Voir la source de Nonprobative photos inflate the truthiness and falsiness of claims

Newman, E. J. (2013). Nonprobative Photos Inflate the Truthiness and Falsiness of Claims (Version 1). Open Access Te Herenga Waka-Victoria University of Wellington. https://doi.org/10.26686/wgtn.17003764.v1


== Abstract de l'autrice == 
''When people evaluate claims they often rely on what comedian Stephen Colbert calls truthiness, judging claims using subjective feelings of truth, rather than drawing on facts. Over seven experiments I examined how nonprobative photos can manufacture truthiness in just a few seconds. I found that a quick exposure to a photo that relates to, but does not provide any probative evidence about the accuracy of claims can systematically bias people to conclude claims are true. In Experiments 1A and 1B, people saw familiar and unfamiliar celebrity names and, for each, quickly responded "true" or "false" to the claim "This famous person is alive" or (between subjects) "This famous person is dead." Within subjects, some names appeared with a photo of the celebrity engaged in his/her profession whereas other names appeared alone. For unfamiliar celebrity names, photos increased the likelihood that subjects judged the claim to be true. Moreover, the same photos inflated the truth of "Alive" and "Dead" claims, suggesting that photos did not produce an "alive bias," but a "truth bias."''

''Experiment 2 showed that photos and verbal information similarly inflated truthiness, suggesting that the effect is not peculiar to photographs per se. Experiment 3 demonstrated that nonprobative photos can also enhance the truthiness of general knowledge claims (Giraffes are the only mammals that cannot jump). In Experiments 4-6 I examined boundary conditions for truthiness. I found that the semantic relationship between the photo and claim mattered. Experiment 4 showed that in a within-subject design, related photos produced truthiness, but unrelated photos acted just like the no photo condition. But unrelated photos were not always benign, Experiment 5 showed that their effects depended on experimental context. In a mixed design, related photos produced truthiness and unrelated photos produced falsiness.'' 

''Although the effect of related photos was robust across materials and variation in experimental context, when I used a fully between-subjects design in Experiment 6, the effect of photos (related and unrelated) was eliminated. These effects add to a growing literature on how nonprobative information can influence people’s decisions and suggest that nonprobative photographs do more than simply decorate, they can rapidly manufacture feelings of truth. As with many effects in the cognitive psychology literature, the photo-truthiness effect depends on the way in which people process and interpret photos when evaluating the truth of claims.''

== Chapter 1 ==
''In this thesis I use the term “truthiness” to refer to a category of phenomena in which the addition of nonprobative information causes people to shift rapidly towards believing a claim (see also the illusory truth effect; Begg, Anas, & Farinacci, 1992).''

=== Judging Truth ===
=== Repetition and Truth ===
=== Conceptual Fluency and Cognitive Availability. ===
=== Imagery and Reality Monitoring. ===
=== Trawling for Confirming Evidence. ===
=== Overview === 
''If photos help boost cognitive availability of related thoughts and images, or help people trawl for and selectively interpret information in line with their hypothesis about unfamiliar celebrities, then photos should increase the truthiness of claims about those celebrities, regardless of whether the claim is that the celebrity is alive or dead (cf. Unkelbach, 2007).''

== Chapter 2 - Experiment 1A & 1B ==
=== Experiment 1A and 1B === 
==== Design ====   
* VI1: photograph: yes, no
* VI2: familiarity: familiar, unfamiliar (i.e. est-ce que la personne est "connue" ou non) 
* VI3: claim: alive, dead 
* VD: C score (déviation à dire "true") 
Résultant en un 2x2x2 mixed design, avec photograph et familiarity within subjects et le claim between subjects.

==== Résultats ====  
* Elle a réalisé une mixed repeated measures (photograph) x 2 (familiarity) x 2 (claim) ANOVA 
* Effet de la photo: 
** Experiment 1A, F(1, 90) = 4.87, p = .03, η<sub>p</sub>² = .05; 
** Experiment 1B, F(1, 46) = 10.53, p < .01, η<sub>p</sub><sup>2</sup> = .19)
[[Fichier:Resultats exp 1a 1b (Newman 2013) .png|600px|vignette|centré|alt=Resultats exp 1a 1b (Newman 2013) |Bias for claims about familiar and unfamiliar names (in Experiments 1a-1b) presented with or without a photograph and collapsed across dead/alive factor. Negative value of c is a bias to say true. In Experiments 1a-2, photos (or words) affected bias for unfamiliar names. Error bars show 95% within-subject confidence intervals for the photo/no-photo effect at each level of familiarity/difficulty (see Masson & Loftus, 2003)]]

==== Discussion ==== 
Newman remarque que les gens ont tendance à trouver vraies les ''"Alive" claims'' plus souvent que les ''"Dead" claims''. 

A ce moment, Newman se demande si c'est un biais de ''truthiness'' ou d'''aliveness'' (i.e. le fait qu'on voit la photo fait penser que la personne est en vie). 

''Photos might promote truthiness because they provide a rich semantic context that facilitates the generation of thoughts and images relating to the claim. But these results also fit with the idea that feelings of truthiness arose because photos are inherently credible or that over time people have come to associate photos with the conclusion that something is real and true (cf. Unkelbach, 2007). ''

== Chapter 3 - Experiment 2 and 3 ==
=== Experiment 2 ===
Pour l'expérience 2, elle se demande si des types d'informations non probantes autres que les photos vont avoir un impact sur la ''truthiness''. Elle a donc cherché à comparer les types de formats (photo et verbal). 

==== Design ====
Variables: 
*VI 1: nonprobative information: yes, no 
*VI 2: format of nonprobative information: photo, verbal
*VI 3: claim: alive, dead
*VD : C score 

Elle a abouti à un design mixte 2 (nonprobative information) x 2 (format of nonprobative information) x 2 (claim) avec le format et le claim comme élément intrasujet. Elle a aussi gardé uniquement les unfamiliar names.

==== Résultats ====
Elle a fait une mixed repeated measures 2 (nonprobative information: yes, no) x 2 (format of nonprobative information: photo, verbal) x 2 (claim: alive, dead) ANOVA. Elle a trouvé un '''effet de la présence de ''nonprobative information''''': F(1, 50) = 10.27, p < .01, η<sub>p</sub><sup>2</sup> = .17.

[[Fichier:Experiments 2 3 - Newman 2013.png|600px|vignette|centré|alt=Experiments 2 3 - Newman 2013|Bias for claims about unfamiliar names (Experiment 2) presented with or without a photograph or verbal information, collapsed across dead/alive factor; easy and difficult trivia claims (in Experiment 3) presented with or without a photograph . Negative value of c is a bias to say true. In Experiment 2, photos (or words) affected bias for unfamiliar names; in Experiment 3, photos affected bias for difficult trivia statements. Error bars show 95% within-subject confidence intervals for the photo/no-photo effect at each level of familiarity/difficulty (see Masson & Loftus, 2003).]]

==== Discussion ==== 
Elle a donc trouvé un effet de la nonprobative information, que ce soit du texte ou de l'image. Ce n'est donc pas directement lié à la crédibilité de l'image. 

=== Experiment 3 ===
Elle regarde l'impact de l'image sur des "''trivia claims''". Elle utilise de phrases produites pour les articles de Unkelbach ([https://psycnet.apa.org/fulltext/2006-22946-014.pdf 2007]) et Nelson & Narens ([https://sites.socsci.uci.edu/~lnarens/1980/NelsonNarens_Jof%20Verbal%20Learning%20and%20Verbal%20Behavior_1980.pdf 1980]). 

==== Design ====
Variables: 
*VI1: photograph: yes, no
*VI2: difficulty: easy, hard
*VD: C score
Ce qui en fait un design (photograph: yes, no) x 2 (difficulty: easy, hard) within-subjects
design

==== Résultats ====
Elle a fait une repeated measures 2 (photograph: yes, no) x 2 (difficulty: easy, hard) ANOVA 
Elle a trouvé un effet de la photo : F(1, 69) = 6.65, p = .01, η<sub>p</sub><sup>2</sup> = .09

=== Summary of Experiments 1-3 ===
Photos did not produce an "alive bias," but a "truth bias." Moreover, the truthiness effect generalized beyond Dead or Alive judgements: Nonprobative photos enhanced the subjective truth of general knowledge claims, too.

Effect of photos on subjective truth is driven not simply by a perception that photos are inherently trustworthy. We speculate that nonprobative photos and verbal information help people generate pseudoevidence (cf. Kelly & Nace, 1994).

our findings suggest that even without repeated exposures or instructions to imagine, the mere presence of nonprobative information such as photos might rapidly inflate the perceived truth of many types of true and false claims 

=== Under what conditions do photos produce truthiness? === 
''There are reasons to predict that relative to related photos, unrelated photos should cause disfluency. That is, unrelated photos should make it more difficult for people to bring related thoughts and images to mind. For instance, in one study, when people saw a target word that was semantically unrelated to a preceding sentence (“The librarian reached for the top shelf and pulled down a book” “napkin”), they rated that target word as less pleasant than when they saw a target that was semantically related (“read”; Lee & Labroo, 2004)'' 

(...) ''The semantically unrelated words were relatively disfluent compared to their semantically related counterparts.''

(...) '''Parallel effects occur with perceptual manipulations; degraded images, difficult fonts, and low contrast colors can lead people to invest more cognitive effort to make sense of stimuli and people tend to be evaluate these stimuli more negatively (Diemand-Yauman, Oppenheimer, & Vaughan, 2011; Petrova & Cialdini, 2005; Reber & Schwarz 1999; see Yue, Castel, & Bjork, in press).'''

(...) ''That is, nonprobative, unrelated information should cause people to disbelieve a claim''

(...) 

''But there are also reasons to predict that unrelated photos would not cause falsiness''

(...) ''adding irrelevant information produces the same effect as giving people no information at all (Gill, Swann, & Silvera, 1998). Thus, a trawling mechanism would predict that unrelated photos would have little if any effect on people’s decisions.''

== Chapter 4 ==
=== Experiment 4 ===
==== Design ====
Elle a fait une manipulation d'un facteur intra-sujets à 3 niveaux (related photo, unrelated photo, no photo). Elle se base à nouveau sur les ''trivia claims'' récupérés de Nelson & Narens (1980) et Unkelbach (2007).

''To orient people to the task, for the first 16 trials they saw easy trivia claims (which tend to be answered correctly 80-100% of the time). Half these easy claims appeared with a related photograph, half with no photograph. To ensure the practice phase did not teach subjects a rule about the relationship between truth and the presence of photos, we paired photos equally often with true and false statements.''

''Immediately after these easy practice trivia claims, the experimental phase began. Subjects saw 24 difficult trivia claims. For one third of trials, a related nonprobative photo depicted the grammatical subject of the statement''

(...) ''for another third of trials, a semantically unrelated nonprobative photo appeared with the claim (for instance,
the claim about macadamia nuts appeared with a photo of a trash can). For the other third of the trials, people saw trivia claims presented without a photo.'' 

(...) ''As a set, the unrelated photos represented a similar range of living and non-living objects as the original set of related photos from Newman et al.''

==== Résultats ====
'''Unrelated photos did not produce truthiness; instead, trials with unrelated photos behaved more like trials with no photos.'''

Une ANOVA à 1 facteur a montré un effet de la photo F(2, 206) = 4.10, p = .02, η<sub>p</sub><sup>2</sup> = .04.

[[Fichier:Experiment 4 (Newman, 2013).png|600px|vignette|centré|alt=Experiment 4 (Newman, 2013)|Bias for difficult trivia claims presented with or without a photograph that was semantically related or unrelated to the trivia claims. A negative value of c is a bias to say true. Error bars show 95% within-subject confidence intervals for the photo factor for Experiment 4 (see Masson & Loftus, 2003)]]

''The finding that related, but not unrelated, photos produced truthiness replicates our earlier work and fits with the idea that unrelated photos would not affect people’s true/false decisions.''

(...)

''But it is possible that this context led subjects to find a semantic relationship between the claim and the photo, for both types of photos. For instance, they might have looked at the thermometer claim paired with a lizard, and thought “Well, the lizard is long and thin, like a thermometer, and has a stripe up the center just like a thermometer.” Such a strategy would have mitigated the perceived incongruency between the photos and trivia claims, leading unrelated photos to become more like related photos and diluting their falsiness. This idea fits with research showing that people will find or create meaning (Bartlett, 1932), especially when they are faced with pairings that don’t have an obvious semantic relationship (for instance, novel metaphors; Grimshaw, Stewart, & Lauwereyns, 2011; Lynott & Connell, 2010). Of course, a critic might argue that unrelated photos produced little bias compared to no photos for other, less interesting reasons: Perhaps people just adopted a strategy of ignoring unrelated photos, much like the way people in a Stroop task can adopt a strategy to ignore the word and focus on the color (Kane & Engle, 2003; Stroop, 1935; see also Besner, Stolz, & Boutilier, 1997). This strategy seems possible, but is unlikely on closer consideration. Because the unrelated photos randomly appeared among related photos, people must have processed photos enough to understand that they were unrelated.''

=== Experiment 5 ===
==== Design ====
Elle a fait un 2 (photo: yes, no) x 2 (relatedness: related, unrelated) mixed design

''We used the same procedure as Experiment 4 with the following changes. Immediately after seeing the easy practice trivia claims, people saw 32 difficult trivia claims (half true, half false). Half the claims appeared with a photo; for half the subjects the photo was always related to the claim whereas for the remaining subjects it was always unrelated to the claim with which it appeared.''

==== Résultats ====
Elle a fait une ANOVA mixte 2 x 2 qui a montré un '''effet d'interaction entre Photo et Relatedness''', F(1, 194) = 9.01, p < .01 ηp2 = .04
* les photos liées ont donc produit un effet de ''truthiness''
* les photos sans lien ont produit un effet de ''falsiness''

[[Fichier:Experiment 5 (Newman, 2013).png|600px|vignette|centré|alt=Experiment 5 (Newman, 2013)|Bias for difficult trivia claims presented with or without a photograph that was semantically related or unrelated to the trivia claims. A negative value of c is a bias to say true. Error bars show 95% within-subject confidence intervals for the photo effect for related and unrelated conditions in Experiment 5 (see Masson & Loftus, 2003)]]

==== Discussion ====
''When it is easier than expected to retrieve something, people interpret that processing discrepancy as a cue to truth; conversely, when it is more difficult than expected to retrieve something, people interpret this discrepancy the opposite way. But when processing matches expectations, there is no discrepancy to interpret.'' 

''If truthiness (and falsiness) depends on a standard, we should see that the pattern from Experiment 5 disappears when people have no standard against which to interpret the ease or difficulty of processing that accompanies claims paired with photos. Accordingly, in Experiment 6 we manipulated the photo factor entirely between-subjects.''

=== Experiment 6 ===
==== Design ====
''single-factor (photo: related, unrelated, no photo) between-subjects design.''

Elle a fait donc en sorte que: 
* 1/3 voient des photos liées à des ''trivia claims'
* 1/3 voient des photos non-liées à des ''trivia claims'
* 1/3 ne voient que les ''trivia claims''

==== Résultats ====
Elle a fait une ''one way ANOVA'' qui n'a montré aucun effet de l'image : F(2, 298) = .75, p = .47, η<sub>p</sub><sup>2</sup> = .01.

[[Fichier:Experiment 6 (Newman, 2013).png|600px|vignette|centré|alt=Experiment 6 (Newman, 2013)|Bias for difficult trivia claims presented with or without a photograph that was semantically related or unrelated to the trivia claims. A negative value of c is a bias to say true. Error bars show 95% confidence intervals for each cell mean in Experiment 6 (see Masson & Loftus, 2003)]]

==== Discussion ====
''These findings fit with the idea that truthiness or falsiness depend on expectations acquired in the experimental context, and only occur when there is a discrepancy in the expected ease of processing''

=== Summary of Experiments 4-6 === 
Bien que robustes, les effets des photos non probantes étaient relativement faibles dans nos études — peut-être parce que '''les gens pouvaient facilement identifier la source du traitement discordant'''. Une voie pour la recherche future pourrait être de rendre la manipulation de la photo moins évidente — la rendant plus difficile à identifier comme étant la source probable du traitement discordant. Une manipulation de ce type pourrait servir à renforcer les effets des photos et de la véracité perçue. (traduit par Mistral)

Pour info, dans sa procédure, elle a fait des pré-tests avec des gens qui devaient "''think aloud''" ce qui lui a donné quelques infos sur ce que les gens "faisaient" des photos, par exemple: 
* "''I’m going to go with yes because they kind of look like peaches, so that would make sense''" (pour l'affirmation "''Macadamia nuts are in the same evolutionary family as peache''")
* "''“I noticed... that tigers don’t really have eyebrows either so it was maybe giving you a bit of an influence.''" (pour "''The Mona Lisa has no eyebrows''" accompagnée d'une image de tigre).

== Chapter 5 - General discussion ==
=== Review ===
Elle a produit 7 expériences. Dans les 4 premières, elle a montré que ''related but nonprobative information can systematically bias people to conclude claims are true.''
* xp1A & 1B: adding a photograph of a celebrity biased people to say claims about a celebrity were true. (...) ''People will draw on tangential cues to inform especially when they do not have access to other information''
* xp2: giving people verbal descriptions of those photos produced the same effect
* xp3: photos that related to general knowledge claims also produced truthiness

Dans les 3 dernières, elle a exploré les conditions pour avoir l'effet de truthiness. Elle a trouvé que ''The semantic relationship between the photo and claim matters''
* xp4: ''in a within-subject design, related photos produced truthiness, but unrelated photos acted just like the no photo condition''
* xp5: ''In a mixed design, related photos produced truthiness and unrelated photos produced falsiness''
* xp6: ''<nowiki>[</nowiki>in a fully between-subjects design<nowiki>]</nowiki> The effect of photos (related and unrelated) was eliminated''

=== Connections with the Broader Literature ===
==== Mental Construal ====
''When people ponder a target stimulus in a concrete rather than abstract way, they tend to report that the target feels subjectively closer in time (or in physical distance) and is more likely to be true''

''Semantically related photos mightlead people to construe the trivia claim in more concrete terms, boosting its perceived truth.''

==== Reality Monitoring and Vividness ====
''people make decisions based on the characteristics of the mental event—using the amount of perceptual detail and ease of imagery as cues to reality''

''In one study, people saw a series of news headlines (true and false) and
judged whether they remembered the news event described in each headline
(Strange et al., 2011). Some of the headlines were presented along with a related “stock photo” (for instance: The false headline Blair under fire for botched Baghdad rescue attempt; won't step down, was paired with a photo of Tony Blair speaking in parliament). When people saw headlines paired with these stock photos, they were more likely to say they remembered the news event—even though the photos did not provide any evidence that the event occurre''

''Messages that are delivered in a vivid way—with concrete language, pictures, and additional details—are more persuasive than messages delivered in a less vivid way—with abstract language, no images and fewer details''

''One explanation for why vivid messages can be persuasive is that they tend to be better remembered; making it easier for people to retrieve and picture the message later — which might also work to bolster the perceived truth of the message''

==== The Seductive Allure of Brain Images ====
Dans une [[On the (non)persuasive power of a brain image|méta-analyse]], son équipe de recherche avait relevé que l'effet de l'impact d'une image de cerveau (présenté dans l'article de [[Seeing is believing: The effect of brain images on judgments of scientific reasoning|McCabe & Castel, 2008]]) était quasi nul. 

''The brain images were added to an explanation that already contained neuroscience, so it might be that these effects are not additive. That is, once people have had their 'hit' of neuroscience through the text, the photo has only a trivial effect on their decisions''

Comme c'était accompagné d'un texte "relativement long", il est possible que le texte donne déjà assez d'informations pour que les images aient un impact marginal. 

Elle dit aussi que ça pourrait être du au design: elle n'a pas eu non plus d'effet avec du between subject ; il y aurait peut être un meilleur effet en within. 

=== Avenues for Future Research ===
==== Does truthiness stick? ==== 
Son design est limité à quelques secondes ; est-ce que l'impact dure dans le temps? (étant donné que les images permettent de mieux se remémorer des souvenirs) 

''We would expect that compared to when people did not see a photo at time 1, photos would help people remember the statement at time 2 and should make it feel more cognitively available, familiar and true''

''at time 2, even if people just remember elements of the photo, it may still help them generate related thoughts images about the claim in the moment—boosting feelings of conceptual fluency and truth''

''if people learn an association between photos and truth at time 1, simply remembering that a statement appeared with a photo might lead people to conclude the statement is true at time 2''

''Given that people tend to remember pictures better than words, over time the effects of nonprobative verbal information might fade more rapidly than the effects of nonprobative photos''

==== Are Some People more Susceptible to Truthiness than Others? ====
Dans ses différentes expériences, il y a 30% des personnes qui ne sont pas tombées dans la ''truthiness'' ; il peut être intéressant de savoir pourquoi. 

Il existe des recherches qui montrent que certaines personnes ont plus facilement tendance à dire qu'elles ont vu une image, là où d'autres ont besoin de plus de preuves pour l'affirmer (Kantner & Lindsay, in press).

=== Practical Implications ===
==== Education ==== 
''What we do not know from our research is whether photos would do more than just bias people’s responses, perhaps photos would influence how much effort people exert when trying to make sense of new information''


==== Media and Public Information Campaigns ====
'' Although related nonprobative photos might boost the truthiness of messages, when paired with warnings or false information, photos could be dangerous.'' (Skurnik, Yoon, Park, & Schwarz, 2005).

''after a delay people remembered the message, but not the information regarding its accuracy''

=== Final Remarks === 

[[Catégorie:Résumés d'articles]]
[[Catégorie:Truthiness]]
[[Catégorie:Images]]
[[Catégorie:ANOVA]]
[[Catégorie:Fluency]]