scispace - formally typeset
Search or ask a question
Journal ArticleDOI

A Longitudinal Study of Pervasive Display Personalisation

TL;DR: The experiences of designing, developing and operating the world’s first comprehensive display personalisation service for mobile users are reported on, and a series of reflections are offered to inform the design of future systems.
Abstract: Widespread sensing devices enable a world in which physical spaces become personalised in the presence of mobile users. An important example of such personalisation is the use of pervasive displays to show content that matches the requirements of proximate viewers. Despite prior work on prototype systems that use mobile devices to personalise displays, no significant attempts to trial such systems have been carried out. In this article, we report on our experiences of designing, developing and operating the world’s first comprehensive display personalisation service for mobile users. Through a set of rigorous quantitative measures and 11 potential user/stakeholder interviews, we demonstrate the success of the platform in realising display personalisation, and offer a series of reflections to inform the design of future systems.

Summary (7 min read)

1 INTRODUCTION

  • Pervasive displays, such as digital signage, are becoming increasingly ubiquitous in their built environment.
  • The authors results indicate that initially their display personalisation system can indeed reach a high conversion rate, but that the level of engagement starts to dwindle significantly over time.
  • The authors further highlight the challenges deployments face in collecting long-term usage data in the absence of reliable sources of information on content activation, and provide detailed insights regarding the long-term maintenance of a user base.
  • Using a combination of data collected from their deployments and controlled small-scale trials conducted in a laboratory setting the authors demonstrate that current commercial technologies can detect entry to display regions with reasonable accuracy, but they are poor at ACM Trans.

2 RELATEDWORK

  • There is a rich history of research into pervasive displays dating back to the 1980s [13].
  • This research includes a multitude of focal areas including, for example, display hardware, interaction modalities and audience behaviour.
  • For the purposes of this paper, the authors concern ourselves only with prior work relating to deployments of pervasive displays and with display personalisation.

2.1 Long-Lived Pervasive Display Deployments

  • Early research into display deployments explored their use as ‘media links’, i.e. using video and audio links to connect together physically separate spaces.
  • A similar system at Bellcore Labs, the VideoWindow [19], connected researchers on two different floors of the building using large projected displays in common areas.
  • The authors deployed a set of applications and content, investigating how user-generated content can improve the use of situated displays in urban settings.
  • One of the largest research-based deployments of pervasive displays was the UBI-Hotspots system, a network of interactive touch-enabled displays located across the city centre of Oulu, Finland [43].
  • Interact., Vol. 1, No. 1, Article 1. Publication date: January 2020. static content (e.g. slideshows), the network and functionality has since been substantially extended.

2.2 Display Personalisation

  • While the falling cost of hardware and the difficulty of reaching the general public through fragmented conventional media has led to the deployment of increasing numbers of public display systems, the vast majority of today’s public displays effectively disappear: people have become so accustomed to their low utility that they have become highly skilled at ignoring them [34, 41].
  • A more indirect form of personalisation was proposed by Müller and Krüger [40] in which the system estimates viewers’ paths between displays to coordinate content across multiple displays.
  • Tacita’s underlying premise is the trust model illustrated in Figure 1; viewers issue content requests to cloud-based content providers with which they already have an established trust relationship (e.g. BBC World Service, Facebook) while display owners only honour requests for screen real estate from similarly trusted content providers.
  • The user’s location is monitored within the mobile application in order to detect if a user has entered the trigger zone of a public display.

3 DEPLOYMENT CONTEXT

  • In order to conduct an investigation of issues surrounding in-the-wild deployments of personalisable pervasive displays, the authors deployed Tacita in the context of Lancaster University’s e-Campus, the world’s largest public display research testbed located.
  • Display nodes run Yarely, a digital signage player that retrieves display schedules from backend services in the form of Content Descriptor Sets (CDSs), an XML-based format describing content items and their scheduling constraints (e.g. date and time restrictions) [10].
  • Stakeholders of the e-Campus deployment who control and populate displays with content include college administrators, departmental officers, and the university’s press office.
  • Whilst a large number of students live on campus, the majority of students and staff commute (typically by bus).

4 SUPPORTING MOBILE DISPLAY PERSONALISATION AT SCALE

  • To support the deployment of Tacita for their long-term trials the authors have redesigned and restructured the original system architecture proposed by Davies et al. [15] (as summarised on page 7 of this paper).
  • The most significant change is the addition of a Display Gateway component described in more detail below.

4.1 Architecture

  • The authors reworked Tacita architecture consists of the following five core components.
  • Display Gateways provide an interface to a display deployment through which Tacita Channels can make requests for dynamic content personalisation on displays within the deployment.
  • Tacita Channels request the display of personalised content by providing the location, display and content identifiers, allowing the Display Gateway to validate the request and, if successful subsequently forwarding the request to the appropriate display node.
  • Firstly, in practice many displays operate behind behind firewalls that block traffic from external sources.

4.2 Integration

  • In its original implementation, their deployed Yarely configuration pulled updates to display schedules at fixed intervals (e.g. once an hour).
  • To support walk-by personalisation, the system was extended to support notifications from the Display Gateway to immediately adjust the content being displayed.
  • Yarely filters content items to produce a minimal set of eligible items, from which one is selected at random using a lottery approach [37].
  • These changes to Yarely are in addition to the development of an e-Campus Display Gateway to support incoming requests from Tacita Channels.

4.3 Localisation and Mapping

  • Fundamental to the Tacita approach is that all localisation is performed on a user’s mobile device and that these devices are also responsible for issuing personalisation requests.
  • To support this, the location and capabilities of displays are encoded in maps that provide detailed information on supported Tacita Channels for each individual display [15].
  • Davies et al. [15] envisaged two mechanisms for the dissemination of maps to the Tacita Mobile Client: announcements (i.e. displays transmitting their capabilities) and map providers (i.e. repositories of maps).
  • In addition, BLE beacons are not well-suited to transmitting large quantities of data and instead typically transmit identifiers for resolution by the receiver [32].

4.4 Tacita Channels

  • Tacita relies on users installing a mobile phone application and subscribing to one or more map providers.
  • The Tacita Mobile Client also allows users to configure each available Tacita Channel with their own preferences .
  • Additionally, the application registers a new significant movement monitoring area with a 100 meter radius around the user’s location.
  • To help lower the barrier of entry for new users at Lancaster University the authors have created a version of the Tacita Mobile Client that is integrated into iLancaster .

5.1 Example Channels

  • Prior research conducted by Clinch et al. [11] used focus groups and surveys to uncover user attitudes to display personalisation and to determine the types of content users imagine that they would wish to view.
  • Users interested in seeing national and international news can use the News Channel to chose from a selection of news sources and categories (e.g. World, UK, Sport, Science etc.) – the display will show a randomly ordered set of news stories from two of the sources chosen by the user.
  • A World Clock Channel allows users to select the name of a town/city/country for which are interested ACM Trans.
  • In both cases the display will show the set of tweets (in a similar format to the news app).

5.2 Implementing Tacita Channels

  • All of their Tacita Channels perform a number of common functions: (1) They provide aWeb-based interface that can be accessed via the Tacita Mobile Client allowing users to submit channel-specific configuration parameters (e.g. location for a personalised weather forecast).
  • (4) They provide content for displays (typically in the form of a dynamic Web page) if the requests to the screen real-estate have been successful.
  • As a result of the framework, creating a typical Tacita Channel that serves a static web page to users as they pass by displays requires less than 20 lines of Python (the library itself comprises approximately 650 lines).
  • While the framework supports most Web content, the Twitter and TV Channels required additional bespoke coding.

6 METHODOLOGY AND DATASETS

  • The authors evaluation combines quantitative and qualitative analysis.
  • Quantitive data allows analysis of real-world usage patterns (Section 7) and benchmarks system performance (Section 8).
  • Qualitative methods are used to understand individual attitudes to personalisation of an established digital signage network (Section 9).
  • In the following sections the authors first describe the methods and set-up for the quantitative analysis and then the methodology for qualitative data capture and analysis.

6.1 Quantitative Analysis of Tacita Performance and Usage

  • The authors study is based on quantitative measurements collected from their deployment at Lancaster University.
  • This dataset consists of 224, 189 events (including 24, 673 content requests) from a total of 147 unique users (this count considers only users who both installed the Tacita Mobile Client4 and interacted with Tacita at least once over the duration of the study).
  • In detail, event (1) was captured directly on the iOS client application as soon as the iOS background process allocated processing time to Tacita (triggered by an iBeacon sighting).
  • For events (2)-(4) , the authors logged each request server-side including a timestamp of the event occurrence with the clocks of all the server components being synchronised.

6.2 Qualitative Analysis of Tacita Usage

  • The authors qualitative studies aimed to develop understanding of patterns of use seen in the quantitative usage data.
  • In addition to informing discussions about value and risk, their engagement with display viewers was intended to capture experience of both using the Tacita Mobile Client to configure personalisation preferences and of viewing personalised content on public displays (including, for example, any potential usability concerns).
  • Demographic information was not collected but all participants were current undergraduate or postgraduate students at Lancaster University.
  • The authors participants had no prior experience of Tacita but were invited to install and use the iOS application on their own phone for the duration of the interview: three participants chose to do this.

7 ANALYSIS OF USAGE PATTERNS

  • The authors begin their analysis by investigating usage patterns of the in-the-wild deployment at Lancaster University with the aim of characterising how personalised displays are used outside controlled settings.
  • In addition, the authors investigate the spatial distribution of personalisation requests across campus.
  • As the source of data, the authors consider the usage and application logs captured through the users’ mobile devices as well as request logs and configuration parameters recorded on Tacita Channels.
  • The authors analysis contrasts with prior research on public display personalisation that has predominately focused on investigating technical aspects such as the use of Bluetooth identifiers to provide personalised content [14] and providing novel forms of content and applications [28] but has lacked the long-term deployments that enable research into usage patterns.

7.1 Channel Selection

  • Display personalisation enables users to express a preference for the type of content they wish to see on pervasive displays.
  • Similar patterns can be observed when considering the total number of requests per day across all users, as shown in Figure 8.
  • Only 1.4% of content requests (356) were triggered manually, i.e. users specifically opened the Tacita Mobile Client to manually request a piece of content.
  • These statistics show that only a minority of users manually adjusted the content while standing in proximity of a display, supporting a view that pervasive displays should be designed with minimal interaction requirements to ensure continued usage.
  • The authors long-term usage data seems to suggest that there is a strong correlation between Channels users report as potentially interesting and those they use in actuality – easing the process of selecting which channels to prioritise.

7.2 Spatial Patterns of Requests and Display Dwell Times

  • The authors next consider the spatial patterns of content requests together with the duration of users’ proximity (dwell times) at different displays, with the aim of understanding how display location and participantmobility influences content selection.
  • The authors separately considered the spatial distribution of content requests for four example Channels one from each category : Bus Timetables (the most popular Channel, b), Live TV (average popularity, b), News (average popularity, c), and Twitter (the least popular Channel, d).
  • The authors also compute the mean dwell time of viewers at each display location, providing us further insights regarding the behaviour of viewers and the characteristics of individual displays locations.
  • For locations with higher dwell times, the prompt delivery of content becomes less important.

7.3 Personalisation Opportunities and Duration of Usage

  • Experience with traditional mobile applications suggest that one of the major challenges is to maintain a high user base.
  • Note that content requests are only issued when users are within the proximity of a Tacita-enabled display and hence the absence of a content-request within a given time-window can not be solely attributed to abandonment/failure of the Tacita Mobile Client application.
  • This is also reflected in a small number of application launches captured through on-mobile analytics (mean 2.1 sessions per day; SD: 4.76).
  • In summary, their usage analysis based on recurring personalisation requests together with the mean number of requests per day provide some indication that display personalisation systems are indeed used over longer periods of time – yet by lower proportions of users than initially expected.

8.1 Overview

  • In particular, its ability to support the most challenging use case of display personalisation: walk-by personalisation.the authors.
  • Figure 17 illustrates the three key sources of delay or inaccuracy that impact on content exposure and accuracy measures: ACM Trans.
  • (2) System Latency describes the delays incurred in processing the request for display personalisation.
  • (3) Beacon Exit Detection Delta represents the time delta between the viewer leaving the viewable area of a display (and the adjacent BLE beacon) and the point at which the background location tracking detects that the viewer is no longer in proximity of the display’s beacon.

8.2 BLE Beacon Performance for Display Personalisation

  • To establish the on-device delay in detecting proximity to iBeacons the authors conducted two experiments: one in controlled laboratory conditions and the other using their real-world testbed.
  • The authors also evaluated the performance of leaving the proximity of a beacon by capturing the time delay between deactivating the beacon transmission and the point at which the beacon was detected as “lost” by the mobile device.
  • Entry and exit detection latencies are highly dependent on (1) the spatial layout of the area in which the display and BLE beacon have been placed, (2) potential background tasks and radio processing on the viewer’s mobile device, and (3) the technology used to detect viewer proximity.

8.3 Tacita System Performance

  • In the previous section the authors considered the on-device performance of detecting beacons in proximity to the user’s mobile device.
  • For the results presented in this section the authors consider the data captured during the long-term in-the-wild trial described in Section 6.
  • Table 5 presents a breakdown of the latencies measured for individual Tacita system components.
  • Latency for detecting and reporting the proximity of beacons is highly dependent on the data connectivity of the user’s mobile device and the speed with which the device detects a proximate beacon.
  • The forwarding of requests from the Display Gateway to content shown on a Public Display is the largest source of latency (median: 1.33; mean: 3.16; SD: 7.19).

8.4 Additional Considerations

  • Requests in Tacita typically fail due to one of three reasons: (1) the requested content is not available, (2) the display is already showing higher priority content or, (3) the display receives multiple, conflicting, personalisation requests.
  • Figure 16c shows the total number of failed requests per day.
  • Compared to the total number of requests, the number of failed requests is relatively low.
  • This suggests that Tacita and its subcomponents are reasonably robust in their handling of personalisation requests.

9.1 Engagement with Display Viewers

  • The authors seven interview participants all owned and used a smartphone: ACM Trans.
  • Campus map, personal calendars, round-robin emails, Instagram, and a Channel that allowed users to broadcast their own content to others, also known as Other suggestions were.
  • Three of the content creators were able to provide a good overview of Tacita’s functions while the fourth professed to no knowledge of the system – for this content creator the authors provided a brief overview of Tacita’s key features.

10 SUMMARY AND CONCLUSIONS

  • The first such real-world deployment to take place at scale (in terms of number of displays, duration of deployment, and number of different items ACM Trans.the authors.
  • Through the process of actually conducting a real-world deployment, the authors identified the need for an additional component not described in prior architectural explorations of pervasive display personalisation.
  • The authors Channel usage data also validates prior observations [11] that personalised social media content is unlikely to be a good choice for public displays.

Did you find this useful? Give us your feedback

Figures (21)

Content maybe subject to copyright    Report

1
A Longitudinal Study of Pervasive Display Personalisation
MATEUSZ MIKUSZ, Lancaster University
PETER SHAW, Lancaster University
NIGEL DAVIES, Lancaster University
PETTERI NURMI, University of Helsinki
SARAH CLINCH, University of Manchester
LUDWIG TROTTER, Lancaster University
IVAN ELHART, Università della Svizzera italiana (USI)
MARC LANGHEINRICH, Università della Svizzera italiana (USI)
ADRIAN FRIDAY, Lancaster University
Widespread sensing devices enable a world in which physical spaces become personalised in the presence of
mobile users. An important example of such personalisation is the use of pervasive displays to show content
that matches the requirements of proximate viewers. Despite prior work on prototype systems that use mobile
devices to personalise displays, no signicant attempts to trial such systems have been carried out. In this paper
we report on our experiences of designing, developing and operating the world’s rst comprehensive display
personalisation service for mobile users. Through a set of rigorous quantitative measures and eleven potential
user/stakeholder interviews, we demonstrate the success of the platform in realising display personalisation,
and oer a series of reections to inform the design of future systems.
CCS Concepts:
Human-centered computing Ubiquitous and mobile computing
;
Ubiquitous and
mobile computing systems and tools.
Additional Key Words and Phrases: mobile computing, smart environments, location-based applications,
pervasive displays
ACM Reference Format:
Mateusz Mikusz, Peter Shaw, Nigel Davies, Petteri Nurmi, Sarah Clinch, Ludwig Trotter, Ivan Elhart, Marc
Langheinrich, and Adrian Friday. 2020. A Longitudinal Study of Pervasive Display Personalisation. ACM Trans.
Comput.-Hum. Interact. 1, 1, Article 1 (January 2020), 46 pages. https://doi.org/10.1145/3418352
1 INTRODUCTION
Pervasive displays, such as digital signage, are becoming increasingly ubiquitous in our built
environment. Estimates suggest that over 45 million digital signs have been deployed globally [
58
].
Authors’ addresses: Mateusz Mikusz, Lancaster University, InfoLab21, Lancaster, U.K., LA1 4WA, m.mikusz@lancaster.ac.uk;
Peter Shaw, Lancaster University, InfoLab21, Lancaster, U.K., LA1 4WA, p.shaw@lancaster.ac.uk; Nigel Davies, Lancaster
University, InfoLab21, Lancaster, U.K., LA1 4WA, n.a.davies@lancaster.ac.uk; Petteri Nurmi, University of Helsinki, InfoLab21,
Lancaster, U.K., LA1 4WA, p.nurmi@lancaster.ac.uk; Sarah Clinch, University of Manchester, Manchester, U.K., M13 9PL,
sarah.clinch@manchester.ac.uk; Ludwig Trotter, Lancaster University, InfoLab21, Lancaster, U.K., LA1 4WA, l.trotter@
lancaster.ac.uk; Ivan Elhart, Università della Svizzera italiana (USI), Lugano, Switzerland, 6900, ivan.elhart@usi.ch; Marc
Langheinrich, Università della Svizzera italiana (USI), Lugano, Switzerland, 6900, marc.langheinrich@usi.ch; Adrian Friday,
Lancaster University, InfoLab21, Lancaster, U.K., LA1 4WA, a.friday@lancaster.ac.uk.
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee
provided that copies are not made or distributed for prot or commercial advantage and that copies bear this notice and
the full citation on the rst page. Copyrights for components of this work owned by others than ACM must be honored.
Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires
prior specic permission and/or a fee. Request permissions from permissions@acm.org.
© 2020 Association for Computing Machinery.
1073-0516/2020/1-ART1 $15.00
https://doi.org/10.1145/3418352
ACM Trans. Comput.-Hum. Interact., Vol. 1, No. 1, Article 1. Publication date: January 2020.

1:2 M. Mikusz et al.
Despite their ubiquity, there is little evidence that such displays are an eective way of communi-
cating with potential viewers. Indeed, systematic observations of viewer behaviour in the proximity
of digital signs suggest that viewers exhibit a phenomenon known as display blindness choosing
to ignore public displays because they perceive them as having little content of relevance [
34
,
41
].
Since the mid nineties, researchers have attempted to address display blindness by enabling
content to be tailored or personalised to viewers within close proximity [
18
]. A popular approach
has been to use mobile phones for detecting potential viewers’ proximity to a screen (e.g., using
Bluetooth) and to tailor the screen’s content according to the individuals in front of them [
15
,
28
,
29
,
39
]. Within commercial settings, another common practice adjusts content based on coarse-
grained audience demographics captured through video analytics [
27
]. However, to date few
attempts have been made at deploying and evaluating display personalisation systems in real-world
settings. As a result, limited information exists about best practices to follow, potential pitfalls
that need to be avoided when deploying display personalisation systems and long-term use of
this personalisation technology. Understanding these factors is fundamental to improving the
eectiveness of personalised pervasive displays in long-term use, and identifying key open research
areas.
This paper contributes by rigorously investigating issues surrounding eld deployments of per-
sonalised pervasive displays and draws on our experiences of instantiating display personalisation
as a service (c.f. a controlled experiment) on a large university campus. We report data from an
ongoing deployment at Lancaster University, with detailed analysis of a period of 165 days during
which time we supported 24
,
673 requests for personalised content across 44 displays, allowing us
to report longitudinal experiences of supporting display personalisation at scale. Our deployment
relies on the Tacita architectural model proposed by Davies et al. [
15
], which we extend to support
our large-scale eld trials. Where possible, we contrast our ndings with those obtained from
controlled laboratory studies, highlighting key discrepancies between real world deployments and
the controlled laboratory tests. Our key contributions are:
Firstly
, we provide the community’s rst robust analysis of how viewers use display personali-
sation systems in real-world settings. We draw on measurements collected from our deployment at
Lancaster University and analyse content requests made by users during the deployment. Analysing
the distribution of requests across content categories, and spatio-temporal patterns across the cam-
pus environment, allows an evidenced investigation of whether people use display personalisation
and, if they do, what exactly they choose to see on the displays. Our results show that people
are indeed willing to exploit display personalisation to provide easy access to relevant factual
information (e.g. real-time transport updates) while rejecting the use of displays for content such
as social media. We also examine the level of engagement users have with our system by looking
at the persistence of content requests over time. Our results indicate that initially our display
personalisation system can indeed reach a high conversion rate, but that the level of engagement
starts to dwindle signicantly over time. In particular, only 40% of users continue issuing content
requests after three weeks. We further highlight the challenges deployments face in collecting
long-term usage data in the absence of reliable sources of information on content activation, and
provide detailed insights regarding the long-term maintenance of a user base. There have been
no previously published studies of how viewers use display personalisation systems in real-world
settings.
Secondly
, we examine the adequacy of a commercial state-of-the-art indoor location tracking
technology based on Bluetooth Low Energy beacons (iBeacon) to support timely personalisation
of displays for viewers. Using a combination of data collected from our deployments and con-
trolled small-scale trials conducted in a laboratory setting we demonstrate that current commercial
technologies can detect entry to display regions with reasonable accuracy, but they are poor at
ACM Trans. Comput.-Hum. Interact., Vol. 1, No. 1, Article 1. Publication date: January 2020.

A Longitudinal Study of Pervasive Display Personalisation 1:3
estimating when a user leaves the region and hence how long the user spent in the vicinity of a
display. We also show that the accuracy of exit detection can be improved using custom ranging
techniques, but even in such cases the performance is sensitive to the conguration of the proximity
detection technologies and the nature of the deployment environment. These ndings are important
because they impact on multiple uses of BLE including the perceived reliability of location-based
display analytics and metrics that are used to gauge the success (or failure) of specic display
campaigns. They also represent a signicant advance of prior studies which have not considered
the use of ranging techniques and have focused on observing performance in single installations.
Thirdly
, we report on the results of a series of structured interviews designed to elicit viewers’
expectations and attitudes together with those of stakeholders responsible for providing content.
We found that most viewers were positive about display personalisation despite privacy concerns.
Content creators appear to see increasing value in display personalisation, especially in terms
of using displays to reach specic groups of viewers and we did not observe the same concerns
regarding the need to identify and potentially moderate personalisation requests that were reported
by Clinch et al. [
11
] thus representing a signicant change in our understanding of attitudes to
personalisation.
Finally
, we reect on our experiences and ndings, providing insights for future deployments
and highlighting open research challenges. We specically highlight changes that are required
to the reported state-of-the-art in personalisation architectures in order to support long-term
deployments at scale.
2 RELATED WORK
There is a rich history of research into pervasive displays dating back to the 1980s [
13
]. This
research includes a multitude of focal areas including, for example, display hardware, interaction
modalities and audience behaviour. For the purposes of this paper, we concern ourselves only with
prior work relating to deployments of pervasive displays and with display personalisation.
2.1 Long-Lived Pervasive Display Deployments
Early research into display deployments explored their use as ‘media links’, i.e. using video and
audio links to connect together physically separate spaces. For example, Kit Galloway and Sherrie
Rabinowitz created the “Hole-In-Space [
21
], a three day art installation in November 1980. The
installation featured two large back-projected displays (plus speakers and cameras) installed in
sidewalk-facing windows of the Lincoln Center for the Performing Arts in New York City and
“The Broadway” department store in Los Angeles. A satellite link between the two cities allowed
the creation of virtual windows in which the video feed of New York was shown on the screen in
L.A. and the video from L.A. in New York. Displays providing media links were also deployed in
(research) workplace settings. The Xerox PARC Media Spaces [
7
,
23
] connected researchers at sites
in Palo Alto and Portland by providing steerable video and audio links in the common area” of
each site. The media links ran 24 hours a day, seven days a week for over two years, nishing only
when the oces in Portland closed. Whilst originally intended to support formal meetings, the
majority of interactions over the links were chance encounters lasting for less than ve minutes. A
similar system at Bellcore Labs, the VideoWindow [
19
], connected researchers on two dierent
oors of the building using large projected displays in common areas.
Pervasive display research has often involved deployments to help explore user behaviour outside
the laboratory. Much of the early research into display deployments focused on urban environments,
e.g. “CityWall” [
47
] that consisted of a touch-enabled display situated in a city centre that showed
content relevant to the context of the deployment, e.g. images and videos tagged with the location.
Viewers were able to use both gestures and touch interaction to reorder and scale media items shown
ACM Trans. Comput.-Hum. Interact., Vol. 1, No. 1, Article 1. Publication date: January 2020.

1:4 M. Mikusz et al.
on the display. Deploying displays in urban spaces to improve engagement within a community
has also been explored by Schroeter et al. [50] who conducted a set of deployments of interactive
displays at bus stops, museums and conferences. Taylor et al
. [56]
utilised interactive public displays
as a way of providing situated voting devices for communities; a two-month deployment led to the
identication of a set of guidelines specic to the design of democracy tools. A larger research-
based deployment of public displays was conducted by José et al
. [29]
who placed a total of 10
displays in various locations including a university, schools and cafes. The authors deployed a set of
applications and content, investigating how user-generated content can improve the use of situated
displays in urban settings. A university deployment was also used by Greis et al
. [25]
, whose three
displays were used to investigate the impact of delays in the moderation of user-generated content,
i.e. the time between the submission of a content item and the time at which the content appeared
on the screen.
Beyond urban environments, Taylor and Cheverst
[55]
deployed the “Wray Photo Display”,
a digital display in a rural village. The deployment consisted of a small number of interactive
touch-enabled displays deployed in key locations within the village including a bookshop and
the village hall. The displays allowed residents to access photos of recent events, information
about events in the future, and to use Bluetooth to upload custom content for other members
of the community. An extended deployment allowed researchers to explore which content their
community of viewers shared and viewed.
One of the largest research-based deployments of pervasive displays was the UBI-Hotspots
system, a network of interactive touch-enabled displays located across the city centre of Oulu,
Finland [
43
]. The deployment consisted of up to twelve in- and outdoor displays and served as a
platform for research into pervasive computing and human computer interaction. UBI-Hotspots
provided easy deployment of web-based display applications to the entire display network which
become immediately accessible by pedestrians and passers-by. In addition to touch-based interaction,
UBI-Hotspots also supported a level of explicit personalisation users were able to register and
authenticate themselves at the display via an RFID tag and, for example, participate in games,
post messages to bulletin boards, and retrieve information such as bus departure times. Whilst
UBI-Hotspots was one of the longest running research-based deployments, the number of displays
has gradually reduced and the system has now been decommissioned [26].
Large-scale ‘in-the-wild’ deployments of public displays (both in terms of number of displays
and physical size of individual displays) are more common as part of commercial display networks
and are typically driven by commercial entities such as advertisement companies. For example,
LinkNYC
1
is a recent example of a large-scale public display deployment across New York City in
which old telephone boxes have been transformed to modern, interaction-enabled public display
kiosks delivering adverts and allowing passers-by to access city-related services and directions.
The deployment consists of, to date, over 7
,
500 displays and has started to expand to other cities
such as London (branded as LinkUK
2
). Previous commercial display deployments include the BBC
Big Screen featuring large display installations situated in over 21 cities across the U.K. to show
major events. Neither LinkNYC, LinkLondon, nor BBC Big Screens attempted to oer personalised
content, but were part of generalised attempts to “transform our urban environments” [31].
Our research builds on the e-Campus infrastructure [
20
], the world’s largest research-focussed
display network, currently consisting of over 85 displays situated across the campus at Lancaster
University. Whilst the initial e-Campus infrastructure consisted of a set of displays showing largely
1
https://www.link.nyc
2
https://www.inlinkuk.com
ACM Trans. Comput.-Hum. Interact., Vol. 1, No. 1, Article 1. Publication date: January 2020.

A Longitudinal Study of Pervasive Display Personalisation 1:5
static content (e.g. slideshows), the network and functionality has since been substantially extended.
We describe e-Campus and relevant extensions in subsequent sections of this paper.
2.2 Display Personalisation
While the falling cost of hardware and the diculty of reaching the general public through frag-
mented conventional media has led to the deployment of increasing numbers of public display
systems, the vast majority of today’s public displays eectively disappear: people have become
so accustomed to their low utility that they have become highly skilled at ignoring them [
34
,
41
].
One approach to tackling this problem is the introduction of personalised content. Our previous
explorations in this domain led to the identication of three distinct classes of display personalisa-
tion [15]:
Walk-by personalisation
in which viewers passing by a single display see content that is relevant
to them (as exemplied in the 2002 lm Minority Report in which the characters are subject
to personalised adverts as they journey across the city).
Longitudinal personalisation
in which viewer preferences for personalised content are realised
as a shift in content on multiple screens in a given geographic area, accommodating prefer-
ences from multiple viewers, typically over an extended period of time. In practical terms this
might mean, e.g., that the content shown on the displays at a university campus automatically
changes during vacation time or that content in a shopping mall adjusts during weekends or
school holidays.
Active personalisation
in which users (inter-)actively engage with a display system to control
personalised applications on a nearby display, e.g. to extend a mobile phone display for better
viewing of complex data.
The selection of appropriate content may be based either on explicit user preferences (as in [
15
,
18
,
43
] or determined implicitly based on contextual information about the viewer (or group of viewers)
currently present in front of a display [
36
]. In this paper we focus on explicit user preferences to
determine content personalisation.
The idea of explicitly personalising public displays as users walk by was rst suggested by
Finney et al. [
18
], who used Active Badges to trigger personalised content such as unread email
messages on nearby displays. Russell and Gossweiler [
49
] investigated the use of public displays
for the delivery of personalised content in combination with an appropriate way to identify
and authenticate the viewer. Their work was one of the early examples to support ‘walk-up
personalisation’ (in contrast to walk by) due to the requirement to explicitly interact and request
personalised content by walking up to the display to authenticate. Other systems have used
IR [
33
], RFID tags [
48
], or custom-built wireless devices [
57
] for proximity detection. Several
systems have explored oering more explicit control in display personalisation. For example,
InstantPlaces [
28
] allowed users to send pictures to a portion of the display allocated to their device,
whilst e-Campus [
14
] and BlueTone [
16
] allowed viewers to take control of the display for one of a
number of predened applications using Bluetooth. A more indirect form of personalisation was
proposed by Müller and Krüger [
40
] in which the system estimates viewers’ paths between displays
to coordinate content across multiple displays. Greenberg et al.’s proxemic interactions [
24
] use
vision-based motion capture to track users’ paths, feeding this information to a display app that
can thus continuously adapt its output. A simpler adaptation is used by Tafreshi et al. [
52
], who
proposed a responsive design approach to public display applications that takes viewer distances
and numbers into account.
In addition to proximity, public displays can also be personalised based on the user’s absolute
location or spatial orientation. AT&T Cambridge’s classic “Sentient Computing” project [
2
] used
ACM Trans. Comput.-Hum. Interact., Vol. 1, No. 1, Article 1. Publication date: January 2020.

Citations
More filters
Journal ArticleDOI
TL;DR: In this article , the authors propose a set of design goals for the implementation and deployment of engaging interactive public display applications in educational settings. And they provide clear guidelines for the design of IPDs for schools by making design teams and stakeholders focus on factors fostering user adoption, social interactions and collaboration, also opened up paths for further explorations regarding display awareness, level of commitment in interactions, the displays integration into structured activities, and display management at the educational institutions.
Abstract: Abstract Interactive Public Displays (IPD) enable new ways of interaction as well as communication channels, extending online communities into physical places and supporting a culture of participation. While educational environments have seen how new digital technologies can enhance learning activities beyond the traditional classroom context, the use of IPDs is still an area insufficiently explored. This paper proposes a set of design goals for the implementation and deployment of engaging interactive public display applications in educational settings. Based on findings from a series of design workshops and two deployment studies in authentic settings, seven design goals were identified and defined. The design goals provide clear guidelines for the design of IPDs for schools by making design teams and stakeholders focus on factors fostering user adoption, social interactions and collaboration. The design goals also opened up paths for further explorations regarding display awareness, level of commitment in interactions, the displays’ integration into structured activities, and display management at the educational institutions.
Proceedings ArticleDOI
17 Oct 2022
TL;DR: A snowballing procedure and systematic literature mapping are conducted to understand whether and how long-term UX evaluations have been conducted in IoT and the challenges and opportunities in conducting longitudinal studies in this field.
Abstract: User Experience (UX) is affected by users' emotions and motivations, among other subjective factors. Hence, changes in UX occur over time, according to the different contexts, and behavioral and internal user aspects. With the rising development of Internet of Things (IoT) systems to support daily activities and, at the same time, the increasing abandonment of these systems, the importance of ensuring good UX levels over time has become an even greater need. In this context, we conducted a snowballing procedure and systematic literature mapping to understand whether and how long-term UX evaluations have been conducted in IoT and the challenges and opportunities in conducting longitudinal studies in this field. We analyzed 55 papers that reported longitudinal evaluations in different IoT systems and 14 papers that reported UX evaluations in the IoT context. We also interviewed five software engineering professionals to learn about their IoT user-testing and evaluation perspectives. Our results provide a panorama of the conduction of longitudinal IoT studies and a glance at the directions the research on UX evaluation for IoT systems is taking. Hence, we indicate future research opportunities for evaluating long-term UX in IoT systems through the longitudinal approach, considering the complexity of these systems and their impact on users' everyday lives.
Proceedings ArticleDOI
09 Mar 2022
TL;DR: In this article , the authors explore how serendipity for remote workers can be created by leveraging IoT technologies, edge computing, high-resolution video, network protocols for live interaction, and video/audio denaturing.
Abstract: Unplanned encounters or casual collisions between colleagues have long been recognized as catalysts for creativity and innovation. The absence of such encounters has been a negative side effect of COVID-enforced remote work. However, there have also been positive side effects such as less time lost to commutes, lower carbon footprints, and improved work-life balance. This vision paper explores how serendipity for remote workers can be created by leveraging IoT technologies, edge computing, high-resolution video, network protocols for live interaction, and video/audio denaturing. We reflect on the privacy issues that technology-mediated serendipity raises and sketch a path towards honoring diverse privacy preferences.
References
More filters
Journal ArticleDOI
Sara Bly1, Steve Harrison1, Susan Irwin1

735 citations


"A Longitudinal Study of Pervasive D..." refers background in this paper

  • ...The Xerox PARC Media Spaces [7, 23] connected researchers at sites in Palo Alto and Portland by providing steerable video and audio links in the “common area” of each site....

    [...]

Journal ArticleDOI
TL;DR: An enhanced version of AT&T Laboratories Cambridge's sentient computing system, which uses sensors to update a model of the real world, is installed throughout an office building.
Abstract: Sentient computing systems, which can change their behaviour based on a model of the environment they construct using sensor data, may hold the key to managing tomorrow's device-rich mobile networks. At AT&T Laboratories Cambridge, we have built a system that uses sensors to update a model of the real world. We designed the model's terms (object positions, descriptions and state, and so forth) to be immediately familiar to users. Thus, the model describes the world much as users themselves would. We can use this model to write programs that react to changes in the environment according to the user's preferences. We call this sentient computing because the applications appear to share the user's perception of the environment. Treating the current state of the environment as common ground between computers and users provides new ways of interacting with information systems. A sentient computing system doesn't need to be intelligent or capable of forming new concepts about the world, it only needs to act as though its perceptions duplicate the user's. In earlier work, we described a prototype of this system and stated our intention to deploy it on a large scale. We have now installed an enhanced version throughout an office building. Over the past year, approximately 50 staff members have used the system daily with a set of trial applications.

531 citations


Additional excerpts

  • ...AT&T Cambridge’s classic “Sentient Computing” project [2] used...

    [...]

Proceedings ArticleDOI
01 Sep 1990
TL;DR: The goal of the VideoWindow system project is to extend a shared space over considerable distance without impairing the quality of the interactions among users or requiring any special actions to establish a conversation.
Abstract: Imagine sitting in your work place lounge having coffee with some colleagues. Now imagine that you and your colleagues are still in the same room, but are separated by a large sheet of glass that does not interfere with your ability to carry on a clear, two-way conversation. Finally, imagine that you have split the room into two parts and moved one part 50 miles down the road, without impairing the quality of your interaction with your friends. That scenario illustrates the goal of the VideoWindow system project: to extend a shared space over considerable distance without impairing the quality of the interactions among users or requiring any special actions to establish a conversation. While the VideoWindow system -a very large screen, full duplex teleconferencing technology that we will describe later in this paper -cannot yet achieve this goal, we believe it can come closer to it than any other system yet invented.

338 citations


Additional excerpts

  • ...A similar system at Bellcore Labs, the VideoWindow [19], connected researchers on two different floors of the building using large projected displays in common areas....

    [...]

Book ChapterDOI
11 May 2009
TL;DR: It is shown how audience expectations towards what is presented on public displays can correlate with their attention towards these displays and possible solutions to overcome this "Display Blindness" and increase audience attention towards public displays are proposed.
Abstract: In this paper we show how audience expectations towards what is presented on public displays can correlate with their attention towards these displays. Similar to the effect of Banner Blindness on the Web, displays for which users expect uninteresting content (e.g. advertisements) are often ignored. We investigate this effect in two studies. In the first, interviews with 91 users at 11 different public displays revealed that for most public displays, the audience expects boring advertisements and so ignores the displays. This was exemplified by the inclusion of two of our own displays. One, the iDisplay, which showed information for students, was looked at more often than the other (MobiDiC) which showed coupons for shops. In a second study, we conducted repertory grid interviews with 17 users to identify the dimensions that users believe to influence whether they look at public displays. We propose possible solutions to overcome this "Display Blindness" and increase audience attention towards public displays.

303 citations


"A Longitudinal Study of Pervasive D..." refers background in this paper

  • ...While the falling cost of hardware and the difficulty of reaching the general public through fragmented conventional media has led to the deployment of increasing numbers of public display systems, the vast majority of today’s public displays effectively disappear: people have become so accustomed to their low utility that they have become highly skilled at ignoring them [34, 41]....

    [...]

  • ...Indeed, systematic observations of viewer behaviour in the proximity of digital signs suggest that viewers exhibit a phenomenon known as display blindness – choosing to ignore public displays because they perceive them as having little content of relevance [34, 41]....

    [...]

Frequently Asked Questions (11)
Q1. What are the contributions mentioned in the paper "A longitudinal study of pervasive display personalisation" ?

In this paper the authors report on their experiences of designing, developing and operating the world ’ s first comprehensive display personalisation service for mobile users. Through a set of rigorous quantitative measures and eleven potential user/stakeholder interviews, the authors demonstrate the success of the platform in realising display personalisation, and offer a series of reflections to inform the design of future systems. 

In addition to the observations that emerged through the process of building and deploying Tacita, this paper also examines usage of display personalisation when made available in an established deployment for an extended period. Their results clearly show that a portion of users are prepared to engage with a system such as Tacita and that they continue to do so for extended periods of time. While in their deployment transport content was the most popular, the authors would not wish to argue that travel apps themselves are the “ killer application ”. Considering their in-the-wild deployment, the authors were able to capture a unique set of performance data providing first-of-a-kind insights into the feasibility of a display personalisation system using BLE beacons under real-world conditions over an extended period of time. 

the insights into dwell times at displays can be used to inform the selection of appropriate content durations or even the location of new displays. 

Previous commercial display deployments include the BBC Big Screen featuring large display installations situated in over 21 cities across the U.K. to show major events. 

The authors also evaluated the performance of leaving the proximity of a beacon by capturing the time delay between deactivating the beacon transmission and the point at which the beacon was detected as “lost” by the mobile device. 

Their results indicate that initially their display personalisation system can indeed reach a high conversion rate, but that the level of engagement starts to dwindle significantly over time. 

The authors partially mitigate against this by only displaying textual tweets and by screening these tweets using a simple “bad word” filter. 

At the same time, it would maintain the priorities held by other core stakeholders, by providing viewing guarantees and, for the first time, allowing those stakeholders to establish the impact of their content on viewer behaviour. 

As a result of the framework, creating a typical Tacita Channel that serves a static web page to users as they pass by displays requires less than 20 lines of Python (the library itself comprises approximately 650 lines). 

the Display Gateway provides a layer of abstraction over underlying signage networks such that deployments of Tacita into new networks now only require modification to a single architectural component. 

By contrast, research systems have experimented with display personalisation for over twenty years [18], but the average duration of any display system (with and without personalisation) is of the order of days or weeks rather than months.