Benjamin LeBrun's ARIA project:听The Impact of Cost-Cutting on Journalism
听
My project sought to analyze the effects of a recent journalistic cost cutting trend dubbed The Mavening. The financial decision characterizing the mavening involves dramatic reductions in staff and budget and have in recent years affected a number of prominent publications (e.g. Newsweek, Sports Illustrated, LA Weekly, Deadspin, The Denver Post). These cost-cutting decisions are usually carried out abruptly and are often due to changes in ownership. For example, just twenty-four hours after Sports Illustrated was sold to the media conglomerate The Maven, nearly half of SI鈥檚 editorial staff was laid off. Critics of the process have hypothesized that it has a negative effect on the journalistic content of affected publications. My project aimed to test this hypothesis.
More precisely, the goal of this project was to generalize about the effects that such financial decision making has on the linguistic style of mavened publications. To do so, we assembled a large data set of articles from a number of publications having undergone a mavening event. Overall, we collected more than 280,000 articles from four mavened publications (Newsweek, Sports Illustrated, LA Weekly, and Deadspin). We then used computational tools to evaluate the periods before and after the moments of transformation for reliable differences in writing style. Our models attempted to capture a trend of stylistic simplification. That is, we expected to see a decline in complex linguistic behaviour in the period following the mavening event.
We found a number of statistically significant changes supporting the stylistic simplification hypothesis. These observed changes include a decline in sensory-based description of events as well as less semantic continuity. There were, however, a number of cases in which our findings did not neatly map onto our hypothesis. As it turns out, the process of modelling, which entails translating our intuitions about certain phenomena into generalizable and verifiable metrics, proved to be both the most challenging and rewarding parts of the project.
On the one hand, building computational models that align with affective judgements about language and culture is a difficult task. A large portion of the project was spent validating our models to assess whether they actually captured the phenomena there were supposed to. In many cases, I felt as though there were a number of limitations to some of our measures. That being said, this was to be expected. The field of computational analysis of culture is relatively new, and as a result, is still in the early stages of building reliable and generalizable metrics to model cultural behaviour at scale. And in fact, while these limitations were disappointing and frustrating at times, they were a reminder that, especially in research, it is the process, not the outcome, that is most valuable.
Indeed, some of the most rewarding moments of the summer involved learning and researching new ways to study language. In particular, the information-theoretic measures of entropy and surprisal, which model language probabilistically, introduced me to a fascinating way of quantifying the processes underlying language. This was one of the main reasons why I wanted to pursue this particular ARIA project: to get an overview of some of the ways we can study language and culture computationally.
In fact, throughout the summer, I found myself taking interest more so in the measures themselves than in their applications. In other words, my curiosity shifted from the insights offered by the analyses to the models used to arrive at such insights. The limitations discussed above point to the fact that there is work to be done in improving the way we study culture at scale. While this can be frustrating at times, the idea of building more accurate and generalizable tools is one that is exciting to me. In that regard, this project facilitated important moments of self-reflection regarding the direction I would like to take in the future. And while these realizations were some of the more unexpected outcomes of the project, they are probably, looking back now, some of the most valuable.
On that note, I would like to thank Mr. Harry Samuel for his generous contribution which supported this project. I feel especially fortunate to have been granted this opportunity. Furthermore, I would like to thank everyone at the Arts Internship Office for their much-appreciated work.
