Where, Wherefore, and How?

Contrasting Two Surveillance Contexts According to Acceptance

Julia van Heek, Katrin Arning, and Martina Ziefle

Human-Computer Interaction Center, RWTH Aachen University, Campus Boulevard 57, 52074 Aachen, Germany

Keywords: Surveillance Technologies, Medical Vs, Crime Surveillance, Technology Acceptance, User Diversity, Con-

joint Analysis.

Abstract: Surveillance technologies are used all over the world for various reasons. In urban environments, surveil-

lance technologies are predominantly used for detecting or preventing crimes. Simultaneously, an increasing

number of technologies are used for medical monitoring at home, but also at clinical facilities, and at public

environments for assuring patients’ medical safety. An intensive policy discussion about perceived ad-

vantages (especially increasing safety) and perceived barriers (in particular the invasion of privacy) comes

along with the use of surveillance technologies. In this paper, it is examined where and for which contexts

the use of surveillance technologies is accepted and under which conditions safety or privacy is perceived as

more important. We investigate the acceptance of surveillance technologies for medical and crime surveil-

lance scenarios using a conjoint analysis approach including four relevant aspects: location of surveillance,

increase in safety, invasion of privacy, and the applied camera type. Results show both, context independent

findings as well as context-sensitive findings: e.g., for crime surveillance, the location is most important fol-

lowed by the trade-off between privacy and safety, while these three factors are of similar importance for

medical surveillance. From a practical viewpoint, the findings might contribute to a differentiated surveil-

lance policy in cities.

1 INTRODUCTION

Nowadays, modern societies face major challenges

in order to cope with on-going urbanization de-

mands: especially the demographic change, which

require novel care and adequate supply concepts as

well as enabling of living in smart, safe, and sustain-

able cities. Along with this comes the increasing

development that higher proportions of people will

live in cities than in all other regions. These substan-

tial urbanization processes lead to consecutive chal-

lenges, which are difficult to balance. Apart from

healthcare, economy, mobility, or governance issues,

the implementation of safe and also well-accepted

technical infrastructures and smart city concepts is

focused worldwide (Ziefle et al., 2014).

Most large cities around the world use surveil-

lance technologies (primarily video surveillance via

cameras) in order to prevent and detect crime for

increasing safety in cities and particularly at public

locations (Dailey, 2013; Barrett, 2013). Progressive-

ly, surveillance technologies in smart city concepts

(Filipponi et al., 2010; Dey et al., 2012) are increas-

ingly connected, integrated, and implemented. This

is especially driven by the motive to increase per-

ceived safety for city residents as this is the essential

condition for the participation in social and econom-

ic life and thus, it is a valuable asset for cities.

In spite of this undisputed positive aspect of sur-

veillance, the violation of public’s or city resident’s

privacy through recording, storage, and processing

of (video) data represents the most discussed draw-

back and barrier of using surveillance technologies

(Patton, 2000). The balance between both poles –

privacy as a value and safety as a value – is quite

intricate. A wide spectrum of resident’s needs, espe-

cially their attitudes towards and requirements for

safety and privacy have to be considered during the

development and implementation of smart city con-

cepts.

This trade-off is not only discussed for the crime

surveillance context, but also matters in the context

of medical surveillance. While surveillance technol-

ogies were mainly used for crime detection reasons

in recent decades, they have been increasingly used

for medical surveillance since the last years. Surveil-

lance technologies - in particular AAL systems

Heek, J., Arning, K. and Zieﬂe, M.

Where, Wherefore, and How? - Contrasting Two Surveillance Contexts According to Acceptance.

DOI: 10.5220/0006362400870098

In Proceedings of the 6th International Conference on Smart Cities and Green ICT Systems (SMARTGREENS 2017), pages 87-98

ISBN: 978-989-758-241-7

(Ambient Assisted Living) - become more and more

popular in the context of chronic illness and medical

monitoring (Leonhardt, 2006; Klack et al., 2011). In

the medical context, one of the major goals of sur-

veillance or monitoring systems is the detection of

emergencies - especially falls - as falls are one of the

most serious health risks for seniors (Rubenstein,

2006). Thus, medical surveillance technologies do

increase not only perceived but also factual safety as

a fast detection of emergencies and falls and consec-

utively immediate help reduce the risk of death by

more than 80% (Gurley et al., 1996). Opposing this

benefit of medical surveillance, there are concerns

about an invasion of the own privacy in terms of a

protection of personal information, e.g., anonymity,

self-determination, personal control of data (Patton,

2000; Wilkowska & Ziefle, 2012). Thus, surveil-

lance generally represents a conflict between two

key motives: on the one hand, the wish to be safe as

well as increase safety in urban environments and,

on the other hand, the wish to stay private and to

protect the own privacy. This conflict is difficult and

almost impossible to address for city planners with-

out understanding the residents’ individual ac-

ceptance of surveillance technologies.

In this paper, the acceptance of surveillance

technologies is empirically investigated differentiat-

ing between two urban contexts. Using a conjoint

analysis, the critical trade-off between privacy and

safety is empirically addressed as well as different

locations of surveillance and different camera types

are taken into account. Comparing a medical and a

surveillance context allows to find out if the ac-

ceptance of surveillance technologies is a context-

sensitive phenomenon. This way, it is examined to

what extent city residents’ evaluation of the trade-off

between safety and privacy varies depending on the

context, the type of technology, and the city location

in which surveillance is applied.

2 ACCEPTANCE OF

SURVEILLANCE SYSTEMS

Currently, more and more different surveillance

technologies are used and integrated into surveil-

lance concepts not only to generate smart houses,

but also to build smart city networks. Thus, this

section presents an overview of the technologies that

are applied for different surveillance contexts as well

as their advantages and drawbacks (2.1). Further, the

current state of the art concerning research on the

acceptance of surveillance is detailed focusing on

which technologies are accepted at which locations,

for which context, and under which conditions (2.2).

2.1 Context-specific Use of Surveillance

Technologies

For the context of crime surveillance, the currently

most applied safety measure in urban environments

is video surveillance (Koh et al., 2016). Video sur-

veillance strongly differs with respect to the camera

type enabling diverse functions (compounded with

other technologies) such as face recognition or pre-

cise location determination. These types of video

surveillance are used in many cities exhaustively

enabling a detailed tracking of criminals and detec-

tion of crimes (La Vigne et al., 2011). Some theoret-

ical research approaches focus also on the integra-

tion of crime surveillance into smart city concepts

(Fyfe, 2004; Dey et al., 2012). So far, many studies

investigated the effectiveness of safety measures

such as improved lighting of public places or video

surveillance for enhancing safety especially at public

high-frequented locations (e.g., La Vigne et al.,

2011, Welsh & Farrington, 2004; Welsh et al.,

2015): all of them found an increased safety percep-

tion by city residents but also a higher “real” safety

in terms of crime reduction and higher rates of crime

detection. In contrast, an invasion of privacy is the

main drawback of (video-based) crime surveillance

as recording, processing, and storage of data is heav-

ily scrutinized (e.g., Welsh et al., 2015; Schwartz,

2012).

For medical reasons, surveillance technologies

were mainly used in the private home environment,

in health facilities, and retirement homes in the last

years. A wide variety of AAL technologies exists

mainly developed to detect falls and enable fast

assistance in emergencies (Cardinaux et al., 2011;

Memon et al., 2014). Using video-based technolo-

gies is in particular advantageous as it enables con-

tactless observation without the necessity to equip

patients with further technologies (e.g., help button,

tag, sensors). Thus, a high number of video-based

health surveillance systems was developed to date

that primarily focus on fall detection in order to

increase safety (e.g., Fleck & Strasser, 2008; Yu et

al., 2012; Chen et al., 2013). Although AAL tech-

nologies were predominantly used at private envi-

ronments, it is reasonable to integrate medical sur-

veillance technologies at public urban environments

to expand existing (crime) surveillance technologies

with regard to medical surveillance as a high number

of emergencies happen in public. In accordance with

demographic change and an aging society, the needs

SMARTGREENS 2017 - 6th International Conference on Smart Cities and Green ICT Systems

for public medical surveillance in cities increase

continuously. Hence, there are also first approaches

to integrate medical or health surveillance into smart

city concepts in order to increase medical safety

(Solanas et al., 2014).

Besides the main benefit of increased safety, pri-

vacy concerns are of major importance in the health

or medical surveillance context (Wilkowska & Zief-

le, 2012). In particular, privacy is a highly relevant

topic for video surveillance, since people in the area

covered by cameras have no option to avoid being

monitored. There are some technical approaches that

try to integrate measures to protect privacy within

the surveillance systems, e.g., by recording and

storage of video data only if accidents were detected

(Aghajan et al., 2007).

Summarizing so far, the use of surveillance tech-

nologies implies a sensitive and critical trade-off

between an increase in safety and the protection of

privacy in a crime as well as a medical surveillance

context.

2.2 Technology Acceptance

Usually, the development and integration of surveil-

lance technologies is a policy decision of planning

authorities in cities and communes without integrat-

ing residents during the technical development and

decision-making process. However, the controver-

sial public discussion about surveillance shows that

it might be supportive to consider resident’s needs

and concerns in early phases of the decision process

in order to enable a pleased and safe life in cities

(Slobogin, 2003; Surette, 2005). While there are a

plenty of studies focusing on technical characteris-

tics of surveillance or the efficiency of applied sur-

veillance technologies (e.g., La Vigne et al., 2011;

Welsh et al., 2015), sparse social science research is

available in this regard. Previous acceptance studies

primarily focus on single surveillance contexts: a

study on crime surveillance acceptance found high-

est acceptance scores for video surveillance at highly

frequented public places, such as train stations, as

well as comparatively higher needs for safety at

public and higher privacy needs at private environ-

ments (van Heek et al., 2015). Another single case

study focused on medical surveillance and also

found higher acceptance scores for video surveil-

lance at public environments. However, general

acceptance for medical surveillance was rather low

due to strong privacy concerns of the participants

(Himmel et al., 2013).

A direct comparison of different surveillance

contexts has not been investigated so far. It would be

useful to analyze whether the acceptance of surveil-

lance technologies and the evaluation of the trade-

off between safety and privacy is a generic or a

context-sensitive phenomenon. As privacy and safe-

ty proved to be important factors of surveillance

acceptance (see 2.1), their understanding is essential

for a successful adoption of surveillance technolo-

gies (Rogers, 2003).

The Technology Acceptance Model (TAM) is

the most known and best-established model explain-

ing and predicting the adoption of technologies and

serves as a basis for numerous subsequent and

adapted models (Davis et al., 1989). However, these

acceptance models cannot be simply applied to the

acceptance of video-based surveillance technologies:

A first reason is that conventional acceptance mod-

els enable an evaluation of complete technologies,

systems, or applications, but they do not allow an

evaluation of single technical functions or character-

istics of a system. Hence, it is not possible to derive

which characteristics of a system lead to adoption or

non-adoption of the complete system yet. As a re-

sult, it is also not possible to derive concrete design

guidelines, e.g., where, how, and under which condi-

tions a video-based surveillance system should be

used. Secondly, questionnaires, designed on the

basis of TAM or adapted acceptance models, do not

allow a simulation of complex decision scenarios, in

which several decision criteria are weighted against

each other. Hence, it is also not possible to infer

statements about relative importance, relationships,

and interactions of several factors.

Summarizing, acceptance, conceptualization, and

integration of video-based surveillance systems in

smart cities could be improved, if designers and

(city) planners could revert on city residents’ prefer-

ences. Thus, the goal of our study was to capture

preferences for video-based surveillance scenarios at

different locations (private vs. public), under consid-

eration of different camera types, benefits in terms

of increased safety, and privacy concerns due to

different data handling purposes. By applying a

conjoint analysis, decision scenarios were simulated

and different attributes’ acceptance as well as their

interrelations were analyzed in detail. In order to

fulfill a direct comparison of surveillance context,

the approach was carried out for a medical (study 1)

and a crime surveillance (study 2) context.

3 METHODOLOGY

In order to understand if acceptance for surveillance

depends on the context, the conjoint analysis ap-

Where, Wherefore, and How? - Contrasting Two Surveillance Contexts According to Acceptance

proach was applied to a medical and a crime surveil-

lance scenario. The factorial design of the conjoint

analysis in both contexts (surveillance vs. medical)

included four attributes that had been identified as

important impact factors on surveillance acceptance

in preceding studies (van Heek et al., 2015; Arning

& Ziefle, 2015): 1) locations of surveillance, 2)

increase in safety operationalized as detection rates

of crimes and medical emergencies, 3) privacy in

terms of different handlings of the recorded data

material, and 4) different camera types. These attrib-

utes were used to identify the most important pa-

rameters and to examine to which extent surveil-

lance scenario decisions based on these attributes

were linked to the surveillance context.

3.1 Conjoint Analysis

Conjoint analyses were developed by Luce and Tuk-

ey in the 1960s and combine a measurement model

with a statistical estimation algorithm (Luce & Tuk-

ey, 1964). Within a conjoint analysis, respondents

assess specific product or scenario configurations

that consist of different attributes and differ from

each other in the attribute levels. This way, conjoint

analyses go beyond the possibilities of conventional

survey-based research approaches: they enable not

only an evaluation of single product or scenario

characteristics, but allow a holistic evaluation of

decision scenarios, a weighting of different attributes

against each other, and a direct simulation of rela-

tionships and interactions (Orme, 2010). Decision

processes and scenario preferences can be simulated

and separated into part-worth utilities of the attrib-

utes and their levels. In this process, the relative

importance of attributes delivers information about

which attribute affects the participants’ choice the

most. The part-worth utilities characterize the most

important or unimportant attribute levels. Further,

preference ratings and preference shares can be

consulted as indicators of acceptance.

For this study, a choice-based conjoint analysis

approach (CBC) was chosen in order to analyse

scenario decisions in which - most probably - more

than one attribute affects the final respondent’s

choice (Sawtooth Software, 2009).

3.2 Attributes and Attribute Levels

The identification and selection procedure of attrib-

utes and attribute levels is the first - and highly im-

portant - step for the conceptualization of a conjoint

analysis, since it affects the generalizability, validity,

and significance of the findings (Rao, 2014). It has

to be ensured that all attributes are considered that

are relevant for the preferences of respondents as

well as for city-planers, policy-makers, or other

important stakeholders. In order to identify the at-

tributes, the results of extensive literature analyses

and preceding studies were used - in which relevant

parameters for the acceptance of surveillance tech-

nologies (crime surveillance (van Heek et al., 2015)

and medical surveillance context (Arning & Ziefle,

2015)) were identified as a basis for the selection of

attributes and levels.

The following four attributes were assessed in

our conjoint study: locations of surveillance, camera

types, increase in safety, and invasion of privacy.

Within the attribute locations, the private home

environment was contrasted with different public

locations as place for camera installation: store,

market, and train station.

Increase in Safety as major benefit of imple-

mented surveillance technologies was operational-

ized as different detection rates of crimes in the

crime surveillance and of medical emergencies in

the medical surveillance context. The attribute levels

were specified as detection rates of 0% (no im-

provement), 5%, 10%, and 20% in both contexts.

Invasion of Privacy as major concern of imple-

mented video-based surveillance technologies was

operationalized as different ways and intensities to

handle and process recorded video data and included

the following levels: merely archiving data (police

and patient data bases), storage in profile data bases

(open to institutions, e.g., health insurance compa-

nies, security services), enabling position determina-

tion, and allowing face recognition.

As last attribute Camera Type was integrated in-

to the study differing in size, visibility, an obtrusive-

ness of the technology: a conventional, clearly visi-

ble, large and tracking camera, a large and visible

dome-camera, a small mini-dome-camera, and a

hidden, integrated, not visible camera.

3.3 Conjoint Questionnaire Design

In both contexts - medical (study 1) and crime (study

2) - identical questionnaire designs were developed

using the SSI web Sawtooth Software (SSI Web,

2016) and consisted of four parts.

The first part addressed demographic characteris-

tics such as age, gender, educational level, type of

residence, and area of residence. Afterwards, the

participants had to answer some context-specific

questions. For the medical surveillance context, the

participants were asked for details concerning health

status and experience with medical emergencies. For

SMARTGREENS 2017 - 6th International Conference on Smart Cities and Green ICT Systems

Figure 1: Example of a choice task (crime surveillance context).

the crime surveillance context, the participants indi-

cated if they have already fallen victim to offenses.

In the second part, the respondents indicated

their perceived threat (of crimes or medical emer-

gencies) and their needs for privacy (measured each

with four items on six-point Likert scales). The

items were checked for reliability and subsequently

summed up resulting in a “need for privacy” and a

“perceived threat” score.

Next, the participants were introduced in the re-

spective scenario. The medical scenario dealt with

the installation of video cameras for medical surveil-

lance purposes at different locations. The cameras

were able to sent an emergency signal to a medical

institution. Similarly, the cameras introduced in the

crime surveillance scenario were also installed at

different locations and should help to detect and

prevent crime by enabling to send alarm signals to

security institutions.

In the fourth part, the CBC choice tasks with four

attributes and each four levels were presented. As a

control, the participants were asked to imagine that

they would be alone during the day. Then, they

should decide under which conditions and at which

locations they would accept the installation of video

cameras and were instructed to select the scenario in

each choice task, that met their individual needs

most closely. An example for a choice task is shown

in Fig. 1. As a combination of all attribute levels

would have let to 256 possible combinations

(4x4x4x4), the number of choice tasks was reduced

and overall, 10 random and one fixed task were

presented to the participants. A test of design effi-

ciency confirmed that the reduced, randomized de-

sign was sufficient and comparable to the hypothet-

ical orthogonal design regarding a sample of at least

100 respondents.

3.4 Data Analysis

For analysing the conjoint data (i.e. estimation of

part-worth utilities, preference simulations) the Saw-

tooth Software was used (SSI Web, 2016)(SMRT,

2016). First, the relative importance of attributes was

calculated, that deliver information about which

attribute affected the participants’ choice the most.

The computed part-worth utilities of all attribute

levels characterize the most important or unim-

portant attribute levels. Finally, preference simula-

tions were run estimating the impact on preferences

if single attribute levels change within a predefined

specific scenario (Orme, 2010). Preference ratings

and shares can be interpreted as indicators of ac-

ceptance. Data was analysed descriptively and, with

respect to the effects of surveillance context and user

diversity, by (M)ANOVA procedures (significance

level at 5%).

3.5 Sample Study 1 and 2

Data was collected in an online survey in Germany

and completion of each questionnaire took on aver-

age 15 minutes.

In study 1, overall 119 participants fully com-

pleted the questionnaire and were included in statis-

tical analyses. 52.9% of these participants were

Where, Wherefore, and How? - Contrasting Two Surveillance Contexts According to Acceptance

female and 47.1% male. The mean age was 28.5

years (SD=11.7) and ranged from 18 to 75 years.

The educational level was high with 38.7% of partic-

ipants holding a university degree and 45.4% a qual-

ification for university entrance. The participants’

majority lived in apartment buildings (60.5%, n=79),

while far fewer people indicated to live in detached

(14.3%), semi-detached (4.2%), or row houses

(15.1%). Most of the participants lived in the city

centre (60.5%), 23.5% on the outskirts, 7.6% in

suburbs, and 8.4% in a village. Further, 29.4% of the

participants indicated to have no experiences with

medical emergencies at all. Nearly half of the partic-

ipants (48.7%, n=58) reported to have already expe-

rienced a medical emergency in their family and

15.1% have been in a medical emergency situation

themselves. Participants indicated to feel only little

threatened by medical emergencies (M=2.3;

SD=1.1; min=1; max=6), but showed very high

needs for privacy (M=5.5; SD=0.6; min=1; max=6).

In study 2, 130 participants completed the ques-

tionnaire and were included in further statistical

analyses. 60.0% of the participants were female,

40.0% were male. The mean age was 32 years

(SD=12.2), ranging from 16 to 77 years. The educa-

tional level was also high with 48.5% holding an

university degree and 26.9% an university entrance

qualification. Similar to study 1, the majority of

participants lived in apartment buildings (60.0%),

and far fewer people lived in a detached (20.0%), a

semi-detached (6.9%), or a row house (13.1%).

43.1% of the participants lived in a city centre, while

22.3% lived on the outskirts, 20.0% in suburbs, and

14.6% in a village. 32.3% (n=42) had no experiences

with crimes at all, while 67.7% had a least fallen

victim to “slight” offenses (e.g., theft). On average,

participants indicated to feel only slightly threatened

by crime (M=2.5; SD=0.9; min=1; max=6). Similar

to study 1, the participants of study 2 showed very

high needs for privacy (M=5.5; SD=0.6; min=1;

max=6). ANOVAs revealed that the samples of

study 1 and 2 did not differ regarding gender, educa-

tional level, type of residence, perceived threat (of

crimes or emergencies), previous experiences, and

privacy needs. However, the results showed signifi-

cant differences only for age (F(1,248)=5.389;

p<.05) and place of residence (F(1,248)=10.436;

p<.05): compared with study 1, the participants of

study 2 were on average a little older and lived more

often outside the city centre.

4 RESULTS

First, the relative importance of attributes is present-

ed for the medical and crime surveillance context.

Afterwards, the part-worth utilities are presented for

all attribute levels comparing medical and crime

surveillance. Further, the results of preference simu-

lations analyses with regard to the trade-off between

safety and privacy are described.

4.1 Context-Specific Acceptance

Factors

The relative importance of attributes was calculated

for the medical and the crime surveillance context

and is shown in Figure 2.

Figure 2: Relative importance of attributes for medical and

crime surveillance.

Overall, (M)ANOVA analyses revealed significant

differences between the medical and the crime sur-

veillance context concerning the relative importance

of the attributes (F(4,249)=21.610; p<.01). For the

medical surveillance context, increase in safety was

the most important attribute (29.5%), directly fol-

lowed by invasion of privacy and locations of sur-

veillance (each 27.0%), which were also very im-

portant. The camera type is comparatively the least

important attribute (16.4%) for medical surveillance

scenarios.

This is in line with the crime surveillance scenar-

io results, where camera type was the least important

attribute (14.4%) as well (F(1,249)=2.725; p=.10;

n.s.). Concerning the other three attributes, a more

heterogeneous picture emerged: in contrast to the

medical surveillance scenario, locations of surveil-

lance was the most important attribute for scenario

decisions (F(1,249)=79.588; p<.01), followed by

increase in safety (23.1%) (F(1,249)=12.300; p<.01),

42,4%

23,1%

20,2%

14,4%

27,0%

29,5%

27,0%

16,4%

0% 10% 20% 30% 40% 50%

locations

increase in safety

(detection rate)

invasion of privacy

(data handling)

type of camera

Relative Importance (%)

medical (n=120)

crime (n=130)

SMARTGREENS 2017 - 6th International Conference on Smart Cities and Green ICT Systems

44,0

20,9

10,6

-75,5

33,7

14,9

-7,4

-41,2

21,7

-4,4

1,9

-19,2

-14,6

-4,5

3,9

15,2

25,8

18,1

11,4

-55,2

48,5

20,7

-17,3

-51,9

-6,0

4,9

25,4

-24,3

-14,1

0,5

10,4

3,2

-100,0 -80,0 -60,0 -40,0 -20,0 0,0 20,0 40,0 60,0 80,0 100,0

train station

market place

shopping mall

own home

20%

10%

archiving (police, emergency)

storage (profile data bases -

insitutions)

position determination

face recognition

integrated, hidden

mini dome

dome

conventional

Locations

increase in safety

(detection rate)

invasion of privacy

(data handling)

type of camera

Average Utilities

medical (n=120)

crime (n=130)

Figure 3: Part-worth utilities of all attribute levels for medical and crime surveillance.

and invasion of privacy (20.2%) (F(1,249)=13.908;

p<.01) which were of similar importance in the

crime surveillance scenario.

4.2 Context-Specific Acceptance

Characteristics

MANOVA analyses were calculated for each attrib-

ute level as dependent and the surveillance context

as independent variable. Figure 3 presents the results

of the part-worth utilities for medical and crime

surveillance.

The results showed a similar evaluation pattern

across scenarios, however, also significant differ-

ences for the attribute locations, (F(3,245)=4.055;

p<.01): surveillance at a store (F(1,248)=0.058;

p=.85; n.s.) or a market (F(1,248)=0.063; p=.42;

n.s.) revealed both rather positive utility values in-

dependent of the surveillance context.

In contrast, surveillance at a train station re-

ceived a significantly higher utility value in the

crime context (44.0) than in the medical context

(25.8) (F(1,248)=11.914; p<.01), while surveillance

at the own home received a clearly lower utility

value for the crime (-75.5) than the medical (-55.2)

context (F(1,248)=5.878; p<.05).

The evaluation pattern of the increase in safety

attribute levels was similar for both contexts, but

differed with regard to the amounts of values

(F(3,245)=3.770; p<.05) due to the higher attribute’s

importance for the medical surveillance context:

detection rates of 0% (F(1,248)=4.809; p<.05) and

5% (F(1,248)=9.986) obtained more negative utility

values for the medical (0%: -51.9; 5%:-17.3) com-

pared to the crime surveillance context (0%: -41.2;

5%: -7.4). Instead, detection rates of 10%

(F(1,248)=5.455; p<.05) and 20% (F(1,248)=8.176;

p<.01) received obviously higher positive utility

values for the medical (10%: 20.7; 20%: 48.5) than

the surveillance context (10%: 14.9; 20%: 33.7).

The most diverse evaluation pattern emerged for

the invasion of privacy attribute levels (F(2,244)=

13.047; p<.01). As the worst way of data and priva-

cy handling, face recognition achieved negative

utility values regardless of the surveillance context

(F(1,248)=0.906; p=.34; n.s.). For the crime surveil-

lance context, archiving of data (by police and

emergency services) obtained the highest positive

utility values (21.7), while it received slightly nega-

tive values for the medical context (-6.0)

(F(1,248)=25.444). Storage of data in profile data

bases (by medical or crime institutions) was rated

with rather positive utility values for the medical

(4.9), but with rather negative values for the crime (-

4.4) surveillance context (F(1,248)=4.143; p<.05).

Finally, position determination received the highest

positive utility values for the medical (25.4) and

Where, Wherefore, and How? - Contrasting Two Surveillance Contexts According to Acceptance

only neutral values for the crime (1.9) surveillance

context (F(1,248)=19.282; p<.01).

The levels of the attribute camera type were part-

ly rated differently for both contexts

(F(4,244)=3.202; p<.05). The hidden and integrated

camera received high negative utilities for both

groups (F(1,248)=0.027; p=.83; n.s.). Further, the

mini-dome camera was not evaluated differently as

well (F(1,248)=2.740; p=.1; n.s.). The dome camera

obtained slightly positive values for the crime (3.9),

but the attribute’s highest positive values for the

medical (10.4) surveillance context (F(1,248)=8.323;

p<.01). In contrast, the conventional large and track-

ing camera received the highest utility values for the

crime (15.2) and only slightly positive values for the

medical (3.2) surveillance context (F(1,248)=7.018;

p<.01).

4.3 Safety Vs. Privacy Decisions

Overall, the results (see 4.1) showed that the safety

and privacy attributes were significantly more im-

portant for medical than for crime surveillance.

What both contexts have in common is that there is

no clear distinction concerning the importance of the

safety and the privacy attribute (Fig. 2).

To examine the trade-off between increase in

safety and invasion of privacy in detail, sensitivity

analyses were carried out using the Sawtooth market

simulator (SMRT, 2016). In these simulations, it is

possible to examine to which extent respondents’

relative preferences for a scenario change if the

levels of an attribute vary while other specific attrib-

ute levels were kept constant. To directly contrast

safety and privacy, two opposite safety and privacy

scenarios with constant attribute levels were formed,

based on the findings of the previously reported

part-worth utilities:

1. high increase in safety + high invasion in pri-

vacy: with the constant levels “detection rate of

20%” and “face recognition”

2. low invasion in privacy + low increase in safe-

ty: with the constant levels “detection rate of

0%” and “archiving of data”

These levels were kept constant while all other

attribute levels (locations and camera type) changed.

The results are pictured in Fig. 4 for the medical and

in Fig. 5 for the crime surveillance context.

For the medical surveillance context (Fig. 4), the

“high safety” scenario (max. 61.6%) reached higher

average preferences compared to the “high privacy”

scenario (max. 20.2%). For all single attribute levels

in the medical surveillance context, the preferences

were significantly higher for the “high safety” than

for the “high privacy” scenario. The acceptance of

the “high safety” and also of the “high privacy”

scenario rose, when surveillance was carried out at

public locations. Changing the attribute level from

private home (32.9%) to a public location (store:

57.5%) in the high safety scenario, caused the high-

est difference in the share of preference (+24.5%),

while the differences between the various public

locations were rather small. Concerning all camera

types in the medical surveillance context, the “high

safety” scenario was consistently favoured by at

least 49.7% difference.

Overall, the decisions in the crime surveillance

context showed a similar pattern (Fig. 5). The ”high

safety” scenario (max. 66.5%) was clearly preferred

compared to the “high privacy” scenario (max.

20.0%) for all attribute levels (locations and camera

types) except of surveillance at the private home

environment (“high safety”: 15.3%; “high privacy”:

14.3%). Here, surveillance for crime detection rea-

sons was not desired and accepted regardless of

different safety and privacy scenarios. Similar to the

results in the medical surveillance context, the high-

est difference (+46.3%) occurred in the “high safe-

ty” scenario, when surveillance at the private home

(15.3%) was changed to a public location (store:

51.6%). In the “high privacy” scenario, there were

only small differences between the various locations

and almost no differences between the camera type

Figure 4: Results sensitivity analyses for medical surveil-

lance.

SMARTGREENS 2017 - 6th International Conference on Smart Cities and Green ICT Systems

Figure 5: Results sensitivity analyses for crime surveil-

lance.

levels. In contrast, the camera types were evaluated

differently in the “high safety” scenario: the large

tracking camera obtained the highest share of prefer-

ences (66.5%), while less visible cameras received

lower agreement (e.g., the invisible, integrated cam-

era 55.8%).

5 DISCUSSION

This study revealed insights into the acceptance of

surveillance technologies for two different surveil-

lance contexts. Using a conjoint analysis approach

and involving the location of surveillance, the ap-

plied camera type, and the trade-off between safety

and privacy as acceptance parameters, decision sce-

narios were simulated for a medical and a crime

surveillance context. The results showed which

acceptance parameters are relatively most important

and which characteristics lead to adoption or non-

adoption of surveillance technologies in urban envi-

ronments. The findings provide valuable insights for

the conceptualization and planning of smart cities

regarding an acceptable implementation and use of

surveillance technologies in urban environments.

5.1 Context-sensitive Acceptance of

Surveillance Technologies

For the crime surveillance context, the location of

surveillance clearly is the most important determi-

nant of surveillance acceptance, while increase in

safety and protection of privacy are of secondary

importance. In contrast, these three parameters are

nearly of equal importance for medical surveillance.

One explanation for the greater importance of loca-

tions in the crime surveillance context could be that

crimes were stronger associated with special loca-

tions than medical emergencies. In contrast, for

medical surveillance, the interaction of perceived

benefit (increased safety, perceived barrier (privacy),

and location of surveillance is important. What both

contexts have in common is that the applied type of

camera technology is comparatively unimportant in

relation to the other three aspects. In contrast to

previous studies, which identified safety and protec-

tion of privacy as important factors for acceptance

without weighting them directly (Slobogin, 2002;

Welsh and Farrington, 2009; Welsh et al., 2015),

this study revealed that acceptance depends on per-

ceived benefits in terms of increasing safety and to a

slightly lesser extent on privacy-related issues. The

rather similar evaluation of safety and privacy for

both contexts shows the importance to analyse this

trade-off in detail and to consider this complex phe-

nomenon in future studies as well as conceptualiza-

tions of surveillance systems in urban environments.

5.2 Context-sensitive Characteristics of

Acceptance

Confirming previous research results (e.g., Welsh &

Farrington, 2009), the use of surveillance technolo-

gies is generally accepted at public locations in ur-

ban areas. In contrast, our findings demonstrate that

surveillance technologies are not accepted at all at

private locations such as the own home. Interesting-

ly, the same acceptance pattern was found for the

crime as well as the medical surveillance context.

Although previous research also indicated this pat-

tern for medical surveillance (Himmel et al., 2013),

we assumed differences between both contexts and

preferences for the home environment in the medical

context as most AAL technologies were used and

were planned to be used in private home environ-

ments.

Concerning the increase in safety, a similar ac-

ceptance pattern was found for both surveillance

contexts as well. The pattern was merely a bit

stronger pronounced for the medical surveillance

Where, Wherefore, and How? - Contrasting Two Surveillance Contexts According to Acceptance

context due to the higher relative importance of the

attribute: the higher the detection rates and thus, the

increase in safety, the higher the acceptance. Low

detection rates of 0% or 5% were completely reject-

ed. Hence, the perceived benefit of surveillance

(increased safety) has to be noticeable for ac-

ceptance and is more important for the medical than

the crime surveillance context.

The most contrary acceptance pattern was re-

vealed for the privacy attribute. For the medical

surveillance context, position determination is the

best evaluated way of handling video data, which is

in line with previous research results. Archiving and

storage of recorded data is not desired for medical

surveillance. In contrast, archiving of data is the

most desired way of data handling in the surveil-

lance context. Interestingly face recognition is per-

ceived as invasion in privacy and is the most reject-

ed way of data handling for medical as well as crime

surveillance.

Although the camera type was relatively unim-

portant, a tendency of preferences for both contexts

can be derived: for crime surveillance, conventional

large cameras are preferred, while a bit smaller and

discrete cameras are desired for medical surveil-

lance. However, hidden and integrated camera tech-

nology is strictly rejected for both surveillance con-

texts. This is surprising and has to be considered, as

current technological developments – in particular in

the field of AAL environments – aim for designing

small, less visible or invisible and seamlessly inte-

grated technologies (e.g., Kim et al., 2012).

5.3 Trade-off between Safety and

Privacy

Previous research results on the trade-off between

privacy and safety indicated that the will to abandon

a piece of privacy for increased safety depends on

the degree of increased safety (Bowyer, 2004). Our

study revealed that only a noticeable increase in

safety (detection rate at least 10%) is perceived

positively independent of the surveillance context.

Although, the analysis of relative importance re-

vealed only slightly higher importance of increase in

safety (crime: 23.1%; medical: 29.5%) in contrast to

the privacy attribute (crime: 20.2%; medical:

27.0%), the sensitivity analyses showed that secure

scenarios were clearly preferred compared to scenar-

ios that focused on privacy.

Thus, safety is much more preferred in a direct

confrontation of safety and privacy for medical as

well as crime surveillance. Thus, safety issues are

more important criteria for the acceptance of surveil-

lance technologies than privacy issues, provided the

technology is efficient and causes a noticeable in-

crease in safety.

5.4 Limitations and Further Research

Although, the applied conjoint analysis approach

was useful to evaluate preferences in different sur-

veillance scenarios and enabled a comparison of

surveillance, there are some limitations that should

be considered for further studies.

A first limitation is that the estimated preference

ratings are ratings on a hypothetical level and do not

mirror actual behaviour. Hence, agreement or rejec-

tion might be higher or lower in real situations

(Ajzen & Fishbein, 1977). A second limitation af-

fects the limited number of attributes in a choice-

based conjoint analysis. It had to be found the right

balance between an economic research design with a

limited number of attributes and a not to high com-

plexity of the research issue. The participants sug-

gested other interesting attributes, which could be

integrated in future research, such as the period of

data storage.

Further, there are some limitations with respect

to the samples of this study. So far, a highly educat-

ed participants group was examined. It is not clear if

the results can be simply transferred to other educa-

tion levels, therefore, future studies will aim at sam-

ples with a more diverse educational level. In addi-

tion, the samples did not have much experience with

crime offenses and medical emergencies. Hence, it

cannot be excluded that the trade-offs between safe-

ty and privacy are quite theoretically evaluated,

without the true understanding of own experience.

As the approach focused on video-based surveil-

lance, it would be useful to analyse the acceptance

of other surveillance technologies with respect to

different surveillance contexts in detail.

Further, it is assumed that the evaluation of sur-

veillance technologies is influenced by current

events such as terrorist attacks and crimes in the

local environment. Hence, longitudinal studies and

comparisons of surveillance acceptance represent an

interesting approach for further research.

Finally, as the approach represents the ac-

ceptance of surveillance with a perspective of a

single country, it would be useful and interesting to

compare surveillance needs and wishes of city resi-

dents of different countries, backgrounds, and cul-

tures.

SMARTGREENS 2017 - 6th International Conference on Smart Cities and Green ICT Systems

ACKNOWLEDGEMENTS

The authors thank all participants for their patience

and openness to share opinions. Furthermore, the

authors want to thank Valentina Kneip for research

assistance. This work was funded by the Excellence

Initiative of the German State and Federal Govern-

ment (Project Urban Future Outline).

REFERENCES

Aghajan, H., Augusto, J. C., Wu, C., McCullagh, P., &

Walkden, J.-A. 2007. Distributed vision-based acci-

dent management for assisted living. In International

conference on Smart homes and health telematics, pp.

196–205, Springer.

Ajzen, I., Fishbein, M. 1977. Attitude-behavior relations:

A theoretical analysis and review of empirical re-

search. Psychological Bulletin, 84(5), 888–918.

https://doi.org/10.1037/0033-2909.84.5.888.

Arning, K., Ziefle, M., 2015. “Get that Camera Out of My

House!” Conjoint Measurement of Preferences for

Video-Based Healthcare Monitoring Systems in Pri-

vate and Public Places. In Inclusive Smart Cities and

e-Health, pp. 152–164. Springer.

Barrett, D., 2013. One surveillance camera for every 11

people in Britain, says CCTV survey. Telegraph. 10

July 2013.

Bowyer, K. W., 2004. Face recognition technology: secu-

rity versus privacy. IEEE Technology and Society

Magazine, 23(1), 9–19.

https://doi.org/10.1109/MTAS.200 4.1273467.

Cardinaux, F., Bhowmik, D., Abhayaratne, C., Hawley,

M. S., 2011. Video based technology for ambient as-

sisted living: A review of the literature. Journal of

Ambient Intelligence and Smart Environments, 3(3),

253–269.

Chen, B.-W., Chen, C.-Y., & Wang, J.-F., 2013. Smart

homecare surveillance system: Behavior identification

based on state-transition support vector machines and

sound directivity pattern analysis. IEEE Transactions

on Systems, Man, and Cybernetics: Systems, 43(6),

1279–1289.

Dailey, K., 2013. The rise of CCTV surveillance in the

US. BBC News. 29

April 2013.

Davis, F. D., Bagozzi, R. P., Warshaw, P. R., 1989. User

acceptance of computer technology: a comparison of

two theoretical models. Management Science, 982–

1003.

Dey, S., Chakraborty, A., Naskar, S., Misra, P., 2012.

Smart city surveillance: Leveraging benefits of cloud

data stores. In 37th Conference on Local Computer

Networks Workshops (LCN Workshops), 2012, pp.

868–876. IEEE.

Filipponi, L., Vitaletti, A., Landi, G., Memeo, V., Laura,

G., Pucci, P., 2010. Smart city: An event driven archi-

tecture for monitoring public spaces with heterogene-

ous sensors. In Fourth International Conference on

Sensor Technologies and Applications

(SENSORCOMM), 2010, pp. 281–286. IEEE.

Fleck, S., Strasser, W., 2008. Smart camera based moni-

toring system and its application to assisted living.

Proceedings of the IEEE, 96(10), 1698–1714.

Fyfe, N., 2004. Zero tolerance, maximum surveillance?

Deviance, difference and crime control in the late

modern city. The Emancipatory City, 40–56.

Gurley, R. J., Lum, N., Sande, M., Lo, B., Katz, M. H.,

1996. Persons Found in Their Homes Helpless or

Dead. New England Journal of Medicine, 334(26),

1710–1716.

https://doi.org/10.1056/NEJM199606273342606.

Himmel, S., Ziefle, M., Arning, K. (2013). From Living

Space to Urban Quarter: Acceptance of ICT Monitor-

ing Solutions in an Ageing Society. In Human-

Computer Interaction. Users and Contexts of Use, pp.

49–58. Springer.

Kim, J. E., Boulos, G., Yackovich, J., Barth, T., Beckel,

C., Mosse, D., 2012. Seamless Integration of Hetero-

geneous Devices and Access Control in Smart Homes.

In 2012 Eighth International Conference on Intelligent

Environments, pp. 206–213.

https://doi.org/10.1109/IE.2012.57.

Klack, L., Schmitz-Rode, T., Wilkowska, W., Kasugai, K.,

Heidrich, F., Ziefle, M., 2011. Integrated Home Moni-

toring and Compliance Optimization for Patients with

Mechanical Circulatory Support Devices. Annals of

Biomedical Engineering, 39(12), 2911.

https://doi.org/10. 1007/s10439-011-0407-1.

Koh, Y., Mohan, A., Wang, G., Xu, H., Malik, A., Lu, Y.

H., & Ebert, D. S. (2016, May). Improve safety using

public network cameras. In Technologies for Home-

land Security (HST), 2016 IEEE Symposium on (pp.

1-5). IEEE.

La Vigne, N. G., Lowry, S. S., Markman, J. A., & Dwyer,

A. M. (2011). Evaluating the use of public surveil-

lance cameras for crime control and prevention. Wash-

ington, DC: US Department of Justice, Office of

Community Oriented Policing Services. Urban Insti-

tute, Justice Policy Center.

Leonhardt, S. (2006). Personal Healthcare Devices. In S.

Mukherjee, R. M. Aarts, R. Roovers, F. Widder-

shoven, & M. Ouwerkerk (Eds.), AmIware Hardware

Technology Drivers of Ambient Intelligence, pp. 349–

370. Springer Netherlands.

Luce, R. D., Tukey, J. W., 1964. Simultaneous conjoint

measurement: A new type of fundamental measure-

ment. Journal of Mathematical Psychology, 1(1), 1–

27.

Memon, M., Wagner, S. R., Pedersen, C. F., Beevi, F. H.

A., & Hansen, F. O., 2014. Ambient assisted living

healthcare frameworks, platforms, standards, and qual-

ity attributes. Sensors, 14(3), 4312–4341.

Orme, B. K., 2010. Getting started with conjoint analysis:

strategies for product design and pricing research, 77–

89.

Patton, J. W., 2000. Protecting privacy in public? Surveil-

lance technologies and the value of public places. Eth-

Where, Wherefore, and How? - Contrasting Two Surveillance Contexts According to Acceptance

ics and Information Technology, 2(3), 181–187.

https://doi.org/10.1023/A:1010057606781.

Rao, V. R., 2014. Theory and Design of Conjoint Studies

(Ratings Based Methods). In Applied Conjoint Analy-

sis, pp. 37–78. Springer Berlin Heidelberg.

Rogers, E., 2003. Diffusion of innovations. (5th ed.). New

York: NY Free Press.

Rubenstein, L. Z., 2006. Falls in older people: epidemiol-

ogy, risk factors and strategies for prevention. Age and

Ageing, 35(suppl 2), 37–41.

Sawtooth Software. (2009). The CBC/HB System for

Hierarchical Bayes Estimation Version 5.0 Technical

Paper. 2015, from http://www.sawtoothsoftware.com/

products/advanced-analytical-tools/cbc-hierarchical-

bayes-module/167-support/technical-papers/sawtooth-

software-products/128-cbc-hb-technical-paper-2009.

Schwartz, A., 2012. Chicago’s Video Surveillance Camer-

as: A Pervasive and Poorly Regulated Threat to Our

Privacy. Northwestern Journal of Technology and In-

tellectual Property, 11, 9.

Slobogin, C., 2003. Public Privacy: Camera Surveillance

of Public Places And The Right to Anonymity (SSRN

Scholarly Paper No. ID 364600). Rochester, NY: So-

cial Science Research Network.

SMRT, 2016. SMRT - Sawtooth Software Market Simula-

tor. Sequim, WA, USA: Software Software Inc.

Solanas, A., Patsakis, C., Conti, M., Vlachos, I. S., Ramos,

V., Falcone, F., 2014. Smart health: a context-aware

health paradigm within smart cities. IEEE Communi-

cations Magazine, 52(8), 74–81.

SSI Web, 2016. SSI Web Sawtooth Software. Software for

online conjoint survey developement. Sequim, WA,

USA: Software Software Inc.

Surette, R., 2005. The thinking eye: Pros and cons of

second generation CCTV surveillance systems. Polic-

ing: An International Journal of Police Strategies &

Management, 28(1), 152–173. https://doi.org/10.1108/

13639510510581039.

van Heek, J., Arning, K., & Ziefle, M., 2015. Safety and

Privacy Perceptions in Public Spaces: An Empirical

Study on User Requirements for City Mobility. In R.

Giaffreda, D. Cagáňová, Y. Li, R. Riggio, & A.

Voisard (Eds.), Internet of Things. IoT Infrastructures,

pp. 97–103. Springer International Publishing.

Welsh, B. C., Farrington, D. P., 2004. Surveillance for

Crime Prevention in Public Space: Results and Policy

Choices in Britain and America. Criminology & Pub-

lic Policy, 3(3), 497–526.

https://doi.org/10.1111/j.1745-9133.2004.tb00058.x.

Welsh, B. C., & Farrington, D. P., 2009. Making Public

Places Safer: Surveillance and Crime Prevention. Ox-

ford University Press.

Welsh, B. C., Farrington, D. P., Taheri, S. A., 2015. Effec-

tiveness and social costs of public area surveillance for

crime prevention. Annual Review of Law and Social

Science, 11, 111–130.

Wilkowska, W., Ziefle, M., 2012. Privacy and data securi-

ty in E-health: Requirements from the user’s perspec-

tive. Health Informatics Journal, 18(3), 191–201.

Yu, M., Rhuma, A., Naqvi, S. M., Wang, L., Chambers, J.,

2012. A posture recognition-based fall detection sys-

tem for monitoring an elderly person in a smart home

environment. IEEE Transactions on Information

Technology in Biomedicine, 16(6), 1274–1286.

Ziefle, M., Schneider, C., Vallée, D., Schnettler, A.,

Krempels, K.-H., Jarke, M., 2014. Urban future Out-

line-A Roadmap on Research for Livable Cities. Smart

Cities, 9.

SMARTGREENS 2017 - 6th International Conference on Smart Cities and Green ICT Systems