Cultural diversity: weights and measure

Cultural diversity seems to be a very French preoccupation. It regularly punctuates the speeches of our ministers, heads of institutions, and other senior officials, most often at the end of a sentence and preceded by the verb "defend." It's true that France has long sought to protect its cultural industries (cinema mainly) from a situation where they would disappear in favour of productions from elsewhere (across the Atlantic mainly), and that cultural diversity has been, for more than twenty years, an important tool in the pursuit of this objective.

The concept, in its political usage, came about from the fight to exclude culture from the 1994 GATT agreement, which itself was preceded by skirmishes where some decried the protectionist measures of others whilst doing the same at home.¹ France was campaigning then for a "cultural exception", a notion that had little legal basis but managed to gather enough support to its cause. The victory of the GATT, snatched at the last minute, convinced many of the need for a more proactive policy of supporting local cultures in the face of globalisation, which could endanger industries that were essential to the collective imagination. The idea of cultural diversity popped up shortly thereafter as the slightly more attractive sibling of cultural exception. It does have the advantage of referring to a goal to be reached rather than a position to defend, which makes it easier for governments to adopt. They can then get away with a bit of protectionism without being accused of being anti-free market. That's how, in 2005, the UNESCO Convention on the Protection and Promotion of the Diversity of Cultural Expressions managed to use the concept of diversity to justify a commitment to protecting cultural industries in the face of ever-growing trade relations.

Too often reduced to a binary distinction between local and foreign works, cultural diversity can be perceived through a wider lens. Here again, France is at the forefront. The SMAD decree, for example, promotes diversity not only by imposing investment obligations in French-language works but also requires that a portion of these have a budget below a certain amount or be produced by independent companies.² Another example is the "Art et Essai" classification, which helps to ensure a large availability of auteur films in cinemas, thereby contributing to the diversity of supply.

The political application of cultural diversity also provoked, in the 2000s, a surge of interest in its economic counterpart. Numerous researchers and university professors sought to measure it within specific sectors. The Ministry of Culture's statistics department even organized a call for research projects on the topic in 2008. However, researchers had difficulty dealing with the subjectivity that is inherent to the measure of cultural diversity, and the subject now intimidates young scholars, such that it still eludes measurement.

Yet, as Fabien Raynaud highlights in his report on the balances within the audiovisual and film industry (2024), "the success of global video-on-demand platforms, which both recommend and produce content and possess considerable financial resources, reinforces the concern for cultural diversity."³ While the omnipresence of major streaming services in the production and distribution of works has reawakened our instinct to protect cultural diversity, it would be wise that we re-examine its measurement, particularly in light of the technological advancements of the past twenty years.

Measuring cultural diversity in order to better protect it

It would be naive to think that it is possible to measure the definitive degree of cultural diversity within a group of works. Nevertheless, the development of a diversity indicator, however imperfect, would allow for both a more effective protection of the diversity of supply and a better understanding of demand.

Without regulation, the diversity of supply tends to diminish, for both economic and psychological reasons. Offering less can, in fact, sometimes yield greater returns. To take the example of cinemas, multi-screening, which involves programming a film on several screens in the same cinema, mechanically reduces the number of films available in order to better ensure occupancy rates. This practice is perfectly logical from an economic standpoint, yet it harms the diversity of supply. This is why the CNC imposed multi-screening limits on cinema operators as early as 2017.

On video-on-demand, the user interface most often presents a succession of thematic rows populated with thumbnails. However, and this is a common practice across all major platforms, the majority of titles are repeated in several locations on the page. The idea here is to avoid what Alvin Toffler termed "choice overload" as early as 1970, referring to a situation where the consumer is paralysed by having to choose from an excessively wide array of options. The diversity of supply is therefore limited in order to encourage the user to make a decision. Video-on-demand in fact creates two types of supply, which are more or less diversified: the whole catalogue, which the user theoretically has access to, and the titles that are actually suggested on-screen, meaning the ones that are more easily discoverable by the user.

In French regulation, a distinction is often made between so-called "generalist" and "thematic" services, the idea being to impose investment and broadcasting obligations on each in accordance with its market positioning. An analysis of the cultural diversity in each service's content offer would undoubtedly facilitate their categorisation. A service that offers a highly diversified selection of works would immediately be labelled as generalist, while classification as a thematic service would be reserved for those whose content offer falls below a certain threshold.

On the demand side, the degree of diversity sought by viewers, listeners, readers, and other "consumers" of cultural goods has long been a subject of debate. The long tail hypothesis, which suggests that an abundance of supply contributes to the formation of niches whose aggregate popularity can exceed that of mainstream hits, has never been proven. It did convey the hope, for a short while, of a more balanced distribution of demand across a greater number of works, all made possible by dematerialisation and its promise of near-unlimited access to culture. However, while the supply has indeed increased tenfold on the internet, the demand curve has not shown a tail as long as one could have hoped.

However, before resigning ourselves to the idea that the diversity of supply does not influence the diversity of demand, it is important to distinguish, as earlier, between content that is available and content that is discoverable. The prescriptive role of the broadcaster should not be underestimated, whether in a dematerialised context (the Netflix homepage, for example) or in real life (in cinemas, for example). This is the argument that Olivier Henrard put forward in his speech at Series Mania:

"The current situation is that, on a European scale, 60% of works viewed are American (in movie theaters as well as on streaming platforms). That is not satisfactory for cultural diversity, but this trend is not inevitable: when a quality local offering is provided, the viewership follows."⁴

How does one measure cultural diversity?

Diversity in the broader sense, before we even get to culture, is rather resistant to objectification. Social science researchers will attest that measuring it is an exercise fraught with difficulties. The diversity of products on a supermarket shelf, for example, can be a matter of subjective appreciation. Consider two shelves offering the same number of different products: on the first, a tube of toothpaste sits alongside sponges and a carton of milk; on the second, there is dish soap, bananas, and cheese. Which of these shelves offers a greater diversity of products? The answer eludes us because it requires assigning a relationship of varying strength to each pair of products, a kind of distance that separates them according to their respective characteristics. Thus we come to wonder whether toothpaste is closer to sponges than to milk, or whether bananas share more in common with dish soap than with cheese. In theory, the aggregation of all the distances that separate each of our products should form more or less compact clusters, the density of which would indicate their degree of diversity.

However, culture cannot be apprehended in the same way as a supermarket shelf – that is precisely why it was excluded from the GATT agreements. The distance separating two artworks is more difficult to estimate than that separating toothpaste from sponges because the objective characteristics of the artworks are not sufficient to describe them fully. This would require taking into account the impression the work has on its audience, the context of its creation, its aesthetic qualities, and so on. Nevertheless, numerous research projects have attempted to measure the degree of similarity⁵ between several artworks by using various "metadata" such as the countries and years of production, the languages spoken, the authors, or even the popularity expressed in audience share, number of views, sales volume, etc. None of these attempts has succeeded in proposing a convincing method for calculating cultural distance. Moreover, a method based solely on subjective assessments appears difficult to apply to large numbers of titles.

Caught between the inadequacy of metadata and the impracticality of a purely subjective approach, cultural diversity may well remain without a means of measurement. However, before giving up, there might be a third path we could explore. Supervised machine learning is a technology that involves training an artificial intelligence model on data annotated by humans. It therefore allows for a significant degree of subjectivity, while promising applicability to a wide range of situations. We hypothesise – perhaps naively, but nothing ventured, nothing gained – that a machine learning model trained on human judgements of the type "A is culturally closer to B than C" could estimate the cultural distance between two works, provided it has sufficient training data. More concretely, we suggest building a tool that, thanks to a significant number of metadata and contextual information on the one hand, and a large volume of human assessments on the other, would autonomously evaluate the cultural disparity within a body of work. A model that knows, thanks to human input, that Wings of Desire is culturally closer to Cléo from 5 to 7 than to Terminator (due to their shared belonging to European "New Wave" movements) should be able to estimate on its own that Pierrot le fou is closer to Fitzcarraldo than to Spiderman.

Furthermore, machine learning offers the advantage of being endlessly adjustable. Adding more training data or weighting its judgements by according greater importance to one variable or another, would ensure the relevance of our model and its adaptability to new situations.

Conclusion

For economists interested in this unique subject that is culture, measuring cultural diversity offers clear benefits, notably a better understanding of the supply and demand for these "products" that are not quite products in the traditional sense.

Politically, the concept of cultural diversity is less prominent today than it was in the 2000s. However, it remains highly relevant. On the one hand, the trade war initiated by Donald Trump suggests a period of isolationism, the opposite of the liberalism of its origins. The circulation of works could be affected in one way or another. Because if certain governments decide, as has been mentioned in recent days, to sacrifice the protection of their cultural industries in exchange for the lifting of tariffs, it is the very foundation of cultural diversity that could be called into question. On the other hand, the pressures exerted by certain American majors and their representatives against the investment and broadcasting obligations imposed on them in Europe compel us to re-evaluate our commitments to diversity, either to renew them or to reduce them. Wherever one stands on with respect to those cultural policies, the task of measuring cultural diversity should be tried once more, if only to bring some clarity to troubled times.