Tobias Kaas Rasmussen and Jakob Hansen Zaar from Political Science at SDU received our first DDC thesis grant to work on their master’s thesis
Tobias Kaas Rasmussen and Jakob Hansen Zaar from Political Science at the University of Southern Denmark received our first DDC thesis grant of 10,000 DKK to work on their master’s thesis titled "Has the Crisis of Trust Been Averted? The Fear of AI is Significant – The Effect is Minimal," which they successfully submitted and defended in July 2024. Here, they investigated how the use of generative artificial intelligence in political communication affects the publics’ trust in politicians.
In reflecting on their experience, Tobias Kaas Rasmussen and Jakob Hansen Zaar explain what receiving the grant meant for their work on their master’s thesis: “We were fortunate to receive the DDC thesis grant for our thesis. The grant allowed us to distribute our survey through the pollster Epinion. This significantly enhanced the academic value of our results, as we obtained a larger and more representative sample than if we had distributed the survey ourselves. Fundamentally, receiving the DDC thesis grant meant a great deal to us. It allowed us to focus on fine-tuning our thesis to achieve a result that was not only a good but also academically relevant and a valuable contribution to the literature on trust and democracy in the digital age”.
You can read more about their interesting project and findings in the abstract below.
Abstract:
The utilization of generative artificial intelligence for strategic political communication has become feasible, yet the implications for trust levels remain largely unexplored. This study examines the relationship between AI-mediated communication and trust in politicians. The extant literature on this subject is nascent and yields divergent findings, underscoring the necessity for precise causal inference. We designed an experiment to assess the impact of politicians employing artificial intelligence as a communication tool on Facebook. The results provide robust evidence that trust in politicians who use AI diminishes. This decline is attributable to an inherent aversion to AI. Furthermore, our findings indicate that this effect is mediated by the perceived authenticity of the politician. Contrary to traditional expectations, our results suggest that contemporary technologies, particularly large language models, necessitate a reevaluation of existing assumptions. While technology has traditionally been viewed as unbiased, our study reveals a growing awareness of the embedded biases within generative AI. This is particularly evident as only factual Facebook posts generated by AI are perceived as less trustworthy, whereas emotional posts do not significantly impact the perceived credibility of the politician. Our contribution to the literature is twofold: (1) we advocate for a revised conceptualization of the public’s relationship with generative AI, and (2) we demonstrate that the decrease in trust when politicians use AI is primarily due to a reduction in perceived authenticity.