Anna Ådahl



Predicting crowds. The aesthetics and politics of today´s digitised crowds and crowd simulations.


On a daily basis, billions of people from different social and economic backgrounds circulate in urban, supervised infrastructures, acting together in a coordinated manner. These synchronised, system-operated crowds initially emerged alongside industrialism, with machines and technologies being developed to accelerate our production modes and systematise our existence. 


Today these crowds also operate in the virtual realm when surfing on their smart devices. Online they become atomised and fragmented, a disembodied yet connected crowd. 


The systematisation and the digitisation of the surveillance and tracking of crowds have been even more intensified with our online existence, where mapping and the ‘quantification of the self[1]’ have enhanced the potentiality of predicting and monitoring our collective behaviour.


The dissimulation by manufacturers of the complexity of these technologies and their networked infrastructure positions us on a superficial level which makes it difficult to grasp how these systems ‘operate’ us and hence how they map and archive our behaviour into big data.


These new technologies use algorithms to think us mathematically as crowds. 

Following this. how does an algorithmically-operated body and crowd articulate itself?


To access a visualisation of a collective programmed body, this research turned to crowd simulation software (e.g. Golaem and Massive) as it provides a computer-generated image of a systematised/-organised crowd. Used for managing our collective movement or illustrating massive crowd formations in films (a phenomenon which emerged when the cost of extras to perform these crowds became too expensive), these crowd simulations are a digital visualisation of a crowd composed of mathematically-programmed digital agents/corpuses proposing a choreography of mute bodies, programmed intelligently using AI and multi-agent systems to act collectively according to each other and the given environment.


These simulations predominantly represent people on the periphery, populating a milieu, or instrumental bodies in crowd management strategies, shaped as an entity seen from afar. Today the agents constituting these crowds are of limited capacity and aesthetics with simplified gestures and physical characteristics for a better homogenous appearance, a background crowd which should not interfere with the main characters acting in the forefront.



These simulated coordinated crowds moving as a homogenous mass (as seen also in tutorials and show reels online) can be seen as a faceless bound materia, which links back to the early theory of the crowd as a mob in which individuality is lost. A mob aka an unruly crowd disturbing the authority in place: this image of a crowd as a revolutionary mob is violent and is what is presented in most of the imagery produced today in crowd simulations. Autonomous digital crowd agents essentially emerged in creating a warrior, a destructive force. An interesting example is that the software Massive created by Peter Jackson, to enable the visualisation of epic battle scenes in the Lord of the Rings film series, was later adapted for crowd management and safety, the crowd's identity being inverted to mimic the crowd as population, fragile and civilian.


When managing crowds and in the programming of crowd simulations, the notion of flow is recurrent. The behaviour of the crowd needs to be fluid to ideally operate in the given environment and act homogenously as an entity. The agents are hence programmed to avoid collision, not only in terms of objects but also each other.  

The vocabulary and notion of flow is also central in our current economics as well as in today’s flows of human refugees. Can the joint use of the notion and vocabulary of flow in these different fields inform us of any common politics? 


From the early crowd formations during the October celebrations to today´s crowd simulations, the systematised organisation of the physical crowd into seductive configurations transpires a gesture of power and control over the crowd members.

It is important to note here that the characteristics of these mathematically-calculated crowd simulations present a totalitarian aesthetic.


The philosophical standpoint in this research is therefore that an organised, supervised crowd is per definition political. Thus when creating and programming a crowd, even if it is a simulation or 'mise en scène' it becomes a political act or image. 


Through the following questions the research will address the above concerns of the politics of crowds and the political crowd through various mediums and methods:


How can new technologies and artistic processes provide new modes of visualisation and critical reflection on the aesthetics and politics of today’s digitised crowds?


How do we understand our embodied selves through these algorithmically-programmed crowd simulations and digitally-supervised crowds?


What image and patterns/gestures can an intelligent computer-generated visualisation propose as a prediction of our collective/crowd behaviour?


This research addresses these questions in the context of crowd simulations and today’s intelligent computational tool operating us as crowds, as a rapidly growing phenomenon coupled with the constant accelerated growth of the world’s population, overcrowding our cities and affecting our collective condition of existence.

Through practice-based research projects, I study and analyse the aesthetics and politics of virtual crowds through their various representations, gestures/media and technologies. Furthermore, I explore the ways in which the pre-programmed default settings impact upon the aesthetics and politics of modelling crowd behaviours.


This study also challenges the traditional depiction of the crowds we can identify in the social sciences as well as in the arts according to which the crowd is received as one unified entity—a representation of the crowd that is drawn from afar, and thus distances the disinterested spectator from those who are part of a crowd. 


Drawing upon Walter Benjamin’s perception of the crowd as an assembly of singularities, my analysis aims to situate itself within these digitised simulated crowds, that is, from the virtual agent/crowd subject’s perspective. From this immanent standpoint, I intend to critically analyse the prevailing aesthetic representation of crowd simulation as a swarm or homogenous mass, a perspective linked to the notion proposed by Le Bon and Freud of the crowd as a mob (as mentioned above). At the same time, I will look at how the mathematically/algorithmically thought body and crowd, both in simulations and surveillance systems, can impact the prediction and modelling/monitoring of future crowd behaviours. 


The practice-based projects are key in the progression and development of this research and have resulted in artworks in various formats. For example, the film Default Character(2016) and the three-channel triptych video installation and film Di-Simulated Crowds(2018) address the political and aesthetic role of default settings within crowd simulation software as well as the aesthetics of today’s multi-target tracking systems and harvesting of data on our online and urban behaviour. 


Another angle of approach, which articulated itself in a performative analytical investigation, used the human body as a tool and reference in staging and enacting digitally simulated crowds. This analysis resulted in a live choreographed performance,And or Or, (2018) (see portfolio), in which a group of dancers enacted the behavioural characteristics and patterns of digital crowd agents. The performing physical bodies embodied, materialised and visualised the simulated crowd in a physical environment (the first performance took place on the 4thof May this year at Marabouparken Art Gallery in Stockholm). The purpose here was to provide insight into/understanding of how the characteristics of a digital crowd simulation differ from a physical crowd and how the simplified characteristics of the default agents propose a standardisation and limited capacity of the crowd agent, as well as to offer an articulated physical understanding of the virtual crowd and to raise greater awareness of the full complexity of our identification with our evolving embodied self in relation to the virtual crowd agent. 


Within the framework of a collaboration with the gaming company Ubisoft, as well as the MIT Digital Lab linked to Linköping University (Sweden), a project will be developed which will use the technologies of virtual reality and crowd simulation software, in order to create an immersive virtual environment populated by interactive intelligent virtual agents. 

The multiple and autonomous interactions between intelligent agents as well as human-agent interactions can provide yet another understanding of the politics involved in the interaction between the organic body and the computer-generated body. This does however pose a technical challenge and requires heavy computational technologies and professional programming skills.


To further investigate the new computational technologies and intelligences such as deep/machine learning, where a computer system can learn by itself, a ,yet hypothetical, project will be developed. A group of programmed crowd agents will be produced who, through their interaction, will learn and develop a common behaviour and intelligence. How will their behaviour evolve when learning from each other? What choreographic patterns will they perform?

The project will be monitored and recorded at different stages of development to see how the collective behaviour of the agents evolves.


Another project, within a similar technological field, will further explore the computer’s capacity to generate new knowledge through machine learning. A computational system will be given data of crowd behaviour and surveillance images and implicitly produce images and sounds, which will represent the computer’s understanding/learning of our crowd behaviour. In other words, this project will look at the computer’s inherent capacity to predict crowd behaviours.


The critical underlying drive of this research is to raise awareness that we belong to a collective, an atomised crowd bound together through the network of our smart devices and the data which is built on us. We are increasingly part of a crowd navigating an invisible web of intelligent multi-agent systems, woven together through our mobile devices and the continuous archiving of our meta and big data, which, in turn, is used to map and capitalise on us, and to predict our collective behaviour.


The infrastructures and devices that operate us as crowds apply new computational technologies, which, in a globalised, online and rapidly-evolving era, have become less and less cognisable.


Fredric Jameson argues that there is a gap between our local experience, the proximities in which we operate and the structural conditions of the system as a whole and that we navigate a world we do not apprehend, creating a sense of alienation, a phenomenon which increases the possibility and rise of polarisation and division, reinforced by our atomised existence online.


A need to grasp and visualise these computational systems that operate our lives has become more and more urgent in our online and 24/7 existence and production modes.


The aim of this research is thus to study and analyse the computational tools of representation and supervision of today’s organised crowds to identify their political agenda and the impact they will have on us and our future collective behaviour/co-existence. 


As well as to identify the political aspect of programming a crowd. How the mathematically standardised behaviours proposed by the digital simulations in concordance with the default settings we use when operating softwares or surfing online, can influence us politically in how we behave within collective structures, act towards each other and potentially alienate us to ourselves and others.



[1]Quantification of the self’ is the possibility, through the apps available on our mobile devices, to upload and track our bodies organic performances, such as daily walked steps, sleep modes and so forth.