Defensive processing and the acceptance of misinformation on social media
2 Poster Text
2.1 Introduction
Psychology-related misinformation is widespread on social media, where unsubstantiated claims about mental health, trauma, and relationships are often presented alongside legitimate psychological research (Karasavva et al., 2025; Starvaggi et al., 2024). Although misinformation acceptance is frequently attributed to knowledge gaps or poor reasoning (Hoffner et al., 2026; Nan et al., 2022; Pan et al., 2021; Pennycook & Rand, 2020), growing work suggests that broader patterns of mistrust may play an important role in how individuals evaluate psychological information online (Pierre, 2020; van Zoonen et al., 2024). One source of this mistrust is inequality-driven mistrust (Jaiswal et al., 2020): skepticism shaped by experiences of marginalization or harm which then drives epistemic mistrust. A key component of inequality-driven mistrust is defensive processing, characterized by heightened vigilance, concern about being misled, and an expectation that information from others may cause harm (Perkins et al., 2025). Because misinformation is often conspiratorial, emotionally charged, and sensationalistic(Baum et al., 2024; Tamul & Hotter, 2019; Xu & Guo, 2018), defensive processing may paradoxically increase susceptibility of misinformation acceptance by heightening attention to threat-consistent cues and narratives that resonate with expectations of harm. The present study tests whether defensive processing predicts acceptance of psychology-related misinformation while controlling for baseline information acceptance.
2.2 Method
In Spring 2025, data was collected from a sample of students and adults with no college experience (n = 184). Participants responded to pre-manipulation items asking about stress, emotion dysregulation, analytical thinking, and inequality-driven mistrust (for sample items, see Figure 1). Participants were then randomly assigned to see majority accurate or inaccurate social posts discussing aspects of psychology, and to rate perceptions of truthfulness and accuracy. In addition to containing accurate or inaccurate information, posts included credibility claims or danger-everywhere themes (for examples, see Figure 2). Lastly, participants responded to post-manipulation items evaluating source credibility, the third-person effect, science populism, anomie, and social support.
2.2.1 Discussion
2.2.1.1 Defensive Processing & Misinformation Acceptance
These results suggest that defensive processing predicts misinformation acceptance. When controlling for accurate information acceptance, there was a positive relationship for all posts (credibility claims and danger everywhere themes). Although individuals high in defensive processing report greater suspicion toward new claims, this does not reliably translate into effective discernment between accurate and inaccurate content. Instead, it may bias attention toward surface-level indicators of legitimacy, such as credibility claims, or information with paranoid themes.
These findings suggest that fear-based messaging strategies (e.g., increasing awareness of the risks of misinformation) may not be effective in decreasing misinformation acceptance. They may also help explain why certain forms of misinformation (e.g., anti-vaccination myths, political conspiracies) gain traction while other types do not. More work is needed to understand the specific impact of inequality experiences on this aspect of mistrust; however, these results raise the possibility that groups experiencing marginalization or discrimination may be at elevated risk for misinformation acceptance.
2.2.1.2 Credibility & Danger
“Credbility-claim” (CC) posts appealed to expert authority and research, while “danger-everywhere” (DE) posts framed themselves as shedding light on pervasive and overlooked dangers. Defensive processing increased misinformation acceptance for both types of posts. However, being exposed to mostly accurate information moderated the relationship between defensive processing and misinformation acceptance for posts with credibility claims. When information is generally unreliable, defensive processing has a null effect on misinformation acceptance. When information is generally reliable, defensive processing predicts greater misinformation acceptance. These findings highlight the importance of informational context in shaping how mistrust operates. When most content in an environment is unreliable, all information is dismissed. In contrast, when the surrounding information is largely accurate, defensive processing appears to disrupt discernment, increasing susceptibility to misinformation that mimics legitimate signals of credibility.
These results have several potential implications. Interventions that focus solely on encouraging skepticism may inadvertently reinforce maladaptive forms of mistrust. Efforts to improve media literacy may need to distinguish between healthy critical evaluation and defensive processing. Finally, platform-level context may shape how individual differences in mistrust translate into belief, suggesting that reducing overall misinformation in an environment may not be sufficient to protect individuals who are high in defensive processing.
2.3 References
Baum, J., Frömer, R., & Abdel Rahman, R. (2024). Emotional content reduces the cognitive effort invested in processing the credibility of social (mis)information. Emotion, 24(6), 1468–1480. https://doi.org/10.1037/emo0001355
Hoffner, C. A., Salomi, V., Apkhazishvili, S., & Edu, S. (2026). Challenging mental health misinformation on social media: The role of eHealth literacy, self-efficacy and presumed media influence. Computers in Human Behavior, 175, 108844. https://doi.org/10.1016/j.chb.2025.108844
Jaiswal, J., LoSchiavo, C., & Perlman, D. C. (2020). Disinformation, Misinformation and Inequality-Driven Mistrust in the Time of COVID-19: Lessons Unlearned from AIDS Denialism. AIDS and Behavior, 24(10), 2776–2780. https://doi.org/10.1007/s10461-020-02925-y
Karasavva, V., Miller, C., Groves, N., Montiel, A., Canu, W., & Mikami, A. (2025). A double-edged hashtag: Evaluation of #ADHD-related TikTok content and its associations with perceptions of ADHD. PLOS ONE, 20(3), e0319335. https://doi.org/10.1371/journal.pone.0319335
Nan, X., Wang, Y., & Thier, K. (2022). Why do people believe health misinformation and who is at risk? A systematic review of individual differences in susceptibility to health misinformation. Social Science & Medicine, 314, 115398. https://doi.org/10.1016/j.socscimed.2022.115398
Pan, W., Liu, D., & Fang, J. (2021). An Examination of Factors Contributing to the Acceptance of Online Health Misinformation. Frontiers in Psychology, 12. https://doi.org/10.3389/fpsyg.2021.630268
Pennycook, G., & Rand, D. G. (2020). Who falls for fake news? The roles of bullshit receptivity, overclaiming, familiarity, and analytic thinking. Journal of Personality, 88(2), 185–200. https://doi.org/10.1111/jopy.12476
Perkins, H., Fisher, R., Akinapally, M., Sullivan, S., & Agramonte, M. (2025). Minority Stress and Epistemic Mistrust: Understanding Inequality-Driven Mistrust. 2025 Society for Personality and Social Psychology Annual Convention.
Pierre, J. M. (2020). Mistrust and Misinformation: A Two-Component, Socio-Epistemic Model of Belief in Conspiracy Theories. Journal of Social and Political Psychology, 8(2), Article 2. https://doi.org/10.5964/jspp.v8i2.1362
Starvaggi, I., Dierckman, C., & Lorenzo-Luaces, L. (2024). Mental health misinformation on social media: Review and future directions. Current Opinion in Psychology, 56, 101738. https://doi.org/10.1016/j.copsyc.2023.101738
Tamul, D., & Hotter, J. (2019). Exploring Mechanisms of Narrative Persuasion in a News Context: The Role of Narrative Structure, Perceived Similarity, Stigma, and Affect in Changing Attitudes. Collabra: Psychology, 5, 51. https://doi.org/10.1525/collabra.172
van Zoonen, W., Luoma-aho, V., & Lievonen, M. (2024). Trust but verify? Examining the role of trust in institutions in the spread of unverified information on social media. Computers in Human Behavior, 150, 107992. https://doi.org/10.1016/j.chb.2023.107992
Xu, Z., & Guo, H. (2018). Using Text Mining to Compare Online Pro- and Anti-Vaccine Headlines: Word Usage, Sentiments, and Online Popularity. Communication Studies. (world). https://www.tandfonline.com/doi/abs/10.1080/10510974.2017.1414068
3 Poster
knitr::include_graphics("images/poster.png")
4 Load Libraries
library(psych)
library(kableExtra)
library(ggplot2)
library(ggrepel)
library(tidyr)
library(dplyr)
# library(gtools)
library(nFactors)
library(corrplot)
library(afex)
library(naniar)
library(stringr)
library(emmeans)
library(lavaan)
library(sjPlot)
library(purrr)
library(DT)
library(svglite)5 Load Data
df1 <- read.csv(file="data/Learning About Psychology (SONA)_August 22, 2025_20.13.csv", header=T, na.strings = c(""," "))
labels1 <- data.frame(t(cbind(df1[1,],samp="Sample Source")))
df1 <- df1[-c(1,2), ]
df1 <- subset(df1, Progress == "100", select=-c(4,14:15))
df1$samp <- "SONA"
df2 <- read.csv(file="data/Learning About Psychology (Prolific)_August 22, 2025_20.14.csv", header=T, na.strings = c(""," "))
labels2 <- data.frame(t(cbind(df2[1,],samp="Sample Source")))
df2 <- df2[-c(1,2), ]
# df2_quality <- subset(df2, select=c(8,173:175))
df2 <- subset(df2, Progress == "100" & Q1.1 == 2 & Q2 < 4, select=-c(4,14:15))
df2$id <- df2$ResponseId
df2$samp <- "Prolific"
common_cols <- intersect(names(df1), names(df2))
df <- bind_rows(
select(df1, all_of(common_cols)),
select(df2, all_of(common_cols))
)
flag <- labels1 %>%
filter(grepl("truthful", X1))
flag2 <- rownames(flag)
a_list <- c("23","34","36","38","48","52","56","60","64","70","72","76","80","84","88")
m_list <- c("40","42","44","46","50","54","58","62","66","68","74","78","82","86","90")
# Identify target columns
target_cols <- grep(
paste(c(a_list, m_list), collapse = "|"),
names(df),
value = TRUE
)
# Recode only for Prolific participants
df[df$samp == "Prolific", target_cols] <- lapply(
df[df$samp == "Prolific", target_cols],
function(x) {
x <- as.numeric(x)
x[x == 4] <- 3
x[x == 5] <- 4
x
}
)5.1 Missing Data
d <- subset(df, select=c(15:79,200:248))
# total percentage of missing data
mean(is.na(d)) ## [1] 0.009459132# total number of participants (rows) with any missing data
(sum(rowSums(is.na(d)) > 0))/nrow(d)## [1] 0.03225806d <- na.omit(subset(df, select=c(id,15:79,200:248)))
backup <- df
df <- subset(df, df$id %in% d$id)5.2 Attention Checks
# attention check
attn <- subset(labels1, grepl("attention", X1))
attn_chk <- rownames(attn)
# Q18_12 == 4
df$Q18_12 <- as.numeric(df$Q18_12)
table(df$Q18_12)##
## 1 2 3 4
## 5 7 4 194# Q28_4 > 9
df$Q28_4 <- as.numeric(df$Q28_4)
table(df$Q28_4)##
## 1 2 3 4 5 6 7 8 9 10 11
## 1 1 3 3 4 3 2 4 1 63 125table(df$Q28_4 <= 9, useNA = "ifany")##
## FALSE TRUE
## 188 22backup <- df
df <- subset(df, df$Q28_4 > 9 & Q18_12 == 4)6 Misinformation Acceptance
df_temp_subset <- select(df, all_of(flag2))
df_temp_subset[] <- lapply(df_temp_subset, function(x) {
x[x == ""] <- NA
as.numeric(x)
})
df_temp_subset_a <- df_temp_subset[ , grepl(paste(a_list, collapse="|"), names(df_temp_subset))]
df_temp_subset_m <- df_temp_subset[ , grepl(paste(m_list, collapse="|"), names(df_temp_subset))]
a_list_cred <- c("23","34","36","38","48","70","84")
m_list_cred <- c("40","44","46","50","54","68","90")
a_list_de <- c("38","48","52","56","76","88")
m_list_de <- c("42","46","50","54","58","62","78","86")
df_temp_subset_a_cred <- df_temp_subset[ , grepl(paste(a_list_cred, collapse="|"), names(df_temp_subset))]
df_temp_subset_m_cred <- df_temp_subset[ , grepl(paste(m_list_cred, collapse="|"), names(df_temp_subset))]
df$acc_cred <- rowMeans(df_temp_subset_a_cred, na.rm = TRUE)
df$mis_cred <- rowMeans(df_temp_subset_m_cred, na.rm = T)
df_temp_subset_m_de <- df_temp_subset[ , grepl(paste(m_list_de, collapse="|"), names(df_temp_subset))]
df_temp_subset_a_de <- df_temp_subset[ , grepl(paste(a_list_de, collapse="|"), names(df_temp_subset))]
df$mis_rating_de <- rowMeans(df_temp_subset_m_de, na.rm = T)
df$acc_rating_de <- rowMeans(df_temp_subset_a_de, na.rm = T)
df <- df %>%
mutate(
across(
c(acc_cred, acc_rating_de, mis_cred, mis_rating_de),
~ ifelse(is.nan(.), NA, .)
)
)
# build combined item sets (unique), then subset once
a_list_all <- unique(c(a_list_cred, a_list_de))
m_list_all <- unique(c(m_list_cred, m_list_de))
df_temp_subset_a_all <- df_temp_subset[ , grepl(paste(a_list_all, collapse="|"), names(df_temp_subset))]
df_temp_subset_m_all <- df_temp_subset[ , grepl(paste(m_list_all, collapse="|"), names(df_temp_subset))]
df$acc_rating <- rowMeans(df_temp_subset_a_all, na.rm = TRUE)
df$mis_rating <- rowMeans(df_temp_subset_m_all, na.rm = TRUE)
df <- df %>%
mutate(
across(
c(acc_rating, mis_rating),
~ ifelse(is.nan(.), NA, .)
)
)7 Inequality-Driven Mistrust
df$Q18_1 <- as.numeric(df$Q18_1)
df$Q18_2 <- as.numeric(df$Q18_2)
df$Q18_3 <- as.numeric(df$Q18_3)
df$Q18_4 <- as.numeric(df$Q18_4)
df$Q18_5 <- as.numeric(df$Q18_5)
df$Q18_6 <- as.numeric(df$Q18_6)
df$Q18_7 <- as.numeric(df$Q18_7)
df$Q18_8 <- as.numeric(df$Q18_8)
df$Q18_9 <- as.numeric(df$Q18_9)
df$Q18_10 <- as.numeric(df$Q18_10)
df$Q18_11 <- as.numeric(df$Q18_11)
df$Q18_12 <- as.numeric(df$Q18_12)
df$Q18_13 <- as.numeric(df$Q18_13)
df$Q18_14 <- as.numeric(df$Q18_14)
df$Q18_15 <- as.numeric(df$Q18_15)
df$Q18_16 <- as.numeric(df$Q18_16)
df$idm_dp <- (df$Q18_13 + df$Q18_14 + df$Q18_15 + df$Q18_16) / 4
df$idm_da <- (df$Q18_2 + df$Q18_8 + df$Q18_9 + df$Q18_10 + df$Q18_11) / 5
df$idm_ib <- (df$Q18_3 + df$Q18_4 + df$Q18_5 + df$Q18_6) / 4
describe(subset(df, select=c(idm_dp, idm_da, idm_ib)))## vars n mean sd median trimmed mad min max range skew kurtosis se
## idm_dp 1 184 2.36 0.69 2.50 2.37 0.74 1 4 3 -0.12 -0.56 0.05
## idm_da 2 184 1.88 0.74 1.80 1.80 0.89 1 4 3 0.67 -0.19 0.05
## idm_ib 3 184 2.33 0.89 2.25 2.31 1.11 1 4 3 0.05 -1.12 0.07psych::alpha(subset(df, select=c(Q18_13, Q18_14, Q18_15, Q18_16)))##
## Reliability analysis
## Call: psych::alpha(x = subset(df, select = c(Q18_13, Q18_14, Q18_15,
## Q18_16)))
##
## raw_alpha std.alpha G6(smc) average_r S/N ase mean sd median_r
## 0.74 0.74 0.7 0.42 2.9 0.031 2.4 0.69 0.45
##
## 95% confidence boundaries
## lower alpha upper
## Feldt 0.67 0.74 0.8
## Duhachek 0.68 0.74 0.8
##
## Reliability if an item is dropped:
## raw_alpha std.alpha G6(smc) average_r S/N alpha se var.r med.r
## Q18_13 0.75 0.75 0.67 0.50 3.0 0.032 0.0012 0.50
## Q18_14 0.64 0.64 0.57 0.37 1.8 0.047 0.0211 0.33
## Q18_15 0.69 0.69 0.60 0.42 2.2 0.040 0.0072 0.43
## Q18_16 0.65 0.65 0.57 0.38 1.9 0.044 0.0129 0.43
##
## Item statistics
## n raw.r std.r r.cor r.drop mean sd
## Q18_13 184 0.68 0.67 0.49 0.42 2.8 0.94
## Q18_14 184 0.81 0.80 0.70 0.61 2.6 0.97
## Q18_15 184 0.74 0.75 0.63 0.53 2.1 0.89
## Q18_16 184 0.78 0.79 0.69 0.59 2.0 0.88
##
## Non missing response frequency for each item
## 1 2 3 4 miss
## Q18_13 0.10 0.27 0.38 0.24 0
## Q18_14 0.18 0.23 0.43 0.16 0
## Q18_15 0.30 0.39 0.25 0.06 0
## Q18_16 0.31 0.42 0.20 0.07 07.1 Item Stacked Bar
resp_labs <- c(
"1" = "Strongly disagree",
"2" = "Disagree",
"3" = "Agree",
"4" = "Strongly agree"
)
long <- df %>%
select(Q18_13:Q18_16) %>%
pivot_longer(everything(), names_to = "item", values_to = "resp") %>%
filter(!is.na(resp)) %>%
mutate(
resp = factor(resp, levels = 1:4, labels = resp_labs),
dir = ifelse(resp %in% resp_labs[1:2], -1, 1)
)
plot_df <- long %>%
count(item, resp, dir) %>%
mutate(n = n * dir)
plot_df <- plot_df %>%
mutate(resp = factor(resp,
levels = c("Strongly disagree","Disagree","Agree","Strongly agree")))
item_labs <- c(
Q18_13 = "I’d prefer to find things out for myself on the internet rather than asking others.",
Q18_14 = "If you put too much faith in what people tell you, you are likely to get hurt.",
Q18_15 = "When someone tells me something, my immediate reaction is to wonder why they are telling me this.",
Q18_16 = "I don’t usually act on advice that I get from others even when I think it’s probably sound."
)
ggplot(plot_df, aes(x = item, y = n, fill = resp)) +
geom_bar(stat = "identity", position = position_stack(reverse = TRUE), width = 0.7) +
coord_flip(clip = "off") +
scale_x_discrete(
labels = function(x) str_wrap(item_labs[x], width = 45)
) +
geom_hline(yintercept = 0) +
scale_y_continuous(labels = abs, expand = expansion(mult = c(0.15, 0.15))) +
labs(x = NULL, y = NULL, fill = NULL) +
theme_bw() +
theme(
axis.text.x = element_blank(),
axis.ticks.x = element_blank(),
panel.grid.major.x = element_blank(),
panel.grid.minor.x = element_blank(),
panel.grid.minor.y = element_blank()
) + theme(
legend.position = c(0.98, 0.98),
legend.justification = c(1, 1),
legend.background = element_rect(fill = "white", colour = NA),
legend.key = element_blank()
) + scale_fill_manual(
breaks = c("Strongly disagree","Disagree","Agree","Strongly agree"),
values = c(
"Strongly disagree" = "grey25",
"Disagree" = "grey45",
"Agree" = "grey65",
"Strongly agree" = "grey85"
)
)
# p <- ggplot(plot_df, aes(x = item, y = n, fill = resp)) +
# geom_bar(stat = "identity", position = position_stack(reverse = TRUE), width = 0.7) +
# coord_flip(clip = "off") +
# scale_x_discrete(labels = function(x) stringr::str_wrap(item_labs[x], width = 35)) +
# geom_hline(yintercept = 0) +
# scale_y_continuous(labels = abs, expand = expansion(mult = c(0.15, 0.15))) +
# labs(x = NULL, y = NULL, fill = NULL) +
# theme_bw() +
# theme(
# axis.text.y = element_text(size = 28), # item labels (y-axis visually, because coord_flip)
# legend.text = element_text(size = 20),
# legend.title = element_blank(),
# axis.text.x = element_blank(),
# axis.ticks.x = element_blank(),
# panel.grid.major.x = element_blank(),
# panel.grid.minor.x = element_blank(),
# panel.grid.minor.y = element_blank(),
# legend.position = c(0.98, 0.98),
# legend.justification = c(1, 1),
# legend.background = element_rect(fill = "white", colour = NA),
# legend.key = element_blank()
# ) +
# scale_fill_manual(
# breaks = c("Strongly disagree","Disagree","Agree","Strongly agree"),
# values = c(
# "Strongly disagree" = "grey25",
# "Disagree" = "grey45",
# "Agree" = "grey65",
# "Strongly agree" = "grey85"
# )
# ) + guides(
# fill = guide_legend(
# keyheight = unit(32, "pt"), # taller rows (main “line height” control)
# label.position = "right",
# byrow = TRUE
# )
# )
#
# p
#
# ggsave(
# filename = "Q18_stacked.svg",
# plot = p,
# device = svglite::svglite,
# width = 18.88,
# height = 8,
# units = "in"
# )8 Regressions
8.1 Overall
# create z-scored versions without overwriting originals
vars <- c(
"mis_rating",
"acc_rating",
"mis_rating_de",
"acc_rating_de",
"acc_cred",
"mis_cred",
"idm_dp"
)
df[paste0(vars, "_z")] <- scale(df[vars])
df$condition <- as.factor(df$condition)
# reg_out1 <- lm(mis_rating_de ~ acc_rating_de, data = df)
reg_out1 <- lm(mis_rating_z ~ acc_rating_z, data = df)
summary(reg_out1)##
## Call:
## lm(formula = mis_rating_z ~ acc_rating_z, data = df)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.04222 -0.55996 -0.01573 0.56740 1.62311
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 5.650e-17 5.804e-02 0.00 1
## acc_rating_z 6.193e-01 5.820e-02 10.64 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.7873 on 182 degrees of freedom
## Multiple R-squared: 0.3836, Adjusted R-squared: 0.3802
## F-statistic: 113.2 on 1 and 182 DF, p-value: < 2.2e-16# reg_out2 <- lm(mis_rating_de ~ idm_dp*condition + acc_rating_de, data = df)
reg_out2 <- lm(mis_rating_z ~ idm_dp_z*condition + acc_rating_z, data = df)
summary(reg_out2)##
## Call:
## lm(formula = mis_rating_z ~ idm_dp_z * condition + acc_rating_z,
## data = df)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.97219 -0.56942 -0.02887 0.55443 1.59197
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.003740 0.083354 -0.045 0.9643
## idm_dp_z 0.202464 0.086620 2.337 0.0205 *
## conditionm 0.008011 0.116667 0.069 0.9453
## acc_rating_z 0.611219 0.058999 10.360 <2e-16 ***
## idm_dp_z:conditionm -0.165975 0.116441 -1.425 0.1558
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.7816 on 179 degrees of freedom
## Multiple R-squared: 0.4025, Adjusted R-squared: 0.3892
## F-statistic: 30.15 on 4 and 179 DF, p-value: < 2.2e-16# reg_out3 <- lm(acc_rating_de ~ idm_dp*condition + mis_rating_de, data = df)
# summary(reg_out3)8.2 Danger Everywhere
# reg_out1 <- lm(mis_rating_de ~ acc_rating_de, data = df)
reg_out1 <- lm(mis_rating_de_z ~ acc_rating_de_z, data = df)
summary(reg_out1)##
## Call:
## lm(formula = mis_rating_de_z ~ acc_rating_de_z, data = df)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.37696 -0.63856 0.09766 0.69106 1.61904
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.04513 0.07056 0.640 0.523
## acc_rating_de_z 0.51135 0.07100 7.202 2.47e-11 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.8811 on 154 degrees of freedom
## (28 observations deleted due to missingness)
## Multiple R-squared: 0.2519, Adjusted R-squared: 0.2471
## F-statistic: 51.87 on 1 and 154 DF, p-value: 2.466e-11# reg_out2 <- lm(mis_rating_de ~ idm_dp*condition + acc_rating_de, data = df)
reg_out2 <- lm(mis_rating_de_z ~ idm_dp_z*condition + acc_rating_de_z, data = df)
summary(reg_out2)##
## Call:
## lm(formula = mis_rating_de_z ~ idm_dp_z * condition + acc_rating_de_z,
## data = df)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.61723 -0.63786 0.09097 0.64020 1.59728
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.05871 0.09587 0.612 0.5412
## idm_dp_z 0.24720 0.10247 2.412 0.0171 *
## conditionm -0.02089 0.14061 -0.149 0.8821
## acc_rating_de_z 0.50354 0.07064 7.128 3.91e-11 ***
## idm_dp_z:conditionm -0.16798 0.14086 -1.193 0.2349
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.8713 on 151 degrees of freedom
## (28 observations deleted due to missingness)
## Multiple R-squared: 0.2827, Adjusted R-squared: 0.2637
## F-statistic: 14.88 on 4 and 151 DF, p-value: 2.837e-10# reg_out3 <- lm(acc_rating_de ~ idm_dp*condition + mis_rating_de, data = df)
# summary(reg_out3)8.2.1 DE Plot
# residualized DV: mis | acc
mis_resid <- resid(
lm(mis_rating_de_z ~ acc_rating_de_z, data = model.frame(reg_out2))
)
# residualized predictor: idm | acc
idm_resid <- resid(
lm(idm_dp_z ~ acc_rating_de_z, data = model.frame(reg_out2))
)
plot_df <- data.frame(
idm_resid = idm_resid,
mis_resid = mis_resid,
condition = model.frame(reg_out2)$condition
)
ggplot(plot_df, aes(x = idm_resid, y = mis_resid, color = condition)) +
geom_point(alpha = .6) +
geom_smooth(method = "lm", se = TRUE, color = "black") +
scale_color_manual(
name = "Condition",
values = c("black", "#E6B10E"),
labels = c("Majority Accurate", "Majority Misinformation")
) +
labs(
x = "Defensive Processing",
y = "Misinformation Acceptance (Danger Everywhere Theme)"
) +
theme_bw() +
theme(
text = element_text(size = 11),
legend.position = c(0.95, 0.95),
legend.justification = c(1, 1)
)
# p <- ggplot(plot_df, aes(x = idm_resid, y = mis_resid, color = condition)) +
# geom_point(alpha = .6) +
# geom_smooth(method = "lm", se = TRUE, color = "black") +
# scale_color_manual(
# name = "Condition",
# values = c("black", "#E6B10E"),
# labels = c("Majority Accurate", "Majority Misinformation")
# ) +
# labs(
# x = "Defensive Processing",
# y = "Misinformation Acceptance (Danger Everywhere Theme)"
# ) +
# theme_bw() +
# theme(
# text = element_text(size = 20),
# legend.position = c(0.95, 0.95),
# legend.justification = c(1, 1)
# )
#
# p
#
# ggsave(
# filename = "dereg.svg",
# plot = p,
# device = svglite::svglite,
# width = 7,
# height = 8.46,
# units = "in"
# )8.3 Credibility Claims
# reg_out1 <- lm(mis_cred ~ acc_cred, data = df)
reg_out1 <- lm(mis_cred_z ~ acc_cred_z, data = df)
summary(reg_out1)##
## Call:
## lm(formula = mis_cred_z ~ acc_cred_z, data = df)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.64949 -0.46702 -0.04269 0.52877 2.12770
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.003379 0.065315 -0.052 0.959
## acc_cred_z 0.506194 0.064773 7.815 5.23e-13 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.8615 on 172 degrees of freedom
## (10 observations deleted due to missingness)
## Multiple R-squared: 0.262, Adjusted R-squared: 0.2577
## F-statistic: 61.07 on 1 and 172 DF, p-value: 5.233e-13# reg_out2 <- lm(mis_cred ~ idm_dp*condition + acc_cred, data = df)
reg_out2 <- lm(mis_cred_z ~ idm_dp_z*condition + acc_cred_z, data = df)
summary(reg_out2)##
## Call:
## lm(formula = mis_cred_z ~ idm_dp_z * condition + acc_cred_z,
## data = df)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.75061 -0.49977 0.00809 0.54241 2.33011
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -0.06194 0.09637 -0.643 0.52130
## idm_dp_z 0.27583 0.09744 2.831 0.00521 **
## conditionm 0.11122 0.13145 0.846 0.39870
## acc_cred_z 0.50802 0.06574 7.728 9.26e-13 ***
## idm_dp_z:conditionm -0.24634 0.12910 -1.908 0.05807 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.847 on 169 degrees of freedom
## (10 observations deleted due to missingness)
## Multiple R-squared: 0.2992, Adjusted R-squared: 0.2826
## F-statistic: 18.04 on 4 and 169 DF, p-value: 2.364e-12plot_model(reg_out2, type="int") + ylim(-2,2)
# reg_out3 <- lm(acc_cred ~ idm_dp*condition + mis_cred, data = df)
# summary(reg_out3)8.3.1 Cred Plot
# residualized DV: mis | acc
mis_resid <- resid(
lm(mis_cred_z ~ acc_cred_z, data = model.frame(reg_out2))
)
# residualized predictor: idm | acc
idm_resid <- resid(
lm(idm_dp_z ~ acc_cred_z, data = model.frame(reg_out2))
)
plot_df <- data.frame(
idm_resid = idm_resid,
mis_resid = mis_resid,
condition = model.frame(reg_out2)$condition
)
ggplot(plot_df, aes(x = idm_resid, y = mis_resid, color = condition)) +
geom_point(alpha = .6) +
geom_smooth(method = "lm", se = TRUE, color = "black") +
scale_color_manual(
name = "Condition",
values = c("black", "#E6B10E"),
labels = c("Majority Accurate", "Majority Misinformation")
) +
labs(
x = "Defensive Processing",
y = "Misinformation Acceptance (Credibility Claims)"
) +
theme_bw() +
theme(
text = element_text(size = 11),
legend.position = c(0.95, 0.95),
legend.justification = c(1, 1)
)
#
# p <- ggplot(plot_df, aes(x = idm_resid, y = mis_resid, color = condition)) +
# geom_point(alpha = .6) +
# geom_smooth(method = "lm", se = TRUE, color = "black") +
# scale_color_manual(
# name = "Condition",
# values = c("black", "#E6B10E"),
# labels = c("Majority Accurate", "Majority Misinformation")
# ) +
# labs(
# x = "Defensive Processing",
# y = "Misinformation Acceptance (Credibility Claims)"
# ) +
# theme_bw() +
# theme(
# text = element_text(size = 20),
# legend.position = c(0.95, 0.95),
# legend.justification = c(1, 1)
# )
#
# p
#
# ggsave(
# filename = "ccreg.svg",
# plot = p,
# device = svglite::svglite,
# width = 7,
# height = 8.46,
# units = "in"
# )