https://orcid.org/0000-0002-0800-9355
![Sample our Humanities journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5939/TOC-Subject-Sample-2021-Humanities.jpg"})
ABSTRACT
While deepfake has emerged as a severe issue in the multimedia environment, most studies examined text-based false claims, leaving the question of what unique features of video-based deepfake news deceives recipients and how it can be corrected. By conducting two online experiments, we study perceived source vividness as a psychological mechanism of the effect of AI-enabled deepfake news on news credibility and engagement intentions. Furthermore, we test how an inserted false-tag onto the fake news can reduce the impact of source vividness experienced by seeing multimodal disinformation on news credibility and engagement intentions. The results suggest that participants who saw deepfake news had higher source vividness than those who saw fake news with other modalities (i.e., text-only and text-photo), and such source vividness increased credibility and engagement intentions of fake news. The false-tag successfully reduced engagement intentions of deepfake news for those who perceived a high vividness of the superimposed interviewee.
Acknowledgments
The authors thank the anonymous reviewers for their helpful comments that improved the manuscript.
Disclosure Statement
No potential conflict of interest was reported by the author(s).
Supplementary Material
Supplemental data for this article can be accessed on the publisher’s website.
Notes
1. A priori power analysis using R package “WebPower” (Zhang & Mai, Citation2018) indicated that a sample of at least N = 158 was needed to detect medium effect (Cohen’s f = 0.25) using a one-way ANOVA (1- β = .80; ⍺ = .05). Covariates were not considered in the priori power analysis, given that the covariates to be included in the main analysis were undecided at this stage (same for Study 2). The participants comprised of 53.8% male, 45.8% female and 0.8% others, with an average age of 39.90 years (SD = 12.85). The education of participants was distributed as follows: less than high school (0.2%), high school graduate (9.6%), some college but no degree (15.5%), Associate degree in college (2-year; 11.2%), Bachelor’s degree in college (4-year; 46.0%), Master’s degree/Professional degree (14.7%), and Doctoral degree (2.9%). Participants’ income was fairly distributed. Approximately 52.1% of participants had an annual household income of $59,999 or below. The mean value of conservatism as a political orientation was 3.45 (SD = 1.83), which was measured on a 7-point scale from 1 (liberal) to 7 (conservative).
3. To prevent any priming effects, participants were also asked to indicate their unfamiliarity with other persons including Chuck Schumer (Senator), Richard Shelby (Senator), and Doug Jones (Senator).
4. A power analysis using R package “WebPower” (Zhang & Mai, Citation2018) indicated that a sample of at least N = 158 was needed to detect medium effect (Cohen’s f = 0.25) using a two-way ANOVA (1- β = .80; ⍺ = .05). Of the total, 46.5% were male and 53.5% were female. The mean age was 43.40 years (SD = 13.13). The education level of participants was: less than high school degree (0.8%), high school graduate (7.8%), some college but no degree (21.6%), associate degree in college (2-year; 11.8%), Bachelor’s degree in college (4-year; 42.4%), Master’s degree/Professional degree (12.7%), and Doctoral degree (2.9%). Regarding participants’ income, approximately 50.6% of participants had an annual household income of $59,999 or below. The mean value of conservatism, measured on a 7-point scale ranging from 1 (liberal) to 7 (conservative), was 3.65 (SD = 1.92).
5. In a similar fashion of Study 1, participants were also asked to indicate their unfamiliarity with other actresses including Angelina Jolie, Jennifer Aniston, and Margot Robbie, for the sake of preventing any priming effects.