Assessing the Effects of Disinformation Campaigns on Citizens: Experiment Design & Challenges
Campaign
Experimental Design
Public Opinion
Survey Experiments
Survey Research
Voting Behaviour
Abstract
Disinformation campaigns significantly affect public opinion and behavior, with examples such as COVID-19 misinformation reducing compliance with public health measures and far-right misinformation narratives boosting support for populist parties in Europe. This study investigates the multifaceted effects of disinformation on public attitudes, behaviors, and emotions, emphasizing the role of AI in crafting and amplifying multimedia disinformation. By conceptualizing disinformation narratives, the research evaluates their differential impacts on diverse audience groups, focusing on key outcomes such as changes in public opinion, electoral behavior, blame attribution, stereotyping, and emotional responses.
The methodological design relies on a randomised control trial (RCT) conducted via web-based surveys in Greece and Cyprus, with a pilot survey targeting university students in March and April 2025. Participants are randomly assigned to control or treatment groups within a quota structure balancing age, gender, region, and socio-economic profiles. The experimental design comprises a three-block structure.
1. The pre-treatment block collects demographic and political background data to serve as control variables.
2. The treatment block uses multimodal manipulations of news stories, framed around a topical issue such as climate change, to expose participants to disinformation, using some main disinformation techniques, including distortion of facts, emotive language, manipulated visuals, false attributions, and conspiracy theories.
3. The post-treatment block measures the dependent variables, such as changes in attitudes, blame perception, stereotypes, voting intentions, and emotional states, providing insights into disinformation's psychological and behavioral effects.
This comparative experiment evaluates the impact of different disinformation modalities (i.e., text, image, audio, and video-based treatments) against a control group exposed to authentic news. The control group serves as a baseline for assessing disinformation's impact, while random participants’ assignment to the control or the treatment groups ensures robustness and minimizes bias.
The findings aim to extend existing conceptualizations of disinformation, offering a comprehensive framework for analyzing and manipulating narratives. Addressing the methodological challenges of studying AI-driven disinformation, this research provides actionable insights to mitigate its societal impact and strengthen public resilience against misinformation.