Store, interface, package, connection - Methods and propositions for multi-situated app studies - Uni Siegen
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Store, interface, package, connection
Methods and propositions for multi-situated app studies
Michael Dieter, Carolin Gerlitz, Anne Helmond, Nathaniel Tkacz,
Fernando van der Vlist, Esther Weltevrede
WORKING PAPER SERIES | NO. 4 | AUGUST 2018
Collaborative Research Center 1187 Media of Cooperation
Sonderforschungsbereich 1187 Medien der Kooperation
Working Paper Series
Collaborative Research Center 1187 Media of Cooperation
Print-ISSN 2567 – 2509
Online-ISSN 2567 – 2517
URN urn:nbn:de:hbz:467-13410
This work is licensed under the Creative Publication of the series is funded by the German Re-
Commons Attribution-NonCommercial- search Foundation (DFG).
NoDerivatives 4.0 International License.
The cover image is a modification of fig. 4 in this paper
This Working Paper Series is edited by the Collabora- ›Companies behind the network connections coming
tive Research Center Media of Cooperation and serves in or out of four popular dating apps (OkCupid, Grindr,
as a platform to circulate work in progress or preprints Tinder, BeeTalk)‹, designed and re-arranged by the au-
in order to encourage the exchange of ideas. Please thors.
contact the authors if you have any questions or com-
ments. Copyright remains with the authors. Universität Siegen
SFB 1187 Medien der Kooperation
The Working Papers are accessible via the website Herrengarten 3
http://wp-series.mediacoop.uni-siegen.de or can be 57072 Siegen, Germany
ordered in print by sending an email to workingpa www.sfb1187.uni-siegen.de
perseries@sfb1187.uni-siegen.de workingpaperseries@sfb1187.uni-siegen.de
Store, interface, package, connection
Methods and propositions for multi-situated app studies
Michael Dieter 0000-0002-5894-7892 @mdieter
Carolin Gerlitz 0000-0002-5716-9355 @cgrltz
Anne Helmond 0000-0003-4327-4012 @silvertje
Nathaniel Tkacz 0000-0002-5796-4105 @__nate__
Fernando van der Vlist 0000-0003-1401-0325 @fvandervlist
Esther Weltevrede 0000-0001-5276-297X @esthr
Abstract This paper discusses methodological approaches to
app studies, focussing on their embeddedness and situatedness
within multiple infrastructural settings. Our approach arises by
paying close attention to the multivalent affordances of apps
as software packages, particularly their capacity to enter into
diverse groupings and relations depending on different infra-
structural situations. The changing situations they evoke and
participate in, accordingly, makes apps visible and accountable
in a variety of unique ways. Engaging with and even staging these
situations, therefore, allows for political-economic, social and
cultural dynamics associated with apps and their infrastructures
can be investigated through a style of research we describe as
multi-situated app studies. The piece offers an overview of four
different entry points of enquiry that are exemplary of this over-
arching framework, focussing on app stores, app interfaces, app
packages and app connections. We conclude with nine proposi-
tions that develop out of these studies as prompts for further
research.
Keywords app studies, mobile apps, situatedness, app stores,
software studies, infrastructure
Introduction: A decade of app economies mational commodities from their inception (Daubs
& Manzerolle, 2015; Morris & Elkins, 2015; Nieborg,
July 10th, 2018 marked the ten-year anniversary of 2015). Today, mobile apps have become significant
Apple’s App Store – a significant milestone in the cultural and economic forms (Miller & Matviyenko,
political economy of software cultures. From the be- 2014). As of May 2018, Google and Apple own the lead-
ginning, the App Store contained not only official or ing app stores worldwide, containing over 3.8 million
‘first-party’ applications developed by Apple, but also Android apps and 2 million iOS apps that generate
apps published by third-party developers. Shortly af- over 86 billion USD in revenue. With the average user
ter, Google launched Android Market on October 22, spending almost 1.5 months using apps per year, apps
2008 – which was later rebranded as Google Play – have become deeply embedded in our everyday lives
and similarly made apps for the Android operating (App Annie, 2018). Apps are ‘mundane software’, not
system available. Contrary to the web, which was only because they support everyday practices, but
originally imagined as a shared information space because they insinuate themselves into our routines
(Berners-Lee, 1996) and only later turned into a and habits (Morris & Elkins, 2015). Most importantly,
commodified space, apps were conceived as infor- apps pose several empirical challenges for media
2 CRC Media of Cooperation Working Paper Series No. 4 August 2018 research because of this tendency to move into the in general highly controllable due to its logical com- background, while still remaining thoroughly en- position as statements and data structures (Hui, 2016, tangled with data-intensive infrastructures and new p. 56). Exploring the infrastructural situatedness of economic models like platforms. apps in different conditions or states can, accordingly, Apps are designed to perform as concrete software give rise to new possibilities for accountability and objects, yet are continually transformed through visibility. Even while encountering at times forms of interaction with diverse socio-technical situations. resistance and constraints (as we will go on to dis- From app stores to our personal devices, apps are cuss), it is possible as a result to tease out different transmitted as packages through seamless, highly forms, values and relations that apps can take by automated downloading and purchase procedures, staging experimental and exploratory situations. through the organised market layouts of stores to In this paper we call this a multi-situated approach icon grids on our devices. For users, they seem like to apps, where each methodological orientation or fixed objects; we drag apps around, bundle them ‘entry point’ at the same time deploys and makes vis- into folders or individually activate them by pressing ible different infrastructural settings. Situatedness, their bounded icons. However, the notion of apps as in our understanding, includes the common under- entirely self-contained also belies their involvement standing as ‘the involvement of the researcher within in the data-flows of multi-sided platforms, and their a research site’ (Vannini, 2008, p. 815), but also in- necessary entanglement with varying hardware de- dicates the standardized, yet differentiated techni- vices and digital infrastructures that make their op- cal agencies at work when using apps. In what fol- erations at once possible and, indeed, valuable. lows we present four different methodological entry The bounded appearance of apps is achieved points through which researchers can actively invoke through the specific ways their identity is regulated different app situations and use these to advance or by software engineering. In the terms of Hui (2016), initiate an inquiry. This may entail performative us- apps are ‘new industrial objects’ consisting of logical age of an app through purposefully orchestrated, statements and structures whose associated milieux personalised user experiences, but might also using includes algorithms, databases and network proto- data flows originally designed for machine read- cols. The file format of the app – the ‘package’ – de- ing. The selection of entry points draws on key sites livers an abstract coherence that nevertheless can be within which many stakeholders engage with apps, decompiled, recombined and reassembled in differ- yet renders them in specific ways to create research ent ways. App packages can be understood on verti- situations, taking inspiration from but also noting cal scales that might include metadata, folder hierar- the challenges this poses for software and platform chies and nested information, but also on horizontal research. The final part of the paper summarizes and scales by relations such HT TP network connections, reflects on these challenges in the form of nine propo- trackers or platform integrations. The discreteness of sitions for situated app studies. an app is a technical achievement woven throughout the entire object, but one that also supports a certain kind of multivalence as it enables its stakeholders Methodological entry points for app studies including app stores, developers, partners or users, among others, to integrate and valorise it in mul- App stores tiple, simultaneous situations. In other words, apps have built-in tendencies to be situated and to situate A key entry point to study the situatedness of apps are themselves within different operative situations: they in stores like Google Play and Apple’s App Store, as are made available in particular ways that follow the well as country-specific and device-specific app stores. principle of extensibility in software development, App stores are the main site for accessing, download- while in turn rendering infrastructures, sensors or ing and distributing apps, and they allow research- networks available for themselves. ers several opportunities to follow the perspectives When we deal with the technical situatedness of different stakeholder groups, including users of apps, it is not just a question of objects or sites and developers. App stores, in this way, function as of research, but concretised systems of relations in key gatekeepers – or indeed, as ‘obligatory passage standardised infrastructures. One might draw a con- points’ (Callon, 1984; Fagerjord, 2015) – by setting up trast, in this respect, with ethnographic approaches the rules for app creation, sorting, and distribution, to global commodities that aim to ‘follow the thing’ and do so by drawing from the economic model of across multiple sites or locales (Marcus, 1995). Since the multi-sided marketplace (Rochet & Tirole, 2006) apps immediately exist as digital objects within tech- to ensure app exposure to potential customers. In or- nical milieux, it is less a case of following an app, than der to account for the research affordances (Welte- re-situating it drawing from a number of unique af- vrede, 2016) of app stores, we need to understand fordances are available to the researcher. Indeed, the specific situations that app stores create for apps. the kinds of reciprocal causality initiated by apps is Indeed, they allow for a multi-dimensional perspec-
Dieter, Gerlitz, Helmond, Tkacz, van der Vlist, Weltevrede · Store, interface, package, connection 3
tive on the relations between apps as organised by stores allow for multiple ways of querying or explor-
the assemblage of stores, their algorithms, rankings, ing the offered app spaces: for instance, by search-
developer and user activities. App stores can, accord- ing for specific apps (e. g., [Facebook]), topics (e. g.,
ingly, be used for addressing relations among apps [pregnancy]), genres (e. g., [messenger]), by brows-
as issue spaces, for platform studies of developer en- ing app categories (e. g., games), ranked lists (e. g.,
gagement and innovation, media concentration, and ‘Top Charts’), or featured lists (e. g., ‘Editors' Choice’).
the conditions of possibility for practices. App stores offer both algorithmic and curational or-
It is important to note that while the two dominant dering practices, introducing research affordances for
app stores are the main focus point of this paper, there exploring app collections based on the demarcation
is an abundance of app stores and repositories. There of the store, and following possible user pathways.
are manufacturer-specific stores (e. g., Samsung Gal- Users are also presented with additional app group-
axy Apps, BlackBerry World), country-specific stores ings on individual app pages. For example, Google
(e. g., Yandex.Store in Russia, Tencent App Store in Play has personalised recommendations (‘Recom-
China), stores by telecom operators, and dedicated mended for you’, ‘You Might Also Like’) and com-
open source and adult stores.1 While app stores are plementary app recommendations (‘Related to this
often considered as a relatively new phenomenon, app’), but it also suggests recommendations based on
they have been around for much longer in the form of topics associated with apps (‘Similar Apps’). In con-
software libraries, distribution platforms, reposito- trast, Apple’s App Store has a section ‘More By This
ries, game stores, package managers, and so forth (cf., Developer’ and ‘You May Also Like’ (technically spec-
Morris & Elkins, 2015). The proliferation of app stores ified as ‘/customers-also-bought’). A key challenge
should be seen in a longer history of software frag- for working with app relatedness are the personalisa-
mentation for different devices, operating systems, tion and localisation effects of app stores – similar to
and world regions, leading to the creation of distinct search engine research – which can determine their
archival databases, platforms, and marketplaces (cf., results based on location, country, language, but also
Basole & Karla, 2011). The open source Android oper- previous user behaviour. Central to, but also beyond,
ating system, for instance, allows third-party devel- the algorithmic ordering of apps are the categories to
opers to build their own alternative, non-official app which developers can assign their apps, which then
stores for Android applications. The multiplication inform similarity calculations, but also the users’ en-
and relevance of app stores has led to the development gagement with apps by browsing these categories.
of various third-party market insights companies like With these various sorting and ordering mechanisms,
App Annie, which aggregates data on app stores, app apps can be understood in not just a single bounded
usage and markets, as well as historical data on app form, but as a digital object that can be diversely situ-
store search results and the most downloaded apps ated, connected to and related with through different
over time. Such companies also offer general guid- queries, topics, categories, developer categorisation,
ance for market research and app store optimization as well as user behaviour.
services and, therefore, provide relevant sources for App stores enable researchers to draw on, but also
considering how app developers write themselves reconfigure these various set-making capacities of
into the calculative processes of stores by strategically app stores. They can be employed as indices of apps
selecting categories, drafting descriptions, and pre- that may be queried for keywords, genres or devel-
senting apps to make them store-ready. These third- oper names, allowing the device to demarcate the
party indices offer alternative, often aggregated ac- sample. One can consider which results are returned
cess points to app stores by placing apps and stores in and where they are ranked across ‘spheres’, per query,
distinct commercial contexts. per country, and per store (Rogers, 2013, p. 118). As
App stores themselves come with different access indices of apps, app stores typically provide indi-
points as some offer web interfaces (e. g., Google vidual pages with details about app titles, function-
Play), while others have limited web functionality ality, version, developer, screenshots, descriptions,
(e. g., App Store), or only have device-based mobile permissions requested, download statistics, reviews,
interfaces (e. g., Yandex.Store). Additionally, most and ratings. Using the DMI Google Play Similar Apps
app stores do not offer systematic access to their data and iTunes Store tools,34 we are able to extract the
via standard A PIs, with Apple’s App Store being one details of individual apps, collect ‘Similar Apps’, and
exception.2 This means there are varying capacities extract their details as well. This similar app data can
for systematic search and data extraction (e. g., via then be used to identify networks of app relations as
A PI calls, web scraping, and manual retrieval). App created by the store.
1 https://wiki.digitalmethods.net/Dmi/WinterSchool
2017AppStoresAsDataInfrastructure 3 https://wiki.digitalmethods.net/Dmi/ToolGooglePlay
2 https://affiliate.itunes.apple.com/resources/documen Similar
tation/itunes-store-web-service-search-api/ 4 https://wiki.digitalmethods.net/Dmi/ToolItunesStore4 CRC Media of Cooperation Working Paper Series No. 4 August 2018
Queries can be deployed to engage with collec- sible to repeatedly collect or log search results and
tions of apps as opposed to focussing on single apps their rankings as they unfold over time (cf., Rogers,
in isolation. First, stores can be used to generate the- 2013). A comparative study of ranking volatility for
matic and issue-oriented collections which can be selected queries across Google Play and Apple’s App
studied as expressions or indicators of cultural dif- Store can yield important insights into their ranking
ference. For example, how are apps employed in the mechanisms, issues related to media concentration,
performance of religion (e. g., Buddhism, Christian- and into the curation and removal of certain apps by
ity, Hinduism, Islam, and Judaism)?5 What solutions app stores.8 Moreover, such approaches do not only
do app stores – as indices of apps – generate or recom- address ranking algorithms, but also ‘ranking cul-
mend when querying for various controversial, par- tures’ – highlighting the ‘distributed and heteroge-
tisan, objectionable content, or sensitive issues (e. g., neous agencies that converge’ in ranked lists (Rieder
abortion, gun control, porn, or mental disorders and et al., 2018).
conditions)?6 The latter, in particular, can be under- The research affordances for collection making
stood as an application of issue mapping (Marres, and indexing apps through app stores opens possi-
2015) from within the organisational logic and struc- bilities to generate research situations that provide
tures of the app stores themselves. insights into the relational and infrastructural situ-
Second, the same approach can be used to explore atedness of apps. As outlined above, app stores allow
different genres of apps and the practices they enable for the identification of how topics, issues or genres
(e. g., health and messaging apps). Here, app titles are addressed through apps, or what app developers
and descriptions offer further and more detailed af- offer as packaged solutions, which practices apps
fordances to engage with the question what features support and how app stores relate these apps to one
apps offer, as they usually provide insights into their another. Interestingly, engaging with apps through
key functionality and interoperability with other app stores resurfaces the role of practices as app store
apps, platforms or infrastructures. How do different algorithms not only sort and organise content, but
secure messaging apps position themselves in their also modes of engagement and user practices (with
self-descriptions and to what extent – and in what regards to topical concerns, with regards to social
terms – do they address issues around security, en- media) as apps are designed to structure behaviour
cryption, and usability?7 App descriptions can be and not meaning.
used as starting point for manual categorisation, for
topic modelling through natural language processing
or be queried for predefined terms and phrases. App interfaces
Third, a large proportion of apps are not created as
standalone objects (only), but are built on top of, or While app stores support research into the relations
in relation, to other apps, software or platforms, for between apps and their economisation, app inter-
instance by drawing on the application programming faces offer entry points for specific inquiries into the
interfaces (A PIs) of platforms for data extraction/ conditions of possibilities for user practices. It has
input or offering support practices for platform en- been a longstanding claim of science and technology
gagement. As platforms explicitly invite and facilitate studies (STS) of human-computer interaction (HCI)
developer engagement with their functionalities and that shaping the user is a central concern of interface
data (Bodle, 2011), app stores can be used to explore design (Woolgar, 1990), particularly through forms
how developers have built on top of platforms and of embedded and enacted scripting (Akrich, 1992;
intensify, support, interpret, alter, or amend their Suchman, 2007). This kind of technical scripting
features, data, and associated practices (Gerlitz et al., raises questions concerning the circulation of power,
2016). subjectivation, and the production of value that are
But the sorting processes of apps can also be made especially pertinent under conditions of platformi-
subject to empirical enquiry themselves, opening up sation and the data-intensive forms of ‘controlled
questions around how ‘app relatedness’ is produced consumption’ that apps facilitate (Andersen & Pold,
in the first place and how the algorithmic politics 2018). The walkthrough method assists with explor-
compare across the different app stores or change ing these questions by systematically documenting
over time. Although algorithmic processes are gen- and abstracting interface features in their normative
erally difficult to account for and interpret, it is pos- infrastructural settings.
The walkthrough method is commonly used in
5 https://wiki.digitalmethods.net/Dmi/SummerSchool software engineering and user-centred design re-
2015DigitalMethodsAppAnalysis search to present a software product to peers or
6 https://wiki.digitalmethods.net/Dmi/SummerSchool
2018AppStoresBiasObjectionableQueries
7 https://wiki.digitalmethods.net/Dmi/SummerSchool 8 https://wiki.digitalmethods.net/Dmi/SummerSchool
2016SecureMessagingAppEcologies 2018AppStoresBiasDieter, Gerlitz, Helmond, Tkacz, van der Vlist, Weltevrede · Store, interface, package, connection 5
stakeholders for review. It is also used in commercial different degrees of emulation of native use situa-
technology reviews and has a longer history in prod- tions, including material devices, location and ac-
uct demonstrations and infomercials. Ben Light et al. tivity history, while other apps allow for more static
(2016) have suggested, however, that walkthroughs walkthroughs.
can be repurposed in a ‘significant departure’ from While in no way exhaustive, we foresee a number
these prior uses in order to perform a critical analy- of deployments of the walkthrough method for app
sis of a given app. For these authors, the core of the studies. As Light et al. note, walkthroughs can be en-
method involves ‘step-by-step observation and docu- acted through different phases of use, including sign-
mentation of an app’s screens, features and flows of ing in, everyday scenarios of routine use and quitting
activity’, all of which can be contextualised within an app. However, as indicated in the examples pro-
the app’s vision, operating model and governance, or vided thus far, we see several opportunities arising
what they call the app’s ‘environment of expected use’ from moving from single to comparative app analy-
(pp. 881–886). In this way, Light et al. propose reap- sis. One can consider, for example, how different
propriating this method within a cultural studies and apps handle key moments of the walkthrough (login,
STS framework as a contribution to the methodologi- terms and conditions, support screens, verification),
cal study of apps. how specific features are organised (action points, in-
In terms of strengthening the interdisciplinary put buttons, notifications), how design patterns are
context for walkthroughs, however, the method ad- implemented, or how navigation paths are arranged.
ditionally benefits from a critical understanding of To illustrate, Figure 1 (p. 6) compares the different
user experience design epistemologies and practices. signup options across mindfulness apps as they ap-
This can, for instance, draw attention to the rise of pear in different stages of the walkthrough.
behavioural design regimes as a mode of scripting Comparative walkthroughs can also be used to
the user that targets the nonconscious dimensions contrast the multi-sidedness of platforms. Here, the
of cognition (Hayles, 2017), leveraging insights from approach involves adopting platform-afforded per-
behavioural economics, cognitive psychology, and sonas – user, developer, advertiser, and so on – in
neurological research, while operationally relying on order to consider how platform providers address
big data and nudging (Yeung, 2017). In this regime of their varying groups on different sides via distinct
design, user journeys contain a series of key perfor- interfaces (cf., Bucher & Helmond, 2018). By de-em-
mance indicators that are ‘sunk’ into interfaces as an phasising the user-centred app walkthrough, multi-
environment to facilitate the captivation of the user. sided walkthroughs can make visible economic and
An emphasis on these characteristics benefits from value creation strategies that are not apparent from
forms of critical design literacy. Indeed, behavioural the user side of the market. This in turn may open
design or dark pattern libraries (Dieter, 2015; Nodder, up a number of political economic areas of inquiry,
2013) might be consulted in this respect to guide the such as how apps reconfigure and regulate platform
analysis of formal components based on plotting user labour on the level of interface design. Finally, walk-
decision-making and actions. throughs can be used for historical analysis, where
It is important to recognise how walkthroughs are versions of an app are considered to detect design,
uniquely performative as a situated rendition of a feature or data capture changes over time. This might
user journey that foregrounds material characteris- be run in an emulator or within environments like
tics of the interface. In this respect, the walkthrough Android Studio, including specific simulations of pe-
is a methodological intervention that inevitably in- riod hardware and operating systems.
volves a user persona to facilitate the process of en- In dialogue with these approaches, there are ad-
gagement within an app. Personas have been a main- ditional opportunities to refine the scope of the walk-
stay of HCI and interaction design as a method that through, particularly in attending to the more formal
produces a realist fiction of a user (Cooper, 2004). or templated aspects allow for data input (Gehl, 2014),
It usually involves empirical research into a market including touchpoints, buttons and forms that might
or audiences for a product, then imagining an ideal be emphasized to indicate platform relations like the
type that can be utilised to develop this software or presence of Facebook or Google logins that speak to
a service. Personas can be used to orient a situated the techno-economic relations of platformisation
enactment of a walkthrough – including processes of (Helmond, 2015). As an example, Figure 2 (p. 7) is a
repetitive or habitual use that might be required to series of comparative walkthroughs of dating apps
investigate personalisation – but also simply for prac- that explore how they are situated within specific
tical purposes of connecting to other existing profiles data infrastructures.9 In this case, the main concern
in social media. Here, we might consider distinguish- involves tracing the magnitude and pacing of inbound
ing between a user persona and a research persona and outbound data flows, linking the micro-practice
that recreates abstract use case scenarios which are
aligned with the interests of the researcher. Depend-
ing on the app, the creation of personas may require 9 https://wiki.digitalmethods.net/Dmi/DataAndDatingFig. 1 : Side-by-side comparison of login or sign-up screens in a selection of mindfulness apps.
6 CRC Media of Cooperation Working Paper Series No. 4 August 2018Dieter, Gerlitz, Helmond, Tkacz, van der Vlist, Weltevrede · Store, interface, package, connection
Fig. 2 : Comparative information visualisation of the user registration process in dating apps (Tinder, Grindr, OkCupid, Christian Mingle, Badoo).
78 CRC Media of Cooperation Working Paper Series No. 4 August 2018
of interface engagement with the distribution of per- The two main formats for mobile apps are .ipa
sonal information. The visualisation strategy, accord- (iOS application archive) files for iOS apps and .apk
ingly, abandons the GUI -screen capture in favour of (Android Package Kit) files for Android apps. They
a more data-centric approach, enabling the quantifi- are both specific types of archive files or compressed
cation and comparisons of informational disclosure software packages that can be extracted to view the
patterns across a set of related apps. code and resources of apps. Developers upload these
Emphasizing the situatedness of app interfaces files to the app stores where they can be subjected
can, accordingly, be further developed through ex- to review before being admitted to these stores for
perimental visualisations to abstract shared features, further distribution. Users typically do not see these
data flows and infrastructural configurations, or application archive files as they are automatically
forms of mapping that draw inspiration from tech- downloaded and installed onto their devices in the
niques in information architecture. The latter might background when using app stores. Device manufac-
be utilised, for instance, to address particular chal- turers such as Apple strictly limit what can be done
lenges in the analysis of user journeys like mapping with their devices and only allow the installation of
branching paths or separate ‘support screens’ to assist apps that have been approved by the official store.
with priming the user for further acts of disclosure. Downloading and installing iOS apps outside of
It may seem obvious, but it is important to stress Apple’s official App Store requires ‘jailbreaking’ and
that the walkthrough approach ultimately works unlocking the device as well as utilising a third-party
with a series of interfaces. The screen captures used app manager such as Cydia (iOS) or Cydia Impactor
to document, annotate or ‘markup’ the walkthrough (desktop). Android, on the other hand, positions itself
are, we suggest, not to be reduced to images and as an open platform and developers can distribute
analysed with semiotic methods. Apps are first and their apps via various third-party Android app stores
foremost operational media; they are applications, and marketplaces, or via their own websites.10 Users
things for doing. Importantly, apps are designed with can then configure their device settings to download
behaviours – not meanings – in mind. App developers and install ‘unknown’ apps from outside Google Play.
aim to get their users to do specific things – to change Such alternate marketplaces or websites, however,
their behaviour – and the walkthrough method can are still generally designed to facilitate downloading
be used to reflect this behavioural focus. While user for standard use, rather than to enable inspection of
experience, usability, and cultural studies-inspired the software package itself.
approaches place the user at the centre of the re- Gaining access to an individual application ar-
search, we see potential in using walkthroughs to chive file (as software) typically requires moving
examine how apps are infrastructurally situated by from the official app stores to so-called app reposi-
teasing out data flows, design strategies and plat- tories or using other dedicated software. Much like
form logics as their broader conditions of possibil- app stores, app repositories are a storage location
ity. The research persona, therefore, plays a special from which software packages may be retrieved and
role as the methodological user surrogate; enabling installed on a computer for personal or research pur-
access to app interfaces, while facilitating heterog- poses (Allix et al., 2016). These repositories are often
enous research situations. presented as online directories of apps with down-
load links to multiple prior versions of the package.
While Cydia contains the largest repositories for iOS
App packages apps, the leading repositories for Android apps in-
clude Aptoide, A PK P ure, A PK M irror, and F-Droid.
Our third entry point allows for the exploration in App repositories may visually resemble the look and
more detail of the embedded infrastructural ar- feel of official app stores and similarly display the
rangements of apps by engaging with them as soft- most downloaded apps, app categories, and various
ware packages, an experience that is usually shielded search options. Repositories such as A PK P ure contain
away from app users. Downloading and installing a wide array apps, but only offer a few versions of an
apps through official app stores is regularly presented app, while A PK Mirror appears narrower in scope but
as a seamless experience where users do not get to with many versions of the most popular apps. This is
see the downloaded app on their devices, but rather to suggest that each repository is to be considered on
only experience the automatically installed version. its own terms; none operate as perfect archives and
In this way, app stores obfuscate the status of apps as each has different affordances for doing (historical)
concrete software objects (cf., Morris & Elkins, 2015). app research. Within software engineering, software
To work against this obfuscation, it is necessary to re- repositories are an important source for studying the
situate apps outside their normative context of con-
sumption by utilising software repositories and tools
for analysis as packages. First, however, let us briefly 10 https://developer.android.com/distribute/marketing-
introduce the different software formats of apps. tools/alternative-distributionDieter, Gerlitz, Helmond, Tkacz, van der Vlist, Weltevrede · Store, interface, package, connection 9
evolution of software (Kagdi et al., 2007). This im- collect various forms of data, usually for advertising,
mediately raises issues about running or emulating authentication, and performance optimisation pur-
older A PK packages to examine the larger techno- poses. We refer to these third-parties as trackers.
commercial ecosystems apps are embedded in over While it is possible to examine A PK files for track-
time or analyse the evolution of the app ecosystem ers directly, we rely on Exodus Privacy – a ‘privacy
at large (cf., Helmond, 2017). Working with app re- auditing platform for Android applications’ – to auto-
positories also raises juridical concerns as it is not mate the process. Exodus scans A PK files, compares
always clear whether an app has been uploaded le- them with its list of known tracking technologies
gally or not. Often there can be issues with malware and generates reports for individual apps. Another
and spam in such app repositories. To prevent these tracker tool, the DMI App Tracker Tracker, is built on
concerns, it is possible to draw on dedicated software top of Exodus Privacy and extends its functionality
such as Raccoon to download current A PK s directly to detect known tracking technologies or other soft-
from Google Play, and to bypass the repositories alto- ware libraries in a set of A PK files collected from an
gether. However, Raccoon still requires authenticat- official app store or app repository. There is a lot of
ing with a Google Account to retrieve apps and will related work in computer science and software en-
only provide the latest app version. gineering examining third-party libraries, including
Situating apps as software packages allows re- advertising libraries, in apps (Book et al., 2013; Ma
search beyond the app’s interface and into the code. et al., 2016). To build a set or collection of apps one
For example, since A PK files are always also valid can follow similar collection-making strategies as
.zip archive files, one can view their contents by previously described for the app stores. The ability to
unzipping the file. Other app packages like IPA files make collections of apps for analysis is itself an affor-
may also be unzipped to view their package contents dance that is gained through such repositories.
and structure, but these contents may be encrypted Examining trackers embedded within mobile
differently (e. g., due to digital rights management apps renders visible a number of otherwise obscured
(DR M) restrictions). Some parts of apps may need stakeholders. Comparative analysis of different apps
further decoding in order to fully view their con- or groupings of apps can also be used to determine
tents and this can usually be achieved with the sup- which advertising or analytics providers dominate
port of additional tools.11 Decoding apps shows the different areas, or which types of apps are loaded
contents of the package including all the necessary with trackers. For example, one previous project ex-
files and resources. The AndroidManifest.xml file, amined tracker code presence within a number of
for example, describes metadata like the name, ver- app sets.12 Additionally, with the help of repositories,
sion, and contents of the A PK file and also includes it is possible to study how the presence of trackers
information about the app’s permissions. Further in an app or group of apps has changed over time,
resources in the package include software develop- which offers insights into changing app stakeholder
ment kits (SDK s) used to build particular modules relations, business model pivots, or otherwise reflect
in the app, including social logins, app analytics and dynamics in the wider economies of app advertising,
advertisement libraries. app development, app analytics, and mobile game
Re-situating apps as software makes them avail- monetisation.
able to the range of enquiries found in critical code
and software studies (Montfort et. al., 2012; Fuller,
2008). One can study the code up close, or parse it App connections
through other diagnostic tools enabling comparisons
across apps or sets of apps (e. g., Appcestry). Data Our fourth entry point for multi-situated app stud-
sourced from files can also be used to complement ies are the network connections that mobile devices
other methods. For example, data on an app’s per- establish and that allow both multi-sidedness and
missions sourced from the AndroidManifest.xml multi-sitedness to be traced. These connections are
file can complement a walkthrough study of when often established on behalf of the apps running them,
and how an app collects user data. For present pur- and are needed for things such as user authentica-
poses, app packages can be used to examine the vari- tion, app updates, advertisements, and serving and
ous stakeholders involved in the production of apps uploading content. It is well-known that mobile de-
(multi-sidedness) as well as the infrastructures for vices keep logs of the countless access points they
transferring and storing app content and data (multi- probe while passing through the access radius of Wi-
sitedness). In particular, through analysing inscribed Fi access points. Indeed, whenever location services
libraries like SDK s, it is possible to identify app tem- are enabled, mobile devices attempt to connect to
plates and third parties embedded within apps that
12 https://wiki.digitalmethods.net/Dmi/WinterSchool
11 https://ibotpeaches.github.io/Apktool/ 2018MappingDataIntensiveAppInfrastructures10 CRC Media of Cooperation Working Paper Series No. 4 August 2018
each and every access point that is listening for such and webs of connectivity are woven, they also serve
access requests. As Mackenzie points out, the impli- as distribution channels to collect and deliver data
cation is that users’ devices are continuously connect- traffic to anywhere between thousands and billions
ing and disconnecting to objects and infrastructures of mobile devices. While A PK archive files can be
‘without knowing exactly how or where’ (2010, p. 5). employed to detect software libraries written into
When approaching apps as being multi-situated in apps (e. g., SDK s) and thereby render static infra-
this sense, network connections are a key entry point structural relations, network connections and net-
for understanding how apps are always and neces- work traffic are ultimately dynamic and ephemeral
sarily bound up with – or indeed ‘tethered’ to (Zit- infrastructural relations. They are established when
train, 2009) – other objects and infrastructures. an app is running – even when running invisibly in
The proposed approach relies on methods from the background – but they are dropped as soon as the
network security specialists (e. g., Enck et al., 2014) app is closed and cannot be rendered from an app’s
which are adapted to study the multi-situatedness package contents. They are triggered by certain spe-
of apps. In order to gain a sense of an app’s connec- cific events, cues, or conditions that not all users or
tive entanglements with other objects, devices, in- devices might meet as in the case of loading person-
frastructures, and services it is possible to capture alised content or advertising, which poses challenges
and log the connections that are being established. to approaches using source code analysis. Instead,
Similar to desktop computers, there are many ap- apps anticipate users or user profiles to trigger these
plications that monitor, track, analyse, and display events or conditions, and the outcomes are specific
inbound and outbound connections from and to a to, and dependent upon them. The permissions re-
device. These applications typically sit somewhere quired by most operating systems to establish net-
‘underneath’ the application to capture the device’s work connections are approved during installation.
low-level network connections and hence might re- Interestingly however, internet permissions are clas-
quire privileged control (i. e., root access). While this sified as ‘normal’ on Android as one of the standard
is more common for advanced Android users, it is dif- permissions – that is, ‘permissions that don't pose
ficult for iOS users to ‘jailbreak’ their iPhones. As a re- much risk to the user's privacy or the device's opera-
sult, network connections cannot always be isolated, tion’ (Android Developers) – that apps need and are
or associated with the apps from which they were granted automatically (and cannot be opted out!).14
derived. Such apps typically log metadata like dates Additionally, apps might not require all the permis-
and times, protocols, packet sizes, IP addresses, and sions they request and might access more data than
bundle IDs of the apps connecting to these addresses. is functionally needed.
These details can be used to distinguish different The boundedness of apps as bundles or pack-
kinds of connectivity (e. g., active or background, ages is challenged when recognising that apps are
inbound or outbound), chart infrastructural rela- routinely extending themselves through these net-
tions to remote hosts and servers, authentication or work connections. On the one hand, researchers
authorization providers (e. g., OAuth, social logins), can observe the topologies, rhythms, and volumes
third-party content delivery networks (e. g., Akamai, of inbound and outbound data traffic flows through
Amazon CloudFront), cloud services (e. g., Amazon network sniffing methods. For example, it is possible
Web Services, Microsoft Azure), ad networks (e. g., to detect and characterise different kinds of connec-
AdMob, MoPub), and thousands of other tracking tions and implicated third parties for different sets of
technologies. Each of these connections provides apps. Merely rendering the networks of third parties
different affordances and insights into how apps are associated with apps visible is arguably a powerful
entangled with other objects and infrastructures, to rhetorical strategy for critical internet infrastructure
account for the multiple sites and sides of apps, and studies. Additionally, there has been a growing inter-
to render the webs of connectivity that apps and mo- est in studying the materiality of internet infrastruc-
bile devices weave. ture and signal traffic (Parks & Starosielski, 2015).
Previous research on tracking technologies, cloud What or who do these connections serve? Further-
infrastructures, and data infrastructures is based more, researchers can also inspect or even intercept
mainly on web corpora.13 However, some of these the data packets transmitted unsecurely across these
approaches may be resensitivised to explore the network connections with packet inspection meth-
multiple sites and sides of apps. For example, net- ods. Using common network data packet inspection
work connections can be studied to gain a sense tools like Wireshark and tcpdump, we collected and
of the many infrastructural relations, dependen- analysed query parameters in HT TP requests, which
cies, data traffic flows, and third parties connected among other things yielded detailed ad requests to
to apps. Once network connections are established,
14 https://developer.android.com/reference/android/
13 https://digitalmethods.net/Dmi/TrackersGuide Manifest.permission#INTERNETDieter, Gerlitz, Helmond, Tkacz, van der Vlist, Weltevrede · Store, interface, package, connection 11
structural relations for serving content, serving ads,
signing in, and sharing content?
Towards situated app studies: nine propositions
The four methodological entry points introduced
above enable the production of different situations
Fig. 3: An encoded MoPub URL with unencrypted within which the multi-sidedness and multi-sited-
HTTP ad request parameters and values (device ness of apps can be made available for research. At
name, bundle ID, gender, age, lat long, screen width, the same time, these entry points raise questions con-
height, language, carrier network, permissions). cerning how researchers in turn situate themselves
methodologically towards apps and their socio-mate-
ad networks (Figure 3).15 Such data is accessible rial relations. In what follows we offer nine proposi-
through any unencrypted HT TP connection and may tions to address the empirical and conceptual chal-
easily be captured. lenges of studying apps and similar digital objects.
Some network utility apps like Network Connec-
tions16 (Android) allow users to live capture network (1) Move beyond ready-made social data. Unlike the
connections while using a device and to export the web or social media platforms, which are considered
logs in standard tabular file formats. Such features to offer a variety of data points ready-made for ‘social’
also point to the possibility to script certain uses, us- investigations and accessible in structured ways via
ers, and use scenarios and to design research proto- open A PIs, apps are rather characterised by hetero-
cols that are more controlled or systematic. In these geneous data formats ranging from prestructured
kinds of studies, it is paramount to craft the research interface-level data, sensor data, and network con-
and methodology carefully so that the situations elic- nections to software libraries and infrastructural
ited through them can be interpreted. One can use data forms. Mobile apps may collect social data, but
a ‘clean’ researcher phone with ‘fresh’ user accounts mostly do not offer structured access via open A PIs,
for the app(s) under study. But one could also con- which means there is an absence of data ready-made
ceive of rich and mature profiles trained to enact a for both developers and researchers. Apps that pres-
researcher’s choreographed situation (e. g., to trig- ent social data through their interfaces (e. g. dating
ger cookies, personalisation, localisation, targeted apps, chat and instant messaging apps) often cannot
ads). Some of these strategies were originally devel- be easily retrieved as the alternative data collection
oped for studying personalisation and localisation in technique of screen-scraping is not accessible on mo-
search engine results (Feuz, Fuller, & Stalder, 2011; bile apps. Engaging with apps through the different
Rogers, 2013). Once network connections data are entry points shows that data or features that may
obtained through dedicated tools, it is sometimes appear as being specific to one app may actually de-
also possible to trace them back to the firms and or- rive from a different data source, an imported plugin,
ganisations behind them (Figure 4, p. 12).17 Network device-based sensors, or external tracking or adver-
connection data include IP addresses of the connec- tising technology. Apps present researchers with the
tions an app establishes, and these can be converted challenge to devise inventive methods that respond
into domain names, locations, and ISPs of their hosts to the heterogeneity, to the different kinds of struc-
by using IP lookup tools. The results can be matched ture, and multiple origins of data in which ready-
to other expert lists containing known infrastruc- made data are both rare and in need of unpacking.
tural technology providers, such as Ghostery for web
tracking technologies and CDN Finder for content (2) Navigate infrastructural resistance. While obfusca-
delivery networks. To create network connection tion is a commonplace technique in computer science
situations, questions to keep in mind are, what are and software engineering (Matviyenko et al., 2015),
the sites (locations, server hosts, data centres, cloud the deliberate efforts or infrastructural effects that
servers) that are being connected to? And which but- render code and data illegible to both human and
tons – or what kind of scripts – trigger these infra- technical interpretation have a significant impact
on app research. Some forms of obfuscation are in-
tentional: A PK files might be designed to frustrate
15 https://wiki.digitalmethods.net/Dmi/DataAndDat decompiling tools; certificate pinning and encrypted
ing; https://www.digitalmethods.net/Dmi/Winter channels can limit the scope of packet sniffing; while
School2018MappingDataIntensiveAppInfrastructures entire app ecosystems like Apple’s iOS can appear
16 https://play.google.com/store/apps/details?id=com. essentially off-limits due to DR M protection. Others
antispycell.connmonitor may result from the fact that functions or data flows
17 https://wiki.digitalmethods.net/Dmi/DataAndDating do not address human users and are instead designed12 CRC Media of Cooperation Working Paper Series No. 4 August 2018
Fig. 4: Companies behind the network connections coming in or out of
four popular dating apps (OkCupid, Grindr, Tinder, BeeTalk).
to be machine-readable. In this sense, it is not un- (3) (Un)do the user. Empirical engagement with apps
common for app studies to encounter considerable opens up a complicated relation to the user: while
infrastructural resistance that might circumvent or both software and platform studies have bracketed
side-track empirical research. Thus, despite recent out users and their practices, app research requires
claims that digital media open up new possibilities to a partial return of the user. Many apps require per-
empirically realise A N T ’s principle to ‘follow the ac- sonalised logins, build on existing social media pro-
tors’ (Venturini et al., 2017), apps pose various chal- files, initiate data flows only through user practices,
lenges to the idea of following actors (or things). But or are tied to users’ specific locations. While disen-
rather than seeing these challenges as black-boxed tanglement from user practices is possible in some
limits, it is more productive to consider instances of situations (such as minimising the personalisation
obfuscation as offering a spectrum of opportunities effects of app stores), or might simply not be an is-
to navigate around resistances and resituate apps to sue in others (A PK-based methods that work with-
open up alternative (albeit partial) perspectives. Pro- out profiles), it is simply not possible when perform-
ducing situations, accordingly, may involve working ing walkthroughs, studying network connections or
both with and against the objectives and infrastruc- sensor-based data flows, as all these build on situated
tural resistance of apps. practices. Apps present researchers with the paradoxDieter, Gerlitz, Helmond, Tkacz, van der Vlist, Weltevrede · Store, interface, package, connection 13
that in order to systematically move away from user- within a single app. Engaging with apps via multiple
centred research, they need to deliberately create situations may require navigating different scalar
them. To mediate this issue, we suggest repurposing levels and relating specific findings in performa-
the design method of creating personas – fabricated tive embodied situations (such as walkthroughs) to
users with habits, preferences, psychographic or de- global app markets or cloud infrastructures. App re-
mographic specificities to inform decision-making search thus requires a heightened sensitivity towards
around product development. Personas can be used scale and scope, towards where and how a particular
to generate research situations by acting as kind of method is positioned.
surrogate user. They can be deployed to create ab-
stract scenarios of use for walkthrough methods, or (6) Engage with apps statically and dynamically. Apps
to activate data flows or scenarios of location aware- are run on devices that continuously collect data and
ness. The distance between the researcher and their make connections (through background sensing,
research persona can also afford a degree of cogni- updates, network-based calculations, and so forth);
tive shelter regarding the behavioural techniques some of which are enabled by specific practices and
of interface design. Generating and maintaining a personas, while others are not. We differentiate be-
research persona can, however, be challenging, as tween methods that can be considered ‘dynamic’
newly created personas may not trigger the same (deployed within a ‘native’ situation) and ‘static’ (ex-
elements (especially ads) as more ‘organic’ perfor- tracted from a native situation). Such a distinction
mances. The use of personas may also stand in con- is significant because dynamic app studies require a
tradiction to an apps’ terms and conditions and can more deliberate effort in controlling the conditions
raise ethical concerns when interacting with other of the situation by setting up personas, re-enacting
users. physical locations through V PNs, or using specific re-
search phones or software environments. Static ap-
(4) Move from content to practices. App research af- proaches such as A PK research or repurposing app
fords a renewed interest in the role of practices – (store) analytics by contrast do not require the enact-
particularly software-enabled mundane routines ment of native use situations.
(Morris & Elkins, 2015) – as opposed to the study of
content prevalent in web and platform research. App (7) Resist presentism. The fast update cycles of apps
interfaces enable specific, embodied, and often con- and the design decision of app stores to only display
text-dependent activities, and, in this way, app stores the latest versions creates a temporality persistently
do not necessarily sort and filter content, but rather focused on the here and now. The stores’ obfuscation
practices. For example, a search for health apps in techniques to prevent access to app packages and
Google Play hardly returns informational apps, but previous versions poses challenges to historical re-
rather the majority of health apps typically offer search approaches. App repositories may function as
meditation, medical, self-tracking and other tools access points to archives for downloading older ver-
and practice-focused solutions. While recognising sions, however they are often incomplete or contain
there is no clear line between content and practice, potentially illegal and spammy software. Moreover,
apps tend to resemble dedicated equipment that sup- apps may come with limited rollouts and exist in dif-
port forms of doing as their privileged mode of en- ferent versions, while specific packages are highly
gagement. Thus, when we study apps, we study the dependent on externally loaded dynamic resources:
conditions of possibility for practices. emulating old app versions may load current assets
in an old framework, mixing up some aspects of old
(5) Scale from situation to infrastructure and back. apps with present material. In addition, there are
The selected entry points allow the creation of re- temporal dimensions to the treatment of apps by app
search situations at multiple scales – from micro- stores and their ranking algorithms, as app ‘fresh-
practices enacted through interfaces to an aggrega- ness’ may determine its ranking and app rankings do
tive bird's-eye view of app store analytics. While the change over time.
app conceals itself as a bounded object, the app store
brings it into relation with other objects by grouping (8) Contest Silicon Valley imperialism. The dominance
the app with ‘Similar Apps’ – or apps that ‘You May of the Silicon Valley-native App Store and Google
Also Like’ – or apps by the same developer. Individual Play hides the fragmented, and culturally and tech-
apps can also belong to a larger group of apps such nically specific landscape of app stores. Each operat-
as Facebook’s ‘family of apps’, conceptualised as ‘app ing system, device manufacturer, country, and type
constellations’ which are collections of ‘mobile apps of app may have its own app store that comes with
that share a single login and have app to app link- specific infrastructural affordances and resistances
ing built in’ (Wilson, 2014). Or so-called 'super apps’ for doing app research. For example, the alternative
like WeChat, which contain multiple services ranging Russian app store for Android devices, Yandex.Store,
from communication and finance to retail and social does not offer a web interface to enable data collec-You can also read
