person will improve their belief accuracy in other words these studies look at group effects which don t necessarily give a good reflection of what an average individual does as a hypothetical example imagine we have a group of 100 people who believe the moon landing was fake then we show them a fact check and 70 people remain unchanged 20 update their views and accept the moon landing was real and 10 dig in their heals and become more convinced it was fake in that situation the average score for the group as a whole will go up showing that correcting the misinformation has a net benefit but most individuals 70 did not update their views and a handful backfired 10 so the key questions become why did so few update their views and where those 10 backfires measurement errors or did encountering corrective information really make them more convinced of the false information a true backfire moving away from these group level effects and trying to understand factors for individuals is a promising approach and seems to be where the field is likely shifting conclusion so where does all of this leave the backfire effect in short to a large extent it does not appear to be a real thing it has never been generalizable and it has only ever showed up for subgroups but even for those subgroups studies reporting backfire effects often can t be replicated numerous large well conducted studies have failed to find backfire effects and the weight of evidence shows that the average effect of correcting misinformation is an improvement in the accuracy of peoples views the opposite of a backfire effect further studies reportin...