AI Determines When Entertainment Became More Important Than Politics

The study of over 100 British newspapers published in a period of 150 years was conducted by a team of researchers from the University of Bristol. The main focus of the study was to find out if major historical and cultural changes could be detected from the content of local press, which has always been one of the main sources of information.

The analysis of 35 million articles was conducted with the help of the neural networks, which are used to "read through" big data bases and large text arrays, and newspaper archive digitized by the company findmypast.

Some of the results were obvious, while other ones turned out quite unexpected. For example, the AI found that 1898 was the turning point from steam to electrical power, and trains overtook horses four years after, foreshadowing the future of transportation.

According to the analysis, the four largest peaks for "panic" corresponded with negative market movements linked to banking crises in 1826, 1847, 1857 and 1866.

The research team tracked the significant increase of "actors," "singers" and "dancers" in the 1890s, while references to "politicians," by contrast, gradually declined from the early 20^th century; Liberals were mentioned more than Conservatives until the 1930s, and "football" was more prominent than "cricket" from 1909.

The analysis also showed that changes in national identity took off in the 20^th century. Since then, people of the country were more often referred to as "British" rather than "English." In addition, males were systematically more present than females during the entire period studied, but there is a slow increase of the presence of women after 1900. Interestingly, the amount of gender bias in the news over the period of investigation is not very different from current levels.

"We have demonstrated that computational approaches can establish meaningful relationships between a given signal in large-scale textual corpora and verifiable historical moments," said Dr. Thomas Lansdall-Welfare, a post-doctoral researcher in machine learning at the Intelligent Systems Lab in the University of Bristol and one of the authors of the study published in the Proceedings of the National Academy of Sciences.

"However, what cannot be automated is the understanding of the implications of these findings for people, and that will always be the realm of the humanities and social sciences, and never that of machines," he added in an official university press release.

The team plans to perform more detailed studies on the same data, moving from general to more specific trends, such as calculating the ratings of politicians, scientists and artists. For example, it is already known that William Gladstone, a British liberal who served as the British Prime Minister for four times, was much more popular in the press in the 19^th century than his main rival, the Conservative Leader and twice-the-premier Benjamin Disraeli.