news & views COMPUTATIONAL SOCIAL SCIENCE Predicting the decay of collective memory Cultural products have a life of their own: academic papers get cited and songs get downloaded. Surprisingly, public attention to these products shows a consistent pattern over time: a constant decline characterized by an inflexion point. This pattern might be due to how cultural products are discussed in the community and archived as cultural memories. Alin Coman T ake the public attention to the 9/11 terrorist attacks as an example. Days after the attacks, New Yorkers tended to congregate in small groups of close friends to discuss the events and their experiences 1 . They communicated with one another as a way to share information and to cope with the trauma. Following the immediate aftermath of the 9/11 events, makeshift memorials, newspaper articles, books on the topic and formal commemorations served as constant reminders of the terrible tragedy. Both the communicative exchanges that occurred right after the attacks (that is, communicative memories) as well as the artefacts that served as reminders in the wake of the attacks (that is, cultural memories) fundamentally impacted the collective memories of New Yorkers. Writing in Nature Human Behaviour, remarkably, Candia and colleagues 2 now show that there might be a universal pattern of decay of attention that characterizes a wide range of cultural products, from academic papers describing the Higgs boson to Taylor Swift’s ‘Shake it Off’. This universal decay is circumscribed by the two factors illustrated in the 9/11 example: the communicative interactions occurring right after the release of a cultural product and the physical records that are subsequently established. The main contribution of the paper is in taking a big-data approach to exploring the dynamics of collective attention, and by extension, to understanding the formation of collective memories. The topic of the formation of collective memories has been intensely debated among the social sciences 3 . Sociologists tend to conceptualize collective memories as objectified symbols that have a physical manifestation. Textbooks, monuments and memorials represent the instantiation of collective memories in society. Sociologists are mainly interested in understanding the structural and institutional struggles that play out as these objectified symbols are constructed, demolished and reconstructed. These struggles, sociologists contend, reveal the fundamental forces that shape societies’ collective memories 4 . In contrast, psychologists conceptualize collective memories as shared individual memories that bear on people’s identities 3 . For them, the main factors that shape collective memories go beyond the forces that are responsible for creating objectified symbols. Rather, it is the reception and filtering through people’s minds of the information present in the environment (for example, the textbook or monument) that is of interest to psychologists. Candia and colleagues take a big-data approach to understand how collective memories — operationalized as public attention — are influenced by both the conversations that people have with one another as well as by the artefacts that are created following initial communicative interactions. The approach taken by the authors involves collecting data that captures the public attention given to a wide range of cultural products, from academic papers published in a physics journal to popular songs rated on Spotify. For each cultural product, they operationalize public attention in different ways: for papers, how many new citations an academic paper received for a period of time after publication, while for movies, how many views their trailers had on YouTube for a period of time after release. They then fit three mathematical functions to try to approximate the observed pattern of public attention: an exponential decay function, a bi-exponential decay function and a log-normal decay function. Results show that a bi-exponential function (that is, decay of attention interrupted by an inflexion point) constituted the best fit of the data. More surprisingly, this decay pattern describes all the types of cultural products they tested. While previous studies established that one cannot predict whether a particular cultural product would be successful or not on the market 5 , this paper indicates that once it’s out there in the world, we know how the public attention to the product will decay over time: in a bi-exponential fashion. There is tremendous promise in the current approach, but further refinement is certainly warranted. First, as useful as this framework differentiating communicative and cultural memories is, empirical evidence of its explanatory power is just tentative. It is inferred that the observed pattern is produced by forces that involve communicative and cultural memory, but this assumption has not yet been tested. Therefore, more direct evidence of this potential mechanism of collective attention should be sought. Second is the question of universality: how universal is the pattern observed? Papers, patents, songs and movies might show this bi-exponential pattern, but other cultural products that clearly fall under the category of collective memories (for example, the 9/11 terrorist attacks) might predictably show a different pattern. In the case of the 9/11 events, one could imagine that a cyclical pattern indicating renewed communicative processes at every commemoration (that is, sinusoidal decay) Credit: Roy Scott/Ikon Images/Getty NATURE HUMAN BEHAVIOUR | www.nature.com/nathumbehav