Ecologies of inaudible machines and their silences: Towards sonic thinking as a way of socio-ecological knowing in the Unheard City

Abstract

This article presents the Unheard City project that uses an Android application to read Bluetooth Low Energy and Wi-Fi on an app supported data walk through a Connected Autonomous Vehicle testbed. We listen to the sonic aspects of the devices found and apply sonic thinking to determine the socio-ecologies within the area. Sound, either through reading the signal as data or sonification, is used as a technical practice to think about the layers within the street and their relationships, drawing on situational analytics and mixing qualitative and quantitative frames of interpretation. The emergent relations are derived from the heard signals that enable a view of the heard software that enables us to understand the road as an increasingly subdivided and specialised networks. These networks connect both humans and technical layers in subdivided areas of the road. Silences are explored as either redundant architecture, present but no longer required, or obfuscated data leading to the socio-ecology becoming less public. Constraints, such as measuring emissions levels, on the ecologies are also considered. Future work will consider other protocols and signal types to extend the analysis.

Introduction

Cameras, routers, and machine boxes are the more visible smart city infrastructures at street level. Each communicates with each other at inaudible frequencies to form emergent ecologies with devices that listen to them. Signals containing device and service information are broadcast as intermittent streams of data through which a device ecology periodically appears. I focus on a project that explores these inaudible technologies that re-contextualise mundane items such as traffic lights and cars (Michael, 2000) as both responding to environmental regulations and implicated in applying ecological models. This engagement focuses on a traffic light within an automated vehicle testbed. This approach follows Ned Rossiter’s (2016) focus on infrastructural software as part of logistical media, where media creates logistical possibilities through informational flow, but also raises its histories. Such a focus often uses the visible devices to explore the socio-ecology work to format the humans around them, rather than the emergent signal world. The inaudible world is noisy yet limited to the devices that are arranged to govern the connected road around them.

Algorithmic cities have the potential to make city, and so lived experience, programmable (Kitchin, 2011). As Gavin Smith (2020) notes, algorithms project an intangible presence into the city through devices. Applications like Waze define an algorithmic spatiality (Fisher, 2020) that work with and clatter existing forms of knowledge about space. These point to a thread that points to the presence of large platforms, such as Google, Apple, Facebook, and Amazon (GAFA), underpinning such cities (Douay, 2018; Hlongwa, 2020; Rosol and Blue, 2022). This approach identifies platforms in which services sit but is problematic at a local level where a service interacts with the public through the mundane places of the algorithmic city. These services present layers and devices that are not audibly run by GAFA. Mercedes Bunz and Graham Meikle’s (2017) focus views the Internet of Things as a series of connections that make the world addressable and support automated cognition through processing data. These addressable connections will be argued to subdivide the city and be hard to trace, either from privacy concerns, changes in devices, or infrastructural systems.

Henri Lefebvre’s (2013) rhythmanalysis provides a useful framing for looking at the algorhythmic city (Coletta and Kitchin, 2017) that focuses on data flows between devices and how the algorithmic city affects humans (Lyon, 2020; Zampoukos et al., 2024). This article draws on this to extend it to exploring the wider relationships that are present within these data flows. The approach in this article looks at the relationships between not only between devices and humans but also goes across protocols to explore how assemblages and the relationships or gaps revealed by them. Using listening as data (Droumeva, 2021) and dealing with the materiality of the translated signal supports an exploration of cartography but also relationships being transmitted.

Algorithmic cities raise questions about what is being listened to and by whom (Mattern, 2020). Sensor systems that are used to instrument cities focus on specific tasks, such as ShotSpotter (Sinha, 2023) or fitness trackers. This article draws from these perspectives within the method, but the approach taken here engages with sensor signals in a physical space. The application itself uses multiple sensors to gain different types of data that is contextualised by location information to enable further analysis. Deborah Lupton’s (2016, 2020) exploration of the quantified and data selves using items such as fitness devices, infrared remote controls, and payment through near field communications explore how the signals support creating data for both the user and how this changes social roles and allowed companies to get greater data. The application supported data walk method introduced in this article fits in between these approaches to devices and signals and enables them to be seen as a socio-ecology of devices that may or may not be in operation. In this experiment, the Unheard City (2020) application is used to listen to streams, that contain messages that connect devices to either other devices or people and can be recorded. The data walk process (Powell, 2021) is developed to include a bespoke application as a digital recording participant is developed to engage with public parts of this programmable city. Like data walks, we are looking for digital infrastructure with the app sampling the sound. The practice draws on sensor ethnography, a way of finding data and ethnographic opportunities that are hard to come by otherwise (Nafus, 2018), that uses sensor data to explore a wider experience that is beyond the human to understand the data generated from that relationship. The mobile applications capture certain types of data, such as Bluetooth and Wi-Fi, from which we can develop readings, drawn from radio signals. This creates new requirements for reading and interpreting the quantitative data, along with any contextual documents, such as photographs, to situate the data within the Holyhead Road in our case. The data is used to augment our understanding of the sensed data and to help understand the links between devices and protocols. One of the features of this data is its temporal nature. The city is seen as a space for a multiplicity of programmes and ecologies that perform multiple models that may or may not mesh or overlap either with each other or the public space but that are not a computer (Mattern, 2021).

Through this, I use the idea of the socio-ecology to explore how society and ecology interact as well as how society creates an ecology of digital devices. I consider socio-ecology as relationships created as situated devices interact with each other to create relationships between device and humans (Romano et al., 2021). While mobile devices and networks enable relationships across platforms and area, the situated aspects of these relationships are key to this project where the local environment feeds into the technical observations. The changing situation may affect the inhabitants’ interactions with each other and so the technical system around them (Romano et al., 2021). The technical system explored in this article is not unique. However, it’s responses would vary with differing contexts that become requirements. This raises challenges for studying these ecologies, especially considering the technical aspects that feed into and from social aspects, to understand potential changing relationships and the possibilities for interpretation. Apart from the relationship between devices and people, a key aspect in this ecology is that of refusal to either provide data or data that can be read.

The devices that we are using place us within and part of the computational ecology, in a similar vein to Nuage Vert (Marres, 2013). Marres approaches the project as an articulation of environmental issues, among other possibilities, through its representations in varying situations. Nuage Vert is an application that operates slightly differently for each deployment (Marres, 2013). Both Nuage Vert and Unheard City are site specific and make features of the specific environment visible. Both projects run the same technical processes, but the results will vary for each situation. The dominant model used is the network to understand the socio-ecology, but the ongoing work requires different models to explore the relationships and question the shape of the relationships. We use the applications to develop different models, either as part of a larger model that is created afterwards or embedded within the walk itself. The interpretational aspects affect the representations through adding qualitative aspects to the quantitative ones present in the digital method (Marres, 2020). The situation in this article becomes even more focussed as the signals, through which the devices interact, are created within this location. As shown later, they can be used to interpret the situation but are equally entities that can be interpreted.

This broadly puts this project in dialogue with Noortje Marres’ (2020) notion of situational analytics and mixing qualitative and quantitative methods for interdisciplinary interpretation (Marres and De Rijcke, 2020). The applications discussed work with the situations of both the device, and its interactions to listen to the devices, and the person using it within the context of the road. Building on the conversion of the data through the tool, the data needs to be interpreted. Our tool that listens to the socio-ecological network is itself a socio-ecological system. We are using these in the locale being explored. On some walks, we took audio recordings that contextualise the environment. In one of these, the sounds are overlaid to compare the road noise next to the traffic lights and a sonification of the devices. A representational strategy is used to sonify a model of the signals being sent from this infrastructure against a local audio recording of the street to listen to the interaction. The applications present protocol objects that require both quantitative and qualitative interpretations while recognizing them as technical and social objects.

This article’s contribution is to use sound/sonics as a method of engaging with media-urban ecologies. These experiments reveal relational sonic qualities about socio-ecologies. As well as thinking about the sound ecology, this article considers the materiality of the heard sound as digital data as a way of engaging with the potential relationships that can be heard. The Unheard City project, discussed below, is read as method of interacting with device ecologies in public areas by using app-driven machine listening to create data. It takes a layered approach to the city to consider how the possibilities of the situated socio-ecology within an algorithmic city might become known. In the first part, I introduce and contextualise the concept that I am calling sonic thinking to approach the data captured by the Unheard City application, discussed the second part. The third section focusses on the mapping opportunities of a street from the data. The fourth section looks at the role of silence before turning to environmental infrastructures as a lens for infrastructural history.

Sonic thinking

Listening to sounds can reveal aspects of the environment that were inaccessible. Sonic thinking is beginning to be recognised as a way of listening to audio objects to think with (Bocquillon, 2022) where sound becomes part of the critical method. This article expands this focus to consider the computational aspects of listening. Using a machine to listen augments our senses by engaging with otherwise inaccessible sounds that then requires a different approach to reading the data as a representation of sound. Through this I use sound as a digital object to be both represented and considered, not just as music (Loughridge, 2023), but as data (Droumeva, 2021) to realise connections that can be sonified or otherwise interpreted. Sonic thinking is involved in forms of the heard object and its use in both culture and industry (Tkaczyk, 2023), which I expand to include how the machine used listens to the emergent socio-ecology. The software interprets the signal as a computational object that can be interpreted. I use these approaches to think about not only the infrastructural aspects but to apply humanist and social science aspects to the representation to aid reflection.

Studying a socio-ecology requires a consideration of the relationships between the data. The city itself has been reflected in the context of electrical walks (Kubisch, n.d.) and Shintaro Miyazaki’s (2013) algorhythmics that create a circuit from the available data to focus on the medium aspects. Both render the signals audible, but not their relationships. The city has been heard as an algorhythm (Coletta and Kitchin, 2017), a sonification that reflects the rhythms in the data, that maps a city. In contrast, the Unheard City applications listen to the signals from the device perspective in the street to generate the relational possibilities of a socio-technical environment and the observed connections rather than taking the macro view of a city or representing unheard sounds for further interpretation. Through this, we can see software infrastructures that exist within the visible infrastructure of boxes and traffic lights, where newer infrastructures are following older ones (Mattern, 2021), such as the Green Light Optimised Speed Advisory lights and environmental sensors discussed later. These items provide the possibilities of the city as programmed entity through device timing and the creation of a software defined network that responds to and enables the relationship while understanding the context of the capturing device, to echo Nafus’s (2021) question about context and positionality. The mapping discussion uses this change of audio context to alter the view of the same infrastructure that give rise to the potential for different sounds and interpretations. These raise other infrastructural readings of the protocols that enable the software to work and the regulatory infrastructure that also informs them.

Sonic methods are used to map the places where particular sounds can be heard. Gallagher and Prior (2014) argue that recordings can play a larger part in geography by recording other cultural aspects, such as background noise, that add context. This move to the wider context into which the research is situated. Here I extend the notion of the urban soundscape (Droumeva, 2017; Stirling, 2021) into the inaudible ranges where machines are required to hear the inaudible audio acts that as a marker for possible ecologies but cannot be removed from the audible ecologies. The road sounds may not be divorced from these inaudible sounds where they are used to shape the traffic, providing a potential insight into the algorithmic timings. The second method that can be drawn from is the sonic cartography where the maps are made through recordings (Stirling, 2021; Wilson, 2016), echoing the infrastructural maps that ‘make visible the invisible’ (Mattern, 2021), though in this project’s case, the inaudible is made audible. The project not only hears the visible devices, such as the traffic light boxes, but also connected devices, such as dashboard cameras, that may not appear to connect to a larger network. The connected road, designed for Connected Vehicles, will be shown to be subdivided into islands that enable traffic to move. These become apparent in the digital representations of the recorded sounds that enable further representation, such as sonification.

I use sonification as a representational strategy to explore the data and patterns as well as considering the computational aspects of sound and the silences that are in the data: either as literal silence or a deliberate hiding of detail in the stream. Here we need to understand the silences within heard object. John Cage’s 4’33 enables the listener to attend the sounds around them (Kahn, 1997) rather than focus on a sound source. Salome Voegelin’s (2010) concept of silence focuses on listening to sounds out of context is refigured as silence in data that shows it to have a different context. Wreford Miller’s (1993) concept that silence as a reduction of hearing to a measurement misses silence as a deliberate gap within listening as data. Silence is a deliberate tactic as well as the result of change, moving away from Agostino’s Di Scipio’s (2015) conception of silence being a shared experience. Here I move away from urban soundscapes towards a socio-ecology created by sound. Silence, in a data sense, points to the background noise of the relationships rather than human spatiality (Johansen, 2020) which enables a challenging of the perceived model. Silence here engages the listener to apply different frameworks to test the reasons for it and to enable recontextualization by the listener as a part of sound. A limitation is that we can only detect what is heard and need to augment the process by recording the context. These silences can be understood as a political aspect through the computational aspects. The message being sent is created by the manufacturer, creating a deliberate silence in some places: notably the devices that connect to cars and their connections to the infrastructure. I argue that we need to understand sound as a material object that can be restructured from radio waves to a computational object. We could read the data to understand these missing pieces of data as silences within the sound. These methods allow us to work through Beth Coleman’s question about the social being at stake in a smart world of data acquisition where decisions may be taken out of human control (Coleman, 2018). This will be made more apparent when considering silence in the data.

These same devices create procedures that animate the world to provide new uncertainties to explore (Thrift, 2011). This article both engages with how the world is written through signals and initial ways to make sense of this world. The device that we use hears the signal and converts it for the reader. The required further reading allows for errors in transcription to appear. These changes rest on the signals’ computational materiality – shown through it changing from a radio to signal into a series of digital objects – and our ability to engage with it. Materiality is used to consider the relationships within the data, either to create sonifications, the communication of data using non-speech audio (Kramer et al., 1997), or visualisations. Sonification uses the data and represents them objects so that we can hear the differing protocols as a form of cross-protocol analysis, to alter Richard Rogers’ (2019) cross-platform analysis that seeks to find a wider set of relationships that expose concerns. I use sound to try and think relationally and about the potential relationships that we might find. Sound here is both the heard object, either through the machine or our ears, and the material object, the way that the sound is presented by the machine to us. This latter part touches on concerns made in infrastructure studies regarding technical literacy (Parks and Starosielski, 2015) that inform these readings. The data contain a series of relationships within it that are constrained by both the protocol and device manufacturers that supports sonifying the missing parts within the data. This informs the representational strategy to consider each observed signal as data that can be read.

Sensing the environment and the world in the context of the Holyhead Road relies on using mobile devices to detect classes of radio signals that are to be represented sonically, after being translated. Sonic thinking considers sound as something to hear but that can be recorded either for representation or considered as a medium object. Having defined this signal space, the project is introduced.

The Unheard City project

The Holyhead Road is an entrance point to the city of Coventry: a residential and business area, and polluted road. Historically, the road was a key part of the British automotive industry as evidenced by the Alvis Park; Alvis being the name of a key early car manufacturer (Taylor, 2018). It is a Connected Autonomous Vehicle (CAV) testbed ¹ and contains various technical apparatus that construct the road as a laboratory for the West Midlands Future Transport project. As the site of the early car industry and the testbed for the CAV, the road appears to be perfect juncture to develop its vehicular past into the future. It also a street with rented houses and a transient population that has connected devices. It poses an issue for considering signals within urban frameworks but also nods to the constantly changing nature of such environments and their possibilities.

Although Wi-Fi and Bluetooth signals use radio waves, they broadcast at levels that cannot be heard by humans. To this end, mobile phone applications are developed to augment our hearing into inaudible ranges using the Android Operating System’s programming interfaces. Listening becomes a form of data (Droumeva, 2021) when heard by machines, but this requires the consideration when being processed. The Unheard City project (UnheardCity, 2020) uses bespoke applications to repurpose Android mobile phones to scan for and record Bluetooth and Wi-Fi signals to explore the devices and their ecologies that appear as shown in Figure 1. The translation from the radio into digital streams provides references through which we can begin to listen to patterns, either as a live sonification or later processing. The devices are situated within the environment from which they collect data. From here we can reconstruct the data to show different layers and features. The application records the device’s location every 5 seconds to enable us to situate the data.

Figure 1.

Outline of mobile device recording public signals of connected applications.

One version of the software application was created as an outreach for an engagement meeting that allowed us to listen to how a room changes at the device level while people are using it. Drawing on Kubisch’s (n.d.) electrical signal detection and Miyazaki’s (2013) detector work, these models not only represent heard data but allow us to represent the data itself. This sonification, more towards an audification by being a minimal translation from data to sound, represents the patterns of devices that advertise their presence without analysing that data. This representational strategy allows us to listen to the devices as background noise that is generated by always present devices to create an awareness of the types of patterns that are available, such as the infrastructural devices, while more transient traffic passes. The sonification represents a distant reading that contains contextual elements to situate the listener within the device space in the same way that one might within a network (Emsley et al., 2017). The representational strategy is reliant upon time and the interpreted situation.

The app acts as an interactive sonification device to allow the walker to understand the underlying appearance of devices and to augment this with the other models. It was altered for a data walk to reflect varying models that have been defined as heuristic models. These models are based on a simple set of rules to suggest device contexts and how the hidden device infrastructure is constituted. Others, such as looking at the hidden information such as the manufacturer identifier or a service id: who makes the device that is detected and what does it do? These raise questions about how the data can be identified raising through the data provided by the standards body. The focus on the state of the manufacturer id supported the development of models such as ‘human in the loop’ and ‘human out of the loop’ (Madaan et al., 2018) as a form of ‘device in the loop’ or ‘device out of the loop’. These formulations point to the relationships within the ecology as devices, such as fitness devices or watches, may need to communicate with an app on a phone or tablet as a device in the loop or bypass these as a device out of the loop. These labels may be uncertain as the device out of the loop may be detecting at a point where it is not advertising a connection, so removing these links from human control (Coleman, 2018). Using these hints from the data as it is processed, the app provides some response to the way in which the low power Bluetooth devices use applications on mobile devices to connect and send their data. The device in the loop needs to be considered as potentially being with or on a human, linking the two together. The move to the constitution of the relationship enables a realisation of uncertainty or gaps and how these come about. While simple, the sonification moves us from the identifying sound and background noise before developing models to understand a device place the constitutes an ecology by sensing it while being part of the ecology.

The second application acts as a digital participant in the data walk that both maps and records the signals to understand inaudible side of the city. The phone is a locative device: it knows where you are and can be used to place you within a location using GPS (Frith, 2015). Larissa Hjorth and Gerard Goggins’ (2024) focus on mobile methods show that phones can be used methodologically to find, create, and locate data. This works to situate the phone within a particular space and time and to extend the methodological considerations to include the phone’s ability to capture and store data. Both applications require thinking with sound as a medium object to identify patterns and aspects: the first requires listening to the patterns in the sonifications and the second requires an understanding of the digital data. Sonic thinking needs to attend not only to the noises but the silences and the position where they appear that enables modelling to test the theories of connection across layers of devices that appear through the protocols used.

The process of modelling raises various challenge to reading and thinking about how these devices create ad hoc ecologies based on the software. It enables a rethinking of how sound can be read and to understand how space and relations are digitally altered. Here we need to take time to explore the data that appears, such as any of the names or identifiers, to look at the ecology that underpins these devices. Some are temporally based, such as the Covid beacon architectures that existed within the tracing applications (Grekousis and Liu, 2021), raising questions about the disappearance of ecologies over time. These architectures are defined through mobile applications, such as the NHS Covid Tracing app (Wymant et al., 2021), that are being switched off. These tracked social ecologies of potential exposure to the Covid virus with the model listening for that beacon within the code. We have shifting registers of ecologies that create different relationships with the city and each other. In one part, the application reads the services being broadcast to search for a key that is defined by two companies within the standard. This model relies in listening to personal devices with the challenge of not tracking them. I consider this as an issue for such work: how to deal with a technological forgetting. We might see these as ways of providing governance through sending infection and notices, how it might be altered, and who creates these alterations. We might here turn to a history of digital infrastructures to understand these temporalities and the power within them.

One of the methodological challenges is the identification of the relevant devices. Using a mobile phone to scan the results captures a series of Bluetooth and Wi-Fi signals. The random devices require shaping into a signal using external models – such as the CCTV cameras operated by Coventry City Council – and adapting Richard Rogers’ (2019) notion of cross-platform analysis to connect the layers. Here we use need to read across protocols to understand the relationship between the devices using Bluetooth and Wi-Fi that can be heard. The heard data, such as the changing decibel levels or services offered, needs to be read to begin to understand the recorded devices. This also requires an understanding the given figures to work out the data that was tested using sonification and visualisation. This pattern appears in two locations where the same sort of device occurs. The representation becomes a way of testing this reading and seeing the road divided into the spaces. In addition to this, the project created models of personal and automotive devices that can be filtered out. Although the manufacturer id is part of the protocol, we need to check the catalogue created by the standards provider to check if it exists or has been added.

Having introduced the applications, I turn to the main features to discuss the data that comes from a particular socio-ecology through its mapping of the road.

Mapping infrastructures

In this section, I change focus from the wider methodological challenges and focus on one ecology, Green Light Optimising Advisory Service (GLOSA), identified in a data walk through an autonomous vehicle testbed. This service broadcasts a signal to oncoming application-enabled traffic to inform of the application of the time to the next green light change, thus optimising the traffic flow. The data walk aimed to explore the presence of devices across the layers and to determine the ecology in which they work. In the following maps, the Wi-Fi data is focussed on the devices set up by Coventry City Council and for the Holyhead Road and the Bluetooth LE devices that appear within the area of covered by these Wi-Fi signals. This focus is taken to identify infrastructural assemblages that support the testbed and data capture within the road.

In this first section, the decibel level for each device can be used as a marker to identify how the road is created into zones where things can be heard by other devices. The sonic mapping process used here relies on reading the decibel level as a level of noise that can be geolocated using the device location. A rough space where these sounds can be heard is determined, that becomes a spatial indicator based on signal strength that enables a signal mapping. The observed decibel levels are relatively quiet that leads to the approach to understand that these decibel levels allow the road to be divided by devices. The decibel level can be read as a method for either determining position of a connected device or control by broadcasting messages.

Cross-protocol analysis between Wi-Fi related to either Coventry City Council or the testbed and small Bluetooth devices indicates a socio-ecology. The devices and associated cameras are visible on or near a traffic light, but the sonic space provides the limits to these areas that overlap to suggest relationships between the Bluetooth devices within each Wi-Fi area. The first representation is the sonification of the devices heard, as shown in Figure 2, and the timing based on the devices present. The sonification, In the Presence of Devices using Time (Socio-Ecology, 2024) in audio1.wav, is between the two main Bluetooth collections becoming quieter as their signal fades in the middle and the associated Wi-Fi at either end. The deeper sound is the Wi-Fi system owned by Coventry City Council with the percussion instruments as the Bluetooth signals from the cameras. The rhythms allow us to hear the times between the different device types that suggests different roles, such as the smaller devices advertising themselves more frequently to cars. These are bounded within a permissioned and addressable space (Bunz and Meikle, 2017) where devices broadcast signals. The weak signal strength, read from the data, is set so the two areas of timings do not overlap that point to the devices operating as part of an infrastructure. This infrastructural relationship suggests a layering of network types that operate at different scales to link the site to the wider algorithmic city.

Figure 2.

Map of the Bluetooth LE signal clusters (smaller circles) and their associations to Coventry Wi-Fi networks (larger circles).

The road is subdivided into a network of devices with each owned and operated by another company, creating micro socio-ecologies that are shaped by and centred on different actors and devices within a particular zone. The mapping in Figure 3 shows the zone (Easterling, 2012) in which the apparatus sets the signalled anticipations for the driver and their vehicle within an enclave. The signal space echoes the infrastructural relationships between devices and who may operate them, such as the testbed or the City Council. Listening to the data, in the Situated Presence of Computational Devices and Street Traffic (Socio-Ecology, 2024) and audio2.wav, allows us to hear how local the data is being created and to attend to that context (Loukissas, 2019) or the data setting. The manufacturer of the Bluetooth devices is not discernible and renders part of this ecology silent unlike the personal device ecology that is heavily populated by Samsung and Apple. Two entwined versions of the road exist: a digital one with its own zones and addresses which joins a larger digital city and the physical one that is walked through.

Figure 3.

Signal clusters for Coventry Wi-Fi, represented by signal strength, creating a localised zone of operation.

Sonification can be used to link the inaudible signals to the audible traffic noise to provide context. In so doing, the city soundscape is made wider to include multiple registers, such as the world around the device. An audification (Kramer, 1994) used a short sample of a voice note that recorded traffic overlaid by a rhythmic sonification of the communication device appearing in the sampling. The relationship displays the multiple temporalities in the system with the device’s appearance based on when it advertises itself linked with the flow of traffic passing the light in the road. The silence of the devices contacting the GLOSA-enabled mobile application that would enable further interpretation of this relationship and how often the timing is sent. Attending to the silences raises questions about what a connected device might mean here: is a fully connected car, one that has the correct application? What other platforms are used to make sense of the signal data? The vehicular traffic situates the devices as part of an urban ecology. These rhythms support a reading whether traffic flow is affected with the regular infrastructural broadcast against traffic down the road against the change between stationary cars and ones in motion. Its situated position within the flows heard at street level works against the mapping and its zones.

The sonic cartography and methods are based on the listening devices, their relationships, and the signal strength level. The data suggests that the road is broken up into sections that enable closer readings of an area as the potential networks become smaller and more localised but also layers that use different protocols. A representation using signal strength suggests that the Wi-Fi systems with Coventry in the names divide the road into two parts. Within this, the signal is linked to a sequence of Bluetooth devices that are created in the locale. Here we need to be able to read the materiality of the devices and the signal strengths, having identified the relevant entities. By reading the device on its own, we can begin to see it appearing in localised clusters, emphasised by their reading across temporal and spatial axes. Here we use cross–protocol analysis to explore any potential links and see that a particular device, without manufacturer, is present in two clusters, both of which are recorded as having a particular Bluetooth device. Having removed personal Wi-Fi details, we are able to see the public Wi-Fi points in the same areas where we also view routers. We can listen to the clusters to operationalise the reading of the signal states where the signals are stronger near each Wi-Fi point, suggesting that the road is being separated into separate parts of control.

In these zones, the timings and lights produce a space where both algorithmic and street layers are combined to use behaviour to derive a model. The most visible aspects are the traffic lights and camera networks, either for local council CCTV or the Automatic Number Plate Recognition camera, within the testbed to calculate traffic flow and reinforce speed limits within the confines of the signals and their zone of influence. Computationally, the light broadcasts the difference between the present and the future that is sent out on a regular cycle to regulate driver behaviour, namely, the flow of traffic through that signal as an economy of efficiency (Rossiter, 2016) that is aligned to calculation. Sonic methods identify the spaces where ecologies are created and where the street is subdivided by signals. Yet the same methods suggest that analysing the organisations and wider relationships that animate the world can be more difficult due to the uncertainties within the recorded data.

The idea of the road as a pure programmable entity is a series of overlapping models and layers. Each layer is constrained by aspects such as the legal, physical, and environmental concerns. Interdisciplinary concerns enable a rethinking of the decibel level to move it beyond being volume but to view them showing zones of control and measurement. A new map of the road is made audible. These readings rely on the data being available to show what is available to animate the world through its connections. Silences and missing parts of the data are considered in the next along and the further need to consider multiple aspects and that the stated road may not be fully knowable.

Silences

As sonic thinking focuses on the noise and how it can be used as a pattern, silence points to the lacuna’s that form part of these same patterns. Silences in the data can be read as a way of finding the hidden parts but equally point to the place where infrastructural literacy fails. The recorded data can be queried against a set of lists offered by the Bluetooth standards consortium, but not all fields can be matched. The non-connectedness of the device removes it from being watched or observed, pointing to an issue for the sonic mapping of infrastructure: that we need to think about the silences that are worth exploring. Reading the protocol identifies moments where manufacturers or services are missing. These can be heuristically sonified in either in the sonification app or afterwards to point to the gaps, whether this is the manufacturer or service offered, custom made, or missing rather than using the standard codes. This provides us with an interdisciplinary marker to understand an enforced silence that may be read as a private ecology. Reading the manufacturer data and being able to identify it as a company can be extended into being a reference to a network and to idea that there are further connections.

We are not allowed to understand some parts, but this directs us to a different reading. Hiding or customising the data demonstrates the devices that animate the world, but that also remove ways of query who creates the possibilities of this animation. One way of listening to silence is to hear the gaps that have been previously thought of a way of engaging with infrastructural literacy. The use of custom codes or spaces in the codes rejects the possibility of identification by those outside of the ecology. A noticeable collection of devices that are hard to read are on or near a traffic light. These devices are separate the road into the areas, as discussed above, but the signal data elides the manufacturer data and presents either a custom service or no service at all. Visibly, the rough ecology can be considered through identifying the device and looking at the policies. Sonically, we can determine a media ecology of devices that operate in a defined locale but are unable to discern the relations within them. While personal ecologies, such as gaming or fitness systems, can be viewed, the traffic light boxes provide neither manufacturer nor discoverable service data to understand who is collecting data from the public sphere of the road. The organisation ecology beneath the devices remains opaque. The situated nature of the phone encourages thinking about the world where the devices occur, such as the traffic sounds, and their physical presence on the street while showing that socio-ecology is made private. Coleman’s concerns about devices making relationships can be extending to consider how their manufacturers are creating the possibility for these networks while obfuscating them. While we can see the media devices the operate the logistic, their ecologies that organise and optimise for efficiency are silent. Each ecology works within an increasingly localised section of the road and requires a slight rethinking of mapping in a different register. The GLOSA ecology shows that the road is both subdivided, but the silence demonstrates that the public data is being recorded by a private network. In the next section, silence and sudden noise is used to consider the infrastructural histories on the street.

Environmental infrastructures

Anomalies can be heard in data that identify silent pieces of apparatus by applying different perspectives. One of the aims of GLOSA is that it helps with environmental issues (Radford and George, 2019). Here we consider the analogue environmental sensor apparatus silently sited in the road. The data walks do not find these devices, but we do see them when looking at the street furniture. A series of vials attached to a lamp post at the end of the road collect air samples and sent to a laboratory to test for Nitrogen Dioxide with the data published annually until 2021 in Annual Status Reports. ² An exploratory sonification applies a legal model of 40 micrograms and calculates the difference to highlight the amount of nitrogen dioxide in the sample, but it presents a growth in sensors as the ecology to measure environmental data. The annual status reports ³ have not been updated and, while non-digital, are useful to help read the way that the data is being read. The documentation links this to a private laboratory who provide the results to the customer, or the Council in this case and provides links to the models and regulations that infrastructure the devices and their relationships. A side effect of the sonification is that we hear an unexpected growth in sensors, where a newer set is added. This led to a walk to look for these sensors and found them on a nearby road, ⁴ but only because we were made aware from the sonification and looking at it from infrastructural perspective. The silence for these devices and their service point to a potential history of the environmental services that is beyond the scope of this article. Capturing signals supports a reading of whether devices are active or not with the silence being useful to identify points of further interest.

Sonifying the environmental data using the legal limits creates a technical imaginary of ecology, based on what can be both sensed and calculated without reference to the context. As a qualitative entity, the mapping shows the addition of devices, but not their removal, over time. Data walking and listening provides a very specific context, that may change within the same day. This gestures to an infrastructural history that is constrained by external factors. Legal factors, such as the legislation on air quality, or organisational challenges where platforms are not supported feed into these changes. The change can be inverted into a wider consideration of electronic waste (Gabrys, 2013) of inactive devices that are left in situ and potentially drawing energy (Gabrys, 2024). These considerations would fall into future work and suggests different sonic methods to enable the construction of new interdisciplinary markers (Marres and De Rijcke, 2020) that focus on the relationship between sound and silence in a particular situation. Listening to the silences enables us to not only see how a road is subdivided, but that these subdivisions may be collecting data by private data to feedback into the system, such as the traffic lights, that affect our behaviour.

Conclusion

In this article, I present the Unheard City project that observes Bluetooth Low Energy and Wi-Fi signals on a Connected Autonomous Vehicle testbed and enables the recordings to be read. Sonic thinking is used to both listen to the data as sonifications or the listen the device’s decibels as well as the protocols. Innocuous boxes become points of reticulation with the computational that enable engagement with the idea of a city as a model and its uncertainties. The data allows us to delve into the relationships by listening to rhythms but to also glimpse the platforms behind the ecology. It becomes possible to read the environment as an algorithmic calculation that feeds back into the signals that is divorced from the environment present in the road and to think about what is not measured or no longer measures. While the system may look towards optimisation and efficiency, a closer exploration raises questions about what might not be counted. The situated nature of the device on the road supports a reflection on the presence of physical infrastructural history. However, this work requires quantitative analysis to engage with the signals with qualitative trying to interpret the emerging relationships across multiple layers with the side effect of working with a technical literacy.

The use of mobile devices within the urban landscape enables listening with data. A part of the practice works with the multiple protocols that works with multiple sensors. This requires a cross-protocol listening to multiple sensors and temporalities that enable a change in position and context. From the street view, there are models, such as environmental monitoring and speed control, that overlap. Changing these points to the different ways in which the urban ecology affects the listener.

Silence is often the result of deliberate gaps in the heard data, where fields are sent empty by the device makers. This works in opposition to the use of silence to enhance awareness of other sounds to engage the listener to apply different frameworks to test the reasons for the silence. The infrastructural practice that becomes visible also shows gaps that hide the relationships.

This approach to socio-ecologies raises the need for an interdisciplinary mixed methods approach where data supplies the impetus for further investigation. I focus on silences as they break the flow and require different ways of thinking to understand them. This is more so within the situated framework that I adopted for this article.

A limitation of this approach is that the listening machine is tied to a set of signals captured by the phone, leading us to miss others. Future work will explore software defined radio to explore other bands, but it creates challenges in identifying how and what to listen. Future work will engage with other devices that interact with signals, such as the mobile application and the protocols that are broadcast. Different signal technologies will be explored to expand the devices and wavelengths heard by using software defined radio to listen to the same bandwidth without being filtered by sensors.