In a recent study, researchers found that liberal-leaning media outlets exhibit more negative sentiments towards artificial intelligence (AI) compared to their conservative counterparts. This aversion in liberal media is primarily driven by concerns over AI’s potential to amplify societal biases. The findings have been published in Social Psychological and Personality Science.
The study’s roots lie in the increasingly prominent role of AI in our lives, impacting sectors from healthcare to job recruitment. Previous research has shown mixed reactions to AI’s growing influence, with some praising its efficiency and others worried about its societal impact, like job loss and privacy issues. This backdrop provided the researchers with a unique opportunity to explore how these divergent views play out in the media landscape, particularly through the lens of political ideologies.
“I was first interested in this topic because AI is a divisive issue,” said study author Angela Yi, a PhD student at Virginia Tech. “Some people embrace AI and some are more opposed to AI. Thus, it is increasingly important to understand which sections of society are more receptive to AI and which sections are more opposed to AI.”
“Given that my research interest lies in political ideology, I thought it would be fascinating to see the heterogeneous reactions to AI across political lines. To that end, I sought to assess partisan reactions to AI by analyzing partisan media sentiment toward AI.”
The study involved an analysis of media articles from both liberal and conservative outlets. Researchers turned to the Dow Jones & Company’s Factiva database to gather a robust selection of articles. The articles were chosen based on specific criteria: the articles must come from a media outlet with a known political leaning (as rated by AllSides’ Media Bias Rating Chart), they had to mention “algorithm” or “artificial intelligence,” be written in English, and published between May 25, 2019, and May 25, 2021. This process led to a substantial dataset of 7,840 articles.
In assessing the emotional tone of these articles, the research team employed a text analysis tool known as Linguistic Inquiry and Word Count (LIWC). This tool enabled them to measure the balance of positive and negative emotional words in each article, providing a nuanced view of the overall sentiment toward AI. In addition to the emotional tone, the analysis also factored in aspects like analytical thinking, authenticity, and the article’s publication date.
The study revealed a clear partisan divide in media sentiment towards AI. Liberal-leaning outlets, such as The New York Times and The Guardian, showed a more negative tone in their AI coverage compared to conservative-leaning outlets like The Wall Street Journal and The Daily Mail. This pattern held even after controlling for factors like the article’s analytical depth and authenticity.
One of the most striking findings was the role of societal bias concerns in shaping these sentiments. The liberal media’s heightened sensitivity to issues like racial and gender biases in AI algorithms was a key factor driving their negative tone. This concern was notably amplified following significant societal events, such as the death of George Floyd, which the researchers identified as a moment that heightened media focus on social biases, influencing AI coverage across the political spectrum.
On the other hand, the study found that concerns over job displacement and privacy issues, often highlighted in debates about AI, did not significantly explain the differences in media tone.
“We find that liberal-leaning media have a more negative sentiment to AI than conservative-leaning media,” Yi told PsyPost. “In other words, liberal-leaning media tends to be more opposed to AI than conservative-leaning media. We also find that liberal media are more concerned with AI magnifying social biases in society than conservative media. These social bias concerns explain liberal-leaning media’s greater aversion towards AI than conservative media. Privacy concerns and unemployment concerns did not explain the results.”
“Furthermore, we examined how media sentiment toward AI changed after George Floyd’s death. Since Floyd’s death sparked a national conversation about social biases in society, his death heightened social bias concerns in the media. This in turn led the media to become even more negative towards AI in their writing.”
While the study offers insightful observations, it’s important to recognize its limitations. The focus on U.S. and U.K. media outlets means the findings may not universally apply, especially in countries with different political landscapes and societal norms.
Future research could explore how media in other regions perceive AI, potentially revealing new dimensions to the AI discourse. Moreover, the study opens doors for further investigation into how media sentiment influences public opinion and policymaking regarding AI.
“Since we quantified the differences in media sentiment toward AI, the next step will be to see how social media conversations change as a function of the partisan media sentiment we see in this paper,” Yi said.
“I want to add that this research is descriptive rather than prescriptive,” the researcher added. “We are showing that there are differences that exist in the media sentiment in regard to AI and algorithms and that these differences are important to quantify and understand.”
The study, “Partisan Media Sentiment Toward Artificial Intelligence“, was authored by Angela Yi, Shreyans Goenka, and Mario Pandelaere.
