Perceptions of Deepfake Technology
The Influence of Recipients’ Affinity for Technology on the Perception of Deepfakes.
Master's thesis by Felix Kleine
3rd semester, 2022
Supervisor: Prof. Dr Sebastian Pranz
Problem and initial situation
Originally, the term ‘deepfake’ was understood to refer to the replacement or substitution of faces, facial features or facial expressions by artificial intelligence (AI). Currently, it is possible to alter entire bodies and voices in audiovisual media content. This paper addresses the issue of how deepfakes are perceived by audiences. It examines the extent to which perception depends on the viewer’s self-assessed affinity for technology.
Methods
First, an overview of the technical fundamentals and the potential of deepfake technology is provided. Then, following a detailed review of the literature on deepfakes, the research question is formulated and seven hypotheses are established.
RQ: Does the recipient’s affinity for technology influence the perception of
deepfakes?
H1: There are no significant differences in participants’ self-assessed affinity for technology depending on their age.
H2: There is no significant correlation between the participants’ self-assessed affinity for technology and their level of education.
H3: The participants’ self-assessment of their affinity for technology results in a higher mean value for male participants compared to female participants.
H4: Participants in the group with a high affinity for technology consider the deepfake advertisement to be just as authentic after the treatment as they did before the treatment.
H5: Participants in the group with a high affinity for technology are, on average, more familiar with the term ‘deepfake’ than others.
H6: There is no correlation between the ATI score and perceptions of the term ‘deepfake’.
H7: Participants in the group with a high affinity for technology perceive the use of deepfake technology as posing a lower risk.
For the study, a questionnaire was designed incorporating the ATI scale to measure affinity for technology. A within-subject design involving an information text made it possible to measure perceptions of authenticity regarding the deepfake. An advertising video served as an example of a deepfake. The online survey, in which 199 German-speaking participants took part, was statistically analysed and interpreted.
Results
The findings from the questionnaire contribute to research on the perception of deepfakes. It has been shown that the field of synthetic media is still in its infancy and constantly evolving. It urgently requires greater attention, not only in research, but also in legislation and in the commercial marketing of tangible goods and services.
To summarise the study’s findings, it was found that the participants’ affinity for technology had little influence on their perception of deepfakes. Male people considered themselves to be more tech-savvy than females. Participants with a greater affinity for technology had more background knowledge of the subject, were more impressed by deepfake technology and tended to view the term ‘deepfake’ less negatively. Although half of the participants are generally opposed to the use of deepfakes in advertising, a high level of fascination with the technology is evident. Of the alternative terms mentioned in the questionnaire, the term ‘AI-generated media’ was selected by the majority of participants as the term that best described the new technology and was perceived as the least negative.
Conclusion and Prospects
No evidence was found to suggest that technological affinity influences the assessment of authenticity or the risk posed by deepfakes. In general, trust in detection software as a countermeasure is low, and almost all participants would like to see a label when deepfake technology is used in advertisements. Half of the respondents find the use of deepfakes in advertisements unacceptable, regardless of the product. However, acceptance is lowest for products and services in the gambling, cosmetics, and hygiene and cleaning sectors. In these areas, the majority reject the use of deepfake technology in advertisements. In contrast, the level of rejection does not appear to be as strong for products and services in the fields of electronics, motor vehicles and tourism. The study provides new insights into current perceptions of deepfake technology in advertisements. Many people remain highly sceptical about the technology, even though there is a clear fascination with it. Its potential in marketing should also have become apparent.