A Quiet Revolution? Cinema Programming in the Era of Artificial Intelligence
Roderik Smits / Erasmus University Rotterdam
AI has reached the point that it is radically reshaping creative and industrial practices in the media industries. While algorithm-based recommendations from streaming services are deeply embedded in media practices, we are now witnessing the transformative impact of AI more broadly on processes of development, production, distribution, and consumption.
Film is one of the industries where such developments are playing out in interesting ways. Netflix, Amazon, and other global tech giants are characteristic of the streaming revolution in the past 10 years, using big data intelligence and algorithm technology to inform film programming and recommendation. As a result, a thriving data-driven culture has emerged that promises to deliver long-desired insights into programming and audience demands with greater precision than ever before.
For good reasons, algorithm approaches to film programming and recommendation are strongly associated with major tech giants. To highlight innovation and sophistication, their approaches are often contrasted with conventional approaches. This has also resulted in varying promotional discourses, with players such as Netflix highlighting the benefits of algorithm programming and smaller independent players with smaller catalogues often highlighting human curation.
While algorithm programming and recommendations are common in streaming, the cinema exhibition sector is also adopting algorithm technology for weekly film schedules, moving beyond traditional human-based approaches.
Already in the late-2000s, well before Netflix and Amazon became global streamers, Eliashberg et al. (2009) performed experiments with a (very basic) scheduling algorithm for a multiplex cinema in Amsterdam (the Netherlands), measured over 14 weeks. While this was an early initiative, experiments with more sophisticated algorithm technologies for cinema programming followed. The multi-national cinema operator VUE emerged as a pioneering player in Europe in the mid-2010s, when they started a collaboration with external algorithm developers to build AI powered beta models for cinema programming.
Clearly, this was a trial-and-error process that demanded a financial investment and a strong belief in an emerging technology. To be clear, VUE’s experimentation occurred around the same time that Netflix’s algorithm-driven approach to recommendations was introduced to the mainstream, prompting concerns about the declining role of human programmers and the potential negative effects on media diversity (Frey, 2019).
Also, the algorithm model itself required significant improvement. VUE has reflected in the trade press on the early development of their model from the mid-2010s: “We went through 53 models in beta over two years and it was around the 30th model that we knew we were on to something very special and started to roll it out across more cinemas as our confidence grew with the technology.”
Another pioneer in the market was the small-scale Dutch company Share Dimension, which developed an algorithm for programming films in cinemas, branded as the product Cinema Intelligence. They started a collaboration in 2015 with the major cinema and data analytics company Vista Group International. As a result, Cinema Intelligence became part of Vista’s broader portfolio of services for cinema operators.
Now, almost 10 years later, Cinema Intelligence is described on the Vista website as an “artificial intelligence powered solution for film forecasting, distribution negotiation, automated scheduling, and business analysis.” It is designed to “create a schedule in minutes” and to “enhance booking decisions with “what if” analysis and professional film insights.”
Despite growing attention to AI and algorithms in other sectors, developments in the film exhibition sector have been relatively quiet, especially in recent academic research. Some companies are innovators or early adopters, while others follow later or continue with tried-and-true approaches. However, the rise of algorithm-driven programming can be significant for major cinema operators. Eliashberg et al. (2009) illustrate this by noting that a multiplex cinema of 10 screens must manage the scheduling of hundreds of showings on a weekly basis.
Researching how widely algorithmic cinema programming is adopted in the industry is challenging due to the lack of data. Internet searches for terms like ‘algorithmic cinema programming’ or ‘algorithmic film scheduling’ related to major cinema chains such as AMC, Cineworld, Cinemark, and Pathé don’t generate meaningful results. Only the development of VUE’s algorithm is covered by the trade press.
In my own research, I found out about algorithm-driven programming in 2018 during an interview with VUE’s Head of Programming in the UK, before the trade press reported about the development. The Head of Programming noted that his hiring, in late 2015, was specifically to bring expertise in big data and quantitative analysis. He also noted that by 2018, his programming team at VUE’s UK head office already included three Data Specialists. They worked with eight Content Managers to programme content across 90 VUE cinemas in the country.
Several years later, I conducted further interviews with multi-national cinema operators to explore the development of algorithm-driven programming. Shifting my focus to the Netherlands, I learned that the two largest cinema operators in the country, Pathé with 30 cinemas and VUE with 20 cinemas, use algorithm-driven programming. However, the third-largest cinema operator, Kinepolis with 19 cinemas, continues to work with conventional programming approaches. The following paragraphs provide insights from my interview with VUE Netherlands.
In November 2022, the CEO of VUE Netherlands mentioned that they also use algorithm-driven programming for cinemas across the Netherlands, centrally managed from their head office in the country. The algorithm considers historical and current film performance data, as well as factors like weather forecasts and major cultural or social events that might affect cinema attendance.
He also noted that they began using the algorithm in 2015, and it took several years to get accustomed, improve the technology, and achieve accuracy. The algorithm generates a schedule that human programmers can manually adjust. Naturally, more adjustments were needed in the initial years compared to later on.
A challenging aspect is the declining role of human programmers, often regarded as playing a critical part in the production of film culture. They are now expected to trust and engage with the logic of a largely automated cinema programme. This shift results in less autonomy for them to make decisions and exert creative control over what was previously ‘their’ cinema programme. At VUE, the majority of the work is done by the algorithm, with human programmers primarily tasked with checking and fine-tuning the results.
As AI continues to transform industry practices, I expect algorithm technology to play a larger role in shaping the future of cinema programming. Algorithms could be integrated into the programming process in various ways: in some cases, the algorithm might be very dominant, while in others, it might simply provide a basis for human programmers to work with or be used for scheduling specific types of films.
There are also various ways in which algorithms can be programmed to produce film schedules. Some cinemas attract different audience groups than others. The algorithm can be designed to encourage either more or less diversity in the film selection, as well as the days, times, and screens on which films are shown.
Further, major cinema chains with many screens benefit particularly from algorithm scheduling for hundreds of weekly showings, as the need for automation is higher in complex and labour-intensive situations. However, the context for smaller cinema operators is different, and they might continue with manual programming.
For future research, it is important to follow the development of algorithms in terms of their sophistication. The term ‘algorithm’ has become a catch-all for all sorts of data-driven innovation. There are now highly complex and advanced algorithm models and much more modest algorithm models. It is meaningful to recognise such differences in order to assess how their innovation compares to more conventional software programmes for film scheduling.
Image Credits:
- 1. A robot and cinema programmer working together (author’s personal collection)
- 2. AI and the film industry
- 3. Major player in the market – VUE
- 4. Screenshot of VUE website in the UK (author’s screen grab)
Eliashberg, J., Hegie, Q., Ho, J., Huisman, D., Miller, S. J., Swami, S., Weinberg, C. B., & Wierenga, B. (2009). Demand-driven Scheduling of Movies in a Multiplex. International Journal of Research in Marketing, 26(2), 75-88. https://doi.org/10.1016/j.ijresmar.2008.09.004
Frey, M. (2019). The Internet Suggests: Film, Recommender Systems, and Cultural Mediation. Journal of Cinema and Media Studies, 59(1), 163–170. https://www.jstor.org/stable/26844141