=Paper=
{{Paper
|id=Vol-2232/paper4
|storemode=property
|title=Towards the Empathic Building - Detection and recognition of well-being of individuals and groups
|pdfUrl=https://ceur-ws.org/Vol-2232/paper4.pdf
|volume=Vol-2232
|authors=Jan Budke,Mario Cichonczyk,Dominic Becking
}}
==Towards the Empathic Building - Detection and recognition of well-being of individuals and groups==
https://ceur-ws.org/Vol-2232/paper4.pdf
R. Zender, U. Lucke, J. Haase, M. von der Heyde, G. Leitner und W. Meyer (Hrsg.):
Proceedings des Workshops "Lernen und Arbeiten im Wandel", Potsdam, 2018 44
Towards the Empathic Building - Detection and recognition
of well-being of individuals and groups
Jan Budke1, Mario Cichonczyk1 and Dominic Becking1
Abstract: This paper reviews the current state of science and technology on the detection and
recognition of well-being of individuals and groups of people. It turns out that most of the current
research projects refer merely to partial areas of this problem. Some concentrate on single values
like temperature, ventilation or lighting separately. However, the results are often encouraging,
suggesting that recognition of general well-being could in principle be possible. Based on these
results, we propose a new approach to measure and detect complacency in order to enable smart
buildings to react to the needs of the inhabitants and users accordingly. Datasets composed of
personal data from portable sensors and wearables or from questionnaires, data from sensors of the
building and additional information like the current utilization of the room result in a clustering
problem which can be solved using methods and algorithms of supervised learning.
Keywords: well-being, comfort, machine learning, smart building, building automation
1 Introduction
Buildings account for 40% of total energy consumption in Germany. Functional
buildings make up less than 10% of the total number of buildings, but they use 40% of
the primary energy consumption of all buildings [Bi12]. The greatest potential for
improving the energy efficiency lies therefore in the restoration and modernization of the
stock. This is one reason why many functional buildings are equipped or upgraded with
building automation and smart technologies [WHHB15].
Another reason is to improve the working and living environment of the inhabitants and
users of such buildings. One goal is to relieve users of menial tasks such as turning on
the lights. In functional buildings, this often leads to situations in which users are
confronted with technologies they did not choose to use and which, in some cases, they
have difficulties to influence or take control of [WHHB15]. In these situations, it is
likely that users do not accept certain technologies. It is often observed that users avoid
them or actively cause sensors to detect false signals or values. This way users override
the benefits of these technologies or even reverse them resulting in increased energy
usage and higher costs. The more complex said technologies get, the higher is the risk of
users rejecting it [RUU10b].
1
University of Applied Sciences Bielefeld, Institute for the Intelligent Building, Ringstraße 95, 32427 Minden,
Deutschland, firstname.lastname@fh-bielefeld.de
Towards the Empathic Building 45
Building automation is also often governed by standards and fixed default values. Other
influencing factors, direct or indirect, are therefore neglected. By looking at the
individual situation, it should be possible to save more energy and reduce costs while
improving the work and life situation of the users and residents at the same time.
These are the main reasons why we propose a system which enables smart building
technology to react according to several different influencing factors related to the well-
being and complacency2 of users, resulting in an empathic building.
In the current situation, residents and users of buildings have to manage many things
manually. In larger rooms, several thermostats must be operated to adjust the
heating/cooling, there are several switches to control and dim the various light sources,
windows must be opened and closed individually. At the same time, we are dealing with
a very inert system in which changes can only be achieved relatively slowly. As a result,
the user may feel uncomfortable or his/her work performance is impaired (e.g. by
cold/heat, too weak or too strong lighting, etc.).
With the current state of knowledge, we define the Empathic Building as a living and
working space that automatically adapts to the needs of one or more residents or users.
In doing so, basic needs as well as preferences of individuals and groups of persons
should be considered. These should be learned without user intervention, so that manual
configuration is no longer necessary.
In this paper we provide an overview of the current state of the scientific and technical
knowledge with regard to our project. Based on this literature review, we will propose
our idea in detail and elaborate on the next steps.
2 Related Work
2.1 Acceptance
Usability and acceptance have a common area of overlap but also need to be taken into
consideration separately since there seems to be no direct interdependency. Products
with a good usability can also be rejected by the users [De13]. If a product is rejected by
users, psychology explains parts of this reaction as reactance behavior. According to
Brehm, a person reacts with resistance if they feel that their freedom is restricted or
threatened [Br72].
This can be interpreted as the restriction of (degrees of) freedom. This restriction of
degrees of freedom can be divided into three categories. For the scenario with the
intelligent building at the center of interest, the narrowing of freedom due to
environmental conditions is of great importance. This means in general that certain
2
Well-being refers to the physical factors and complacency to the psychological factors.
46 Jan Budke, Mario Cichonczyk and Dominic Becking
requirements, laws, scarcity or similar factors cause that a decision-making alternative is
no longer available. This type of constraint is of particular importance for the discussed
subject, since the technologies of intelligent buildings often require a certain rethinking.
Regularly, familiar behavior patterns of users have to change because functions are
implemented differently or have been taken over by the building control. For example,
the way in which windows are opened, blinds are closed or the lights are controlled
[RUU10a].
Since reactance is motivated by the urge to restore threatened or lost liberties, this also
means that it is not directly observable [RUU10a]. Therefore, a distinction must be made
between reactance and reactance effects. There are several classes of reactance effects,
but the most important one for this subject is the direct restoration of freedom. In this
case, a familiar means of interaction is no longer supported or offered. The person
showing reactance tendencies tries to take this alternative anyway, although this
behavior might be sanctioned. However, the threat of this sanction is too weak to make
the proposed alternative attractive enough.
Ehrenbrink, Hillmann, Weiss and Möller [Eh16] performed a study that analyzed self-
adaptive and user-adaptable systems and compared their effects on users' reactance while
interacting with a dialogue system. Hong's Psychological Reactance Scale and an
ensuing factor analysis was used for the questionnaire. The resulting findings show that
interaction with self-adaptive systems compared to user-adaptable systems can increase
reactance.
In this context, another paper is interesting and should be mentioned: Heine, Buhr,
Eschweiler and Weimar [He16] focus their work on the challenges of our ageing society
and try to improve user involvement of elder end users in smart environments in order to
create a better acceptance. Therefore, a living lab called 'LebensPhasenHaus' was
established, where users can experience different smart systems and the main purpose of
this house is to raise awareness and inform elder people. The results show that the elder
users were comfortable with the system and after a while they explored the system on
their own, adapting the environmental parameters according to their needs.
2.2 Temperature and Thermal comfort
The standard specification ISO 7730:2005 states that there are six parameters for human
thermal comfort: air temperature, radiant temperature, humidity, air speed, clothing level
and activity level [IO05].
Early on, several researchers have shown dependencies between performance capability
and room temperature. Furthermore, in the work of Wyon, Andersen and Lundqvist
[WAL79], this performance reduction was quantified for the first time. In one example
for seated activity, light clothing and a room temperature of 26°C, a performance
reduction of 13% compared to a temperature of 23°C can be expected. At the same time,
memory performance increases substantially while the temperature rises from 23°C to
Towards the Empathic Building 47
26°C, henceforth decreasing again with rising temperature. In practice, complaints are
expressed regularly although technical measurements stay in a zone where the highest
level of comfort should be. Obviously, the perceptions of some users do not correlate
with the recommendations derived from said research. Fiedler argues that these feelings
are strongly dependent on individual factors and preferences and that they are not
generally the same for all people [Fi16]. However, he confirms that temperatures
between 20°C and 25°C are usually desired in individual offices and between 22°C and
24.5°C in open-plan offices.
The results of [CL12] suggest that the surface temperature of the hand provides
meaningful values. In some cases, additional sensors at other parts of the body provide
useful data which at the same time limit the user's personal freedom. In a study by Choi
and Loftness the skin temperature was used to adjust the room temperature. Skin
temperature was measured at different parts of the body and the data was analyzed to
assess a correlation between skin temperature and comfort. The results showed that the
temperature differences and the changes of temperatures had a much stronger effect than
the actual current values. The researchers also found that the measured skin temperatures
at the wrists provided data which could be interpreted better than that of any other body
segments. This data allowed to distinguish even two closely neighbored sensations like
slightly warm and neutral.
Furthermore, it could be shown that people who have the notion that their influence on
the indoor climate is high estimate their productivity higher as those with a low
perceived possibility of influence and control [LB99]. In conclusion, the facility
management of a building has a direct and indirect effect on the productivity of people
working in it. Contrary to these findings, most commercial buildings use fixed settings.
Zhang, Lam and Wang [ZLW14] in contrast focus their work on the idea that
participatory building management systems provide a better energy efficiency while at
the same time offering greater thermal comfort. It is crucial for these occupant-
participatory approaches that the occupants are trustworthy and do not lie to the system
by giving false feedback. In order to achieve this trustworthiness, they propose a
strategy-proof framework for thermal comfort voting schemes.
2.3 Air Quality
In recent years, the topic of air quality has drawn considerable attention due to
increasing problems with air pollution in big cities. Many research projects deal with the
monitoring of air quality (e.g. [Ki13]). People spend a lot of their time (nearly 90%
[EP04]) indoors and therefore good indoor air quality has an effect on our comfort and is
a vital part of human health. For this reason, some research activities concentrate on the
measurement, visualization and improvement of air quality in buildings. For example,
Frešer, Gradišek, Cvetković, and Luštrek [FGCL16] state that in addition to thermal
comfort, air quality affects the productivity of employees. They designed a system which
monitors temperature, humidity and CO2 concentration and detects whether the windows
48 Jan Budke, Mario Cichonczyk and Dominic Becking
are opened or closed. They combine this data with a self-designed ontology to
recommend actions via a smartphone application to users. For the implementation of the
predictive functions, different regression algorithms were tested and the Support Vector
Regression provided the best results and was selected. The system was evaluated and it
was shown that the occupants comfort was improved.
Kim and Paulos [KP09] work is based on air quality being difficult to rate without
measurements since it cannot be detected through sight and smell. They developed a
system called 'inAir' to solve this problem by measuring and visualizing these
parameters in such a way that also non-scientists are able to understand. Their software
creates historical and real-time visualizations of the data from different sensors. These
charts are displayed in an iPhone-App. The system was validated in a two-week study
with five private households which proved the relevance of their work showing that
these visualizations were considered helpful. Min, Forys and Schmid [MFS14] follow a
similar approach. They developed 'AirFeed', an interactive air quality monitoring system
for indoor use. The system detects particles, humidity, and temperature and sends
measurements to a server which analyzes this data after combining it with regional
information and user feedback (see figure 1). One result from their ongoing work is the
successful mapping of human activities to different levels of air quality. The amount of
experimental observation is not yet sufficient and a more extensive evaluation will be
needed.
Fig. 1: The AirFeed System Architecture [MFS14]
The paper by Snow, Soska, Chatterjee and Schraefel [SSCS16] should also be mentioned
in this context, since it provides a social perspective to the topic of air quality in
naturally ventilated office environments. They performed 15 qualitative structured
interviews, asking about social factors of opening windows and doing thermostat
adjustments in shared offices. They found that social factors affect behavior in such a
way that some people accepted sub-optimal conditions (too hot, too cold) in order to
avoid discussions and arguments.
Towards the Empathic Building 49
2.4 Mood Recognition
There already are some interesting research projects regarding the subject of mood
recognition whose results could at least partly be used for the establishment or
improvement of an empathic building. For example, LiKamWa Liu, Lane and Zhong
developed a system called 'MoodScope' [LLLZ13], a mood sensor based on smartphone
usage. The user was asked about his mood 4 times a day and the answers were matched
with an automatic recognition done by the system at the same time. To predict the daily
average mood, they used multi-linear regression and sequential forward feature
selection. MoodScope was tested collecting two months of usage data and self-reported
mood provided by 32 participants. The results validate their approach, although they
admit that 'not every factor that impacts user mood can be captured by a smartphone'
[LLLZ13]. One problem with this approach was that the user was questioned at fixed
intervals and not dependent on the current situation or in an adaptive way. Also, the
average mood does not account for immediate situations. Another problem was the
insufficient anonymization of data.
Additionally, Bachmann et al. [Ba15] focused their work on the aspect of how to use
smartphones for less obtrusive mood recognition. In their paper, they describe a
smartphone app that can be used to measure the mood, fatigue and stress of a user. They
designed the system in order to use it in the context of health care, e.g. for the
monitoring of patients. Based on first measurements, they analyzed which features
provide the best recognition results and showed that space-time attributes, such as
daytime, weekday and place provide the most information about fatigue and stress and
can be measured easily. They also found a correlation between high fatigue or stress and
poor mood.
With the help of this app, they did a formative user study with nine persons in order to
validate the usability of the system and to get training datasets with additional self-
reports. For example, the app asked for the user's mood using Likert scaling3 after certain
events occurred (e.g. message received/sent, call finished) but they tried to avoid too
many notifications at the same time. Then the classifiers used for mood-, fatigue-, and
stress-recognition were trained with naive Bayes and ten-fold cross-validation. Overall,
this gave an accuracy of 76% and after the heart rate was additionally included, the
detection rate could be increased further. As a final advice, Bachmann et al. [Ba15]
recommend other researchers to try out different algorithms and then determine features
and classifiers.
Exler, Schankin, Klebsattel and Beigl [ESKB16] also worked on the topic of mood
assessment. In their approach, they combined smartphone based self-reports, a
smartwatch and a wearable ECG monitor. Thus, this enabled them to collect subjective
feedback and objective body data at the same time. Their research shows that mood and
biophysical reactions of the human body correlate and that, in addition to heart rate, the
3
The Likert scale is a method for measuring personal attitudes. Respondents specify their level of agreement or
disagreement on an agree-disagree scale for a series of statements.
50 Jan Budke, Mario Cichonczyk and Dominic Becking
temporal aspect is also of great importance. Decision trees were trained during three
weeks of usage and then tested showing a high recognition accuracy (up to 91%). It can
be said that the app they created and the combination of smartphone and smartwatch is
in principal a suitable approach for mood recognition. However, they already have
several plans for future enhancements, like noise level recognition, an ambient light
sensor and the correlation with the user's location (shop, restaurant, office).
Niforatos, Elhart and Langheinrich [NEL15] created an app on four interactive touch-
enabled public displays on a university campus in order to gather information on well-
being of students (see figure 2). In their approach they measure the heart rate of
bypassing students with a camera-based estimation and combine results with answers of
some self-reporting questions. At the time being, their work is merely a position paper,
describing ideas and research questions but no validated results are shown yet. For the
future they plan to work on this topic and answer some questions regarding heart rate
variation during a student's day, specific periods of increased heart rate and whether
students perceive the measurement as rather useful, playful or disturbing.
Fig. 2: Prototype for mood recognition app by Niforatos et al. [NEL15]
2.5 Combined Factors & Machine Learning
Carreras, Higuera, Perálvarez and Hertog [CHPH14] created a new smart lighting and
climate system by analyzing different comfort indicators. Their system was based on two
components: the intelligent light engine and the smart thermostat. The intelligent lighting
engine was able to detect natural light levels and adjust the artificial light sources to
increase the visual comfort and reduce energy costs. The smart thermostat included a
sensor and actor system to monitor users, temperature, air velocity, CO 2, relative
humidity and illuminance levels. By adjusting comfort indicators, the system allowed to
Towards the Empathic Building 51
reduce the energy consumption up to 23% in lighting and 43% in HVAC. The authors
used input variables like occupancy user detection, daylighting levels, the position of the
blinds, the schedule of work and the daylighting indexes. The thermostat also calculated
two thermal comfort indicators to predict the satisfaction of a group of people. They are,
however, not based on data generated by the users themselves.
The importance of user participation in smart building technology is shown by the
research of Harfield and Rattanongphisat [HR17]. They presented an alternative to plain
building automation with a monitoring platform that empowers users to make intelligent
decisions about energy efficiency. Their system used temperature, humidity, power and
occupancy information to provide users information on how to reduce energy
consumption. The evaluation showed that by analyzing the collected data, areas with
potentially high energy savings could be identified. The authors also pointed out that
reacting to users leads to better energy savings than full automation.
3 A New Approach
The idea for the system proposed in this paper arose from a subjective impression most
public speakers can relate to. People like lecturers in universities develop a sense of or a
connection to the mood of their audience. Standing in front of a crowd, it is often
subjectively possible to assess if this group of people is for example relaxed, alert, tired
or stressed. If a human is able to sense this, there might be indicators which can be
measured and analyzed by an algorithm which then can make assumptions about the
mood of a group of people.
In the intelligent building, we can identify three groups of variables which have
influence on the well-being and the mood of users. First there is the direct influence.
This category includes objectively measurable variables like temperature, lighting, air
quality, noise or even olfactory influences. Related work showed that users have
individual needs, which are not met by setting fixed values that define the climate of a
room. By reacting to these needs it was shown that the satisfaction and the well-being of
the users is improved and at the same time it is possible to reduce energy consumption
and costs. Thus, acceptance of smart technologies can be increased. Individual variables
in this group are already comparatively well covered by research. However, there is no
comprehensive examination and the consideration of interactions regarding several
variables or groups of people instead of individuals.
The second group of variables is the indirect influence, or well-being and complacency
as being externally affected, with factors like time of the year/season, room size, the
number of the people inside the room, the time of day but also floor plans or similar
information. Typically, variables of this group affect users subconsciously and are
difficult to influence or cannot be influenced at all. Nevertheless, they play an important
role for well-being and complacency.
52 Jan Budke, Mario Cichonczyk and Dominic Becking
The third group is the psychological or subjective factor. Since people have different
views, conceptions, wishes and needs, all of the previously mentioned influences have
an individual effect. Even for the parameter temperature, when examined as an isolated
variable, related work shows that there are different personal wishes and preferences for
its value. Additionally, the basic emotional state of a person is an influencing factor.
This third group, therefore, plays an important role and enables a classification of data
from the two previously mentioned because it is the only valid source of information
about complacency and well-being of a person. For statements to be made, their data
must be interrelated. Data for this third category can for example be obtained by
questioning the users or gathering personal information using portable sensors or
wearables. As the related work shows, data like differences in skin temperature, heart
frequency or skin conductance can give significant information on the current mood of
the user.
It is obvious that only the first group, the direct influence, can be controlled externally,
in our case by the intelligent building. However, all groups of variables should be used
to determine the degree of satisfaction of users. We assume that it is also possible to
identify parameters that need to be adjusted in order to increase the well-being and
complacency and by doing so reduce the reactance towards intelligent building
technologies and ensure that building automation is able to achieve its objectives:
improving the living and working environment of the inhabitants and reducing costs and
energy consumption.
To realize this, we deem it important first to determine significant variables and the
corresponding sensor technology to measure them in sufficient accuracy to collect data
which in a next step allows to make correct and meaningful statements. It has to be
ascertained whether the statements and results that have proven to work in scenarios
with individual persons and limited parameters can be applied to groups of persons as
well. It is to be assumed that the group characteristics differ from individual
considerations. It is also necessary to examine the extent to which the three groups of
variables influence each other and also to find possible dependencies within each group.
Hence, we will set up a research and data collection program which will investigate this
issue. On the basis of the present variables, it can be assumed that a clustering problem is
given here and that it can be solved with methods and algorithms of machine learning.
There are two possible approaches. On the one hand, the unlabeled data can be used as
an unsupervised learning task. On the other hand, the information from the third group,
the data generated by the user himself, can be used to label direct influence and indirect
influence data. Thus, approaches of semi-supervised learning are also possible. We plan
to follow along both paths to evaluate the success rate in both cases. A big advantage of
the second approach is that - once a model which generates reliable statements is
established - this data source is not needed anymore. Personal information is always
sensitive with respect to privacy and security and portable sensors and wearables can
restrict users or cause discomfort. This is why an important goal is also to minimize
Towards the Empathic Building 53
inferences for the users. A model can also be used to make predictions. This is especially
important in the context of buildings since they form an inert system and it takes time to
react to new situations.
We plan to do further research on parameters and variables which are related to well-
being in buildings. The related work showed that some areas are better covered than
others. Thus, preliminary tests may be needed for variables which require further
research. Once the variables are defined, sample data in real world scenarios can be
gathered. For this we will work with an accordingly equipped functional building which
also acts as a living lab. The data will be preprocessed to get a better understanding of
the sample data. At this point, the results from questioning the users can be used to label
data, at least to a certain extend. Machine learning algorithms (e.g. k-means, mixture
models, hierarchical clustering) can then be used for the following tasks:
1. Determination of the statements which should be identified by the combination
of the data, or in other words definition of goals for the algorithms.
2. Selection and testing of promising algorithm candidates from the fields of
unsupervised and semi-supervised learning.
3. Further development and adaptation, as well as tuning of the algorithms by
evaluation and optimization.
4. Validation of results with information on well-being from user interviews.
4 Architecture
Based on the explanations in chapter 3, we have developed an architecture (figure 3) to
analyze inhabitant well-being. For this purpose, the data from the groups of direct
influences (shown in yellow), subjective influences (shown in red) and indirect
influences are analyzed first. The sensor data are mostly interpreted as time series, since
our first preliminary tests in feature engineering have shown that the consideration of
temporal dependence offers a higher information density and therefore more accurate
classification potential. Based on the data, a personal model of each user is created. This
already allows statements to be made about the well-being of separate individuals.
54 Jan Budke, Mario Cichonczyk and Dominic Becking
Fig. 3: Architecture
In a second step, the personal model is also used as input to create a group model. Again,
data from direct and indirect influences are used to generate features. By combining the
statements from the models of the first step, gathering information about the status of the
group is possible. The group model now contains information about the current state of
the room and the well-being of the users and thus also about possible grievances of the
variables. Based on this information, building automation can now make changes if
necessary, which are likely to improve the situation. The evaluation of this assumption
shall be a major research target.
In a preliminary test, sample data of an intelligent building was used to create a set of
promising feature transformation functions. Using these features, simple queries were
formulated at first. Trying to detect weekends or occupancy purely based on the
variables temperature and CO2 yielded very promising results with a success rate of up
to 100%. This shows the viability of the approach.
5 Conclusion
This paper reviews the current state of science and technology on the detection and
recognition of complacency and well-being of individuals and groups of people in
Towards the Empathic Building 55
intelligent buildings. We showed that most of the current research projects refer merely
to partial areas of this problem and concluded that there is a need of further research.
However, the results are often encouraging, suggesting that recognition of general well-
being could in principle be possible. Unlike a large part of the related work, we propose
a comprehensive consideration of all variables to detect complacency and the well-being
of users in order to enable smart buildings to react to the needs of the inhabitants and
users accordingly.
Our goals are to establish a correlation between sensor data and subjectively felt well-
being and to identify well-being or discomfort in groups of people. For this we want to
create and utilize a broad dataset which contains all variables related to well-being. The
resulting clustering problem can be solved by using methods and algorithms of
unsupervised learning. An architecture for a system that meets the requirements as well
as first promising preliminary tests were presented.
We address this problem because smart building technologies become more important
and start to be a lot more common in everyday life.
6 Future Work
After proving that it is possible to measure and detect complacency, the next step would
be to use this information and enable smart buildings to react to the current needs of the
inhabitants and users accordingly. For this it might be useful to first create a simulator to
test different common scenarios but also special ones in order to examine the robustness
of the system. The simulator can thereby help to further improve and optimize the
machine learning algorithms used.
Subsequently, a prototype could be developed and be used to evaluate the system in a
real case scenario. For this we can work with a smart building on campus which is built
and equipped as a living lab and hosts offices, seminar rooms, labs, a canteen, a library
and recreational rooms.
Bibliography
[Ba15] Bachmann, Anja; Klebsattel, Christoph; Budde, Matthias; Riedel, Till; Beigl, Michael;
Reichert, Markus; Santangelo, Philip; Ebner-Priemer, Ulrich: How to use smartphones
for less obtrusive ambulatory mood assessment and mood recognition. In (Mase,
Kenji; Langheinrich, Marc; Gatica-Perez, Daniel; Gellersen, Hans; Choudhury,
Tanzeem; Yatani, Koji, eds): the 2015 ACM International Joint Conference. pp. 693–
702.
56 Jan Budke, Mario Cichonczyk and Dominic Becking
[Bi12] Bigalke, Uwe, ed. Der dena-Gebaeudereport 2012: Statistiken und Analysen zur
Energieeffizienz im Gebaeudebestand. Zukunft Haus, Energie sparen - Wert
gewinnen.Berlin, 2012.
[Br72] Brehm, Jack: Responses to loss of freedom: A theory of psychological reaktance.
General Learning Press, Morristown, 1972.
[CHPH14] Carreras, J.; Higuera, J.; Perálvarez, M.; Hertog, W.: Hybrid smart lighting and climate
control system for buildings. In: IET Conference on Future Intelligent Cities. 4-5
Dec.2014.
[CL12] Choi, Joon-Ho; Loftness, Vivian: Investigation of human body skin temperatures as a
bio-signal to indicate overall thermal sensations. Building and Environment, 58:258–
269, 2012.
[De13] Dettmann, André; Jentsch, Martin; Leiber, Paul; Bullinger, Angelika C; Langer,
Dorothea: User in the loop: Konzeption und Entwicklung von Nutzerschnittstellen
für"Mobilityon- demandKonzepte. In: Conference on Future Automotive Technology-
Focus Electromobility. volume 18, pp. 03–19, 2013.
[Eh16] Ehrenbrink, Patrick; Hillmann, Stefan; Weiss, Benjamin; Möller, Sebastian:
Psychological Reactance in HCI: A Method Towards Improving Acceptance of
Devices and Services. In: Proceedings of the 28th Australian Conference on
Computer-Human Interaction. OzCHI ’16, ACM, New York, NY, USA, pp. 478–482,
2016.
[EP04] EPA, GB: Buildings and the Environment: A Statistical Summary Compiled by: US
Environmental Protection Agency GB Workgroup. Environmental Protection Agency
Green Building, Washington, DC. Available at: www. epa.
gov/greenbuilding/pubs/gbstats. Pdf (accessed 2 October 2008), 2004.
[ESKB16] Exler, Anja; Schankin, Andrea; Klebsattel, Christoph; Beigl, Michael: A wearable
system for mood assessment considering smartphone features and data from mobile
ECGs. In (Lukowicz, Paul; Krüger, Antonio; Bulling, Andreas; Lim, Youn-Kyung;
Patel, Shwetak N., eds): the 2016 ACM International Joint Conference. pp. 1153–
1161.
[Fi16] Fiedler, Eckehard: DIN EN ISO 7730 in der Praxis. In: Proceedings of the
TGAKONGRESS. Berlin, Germany, pp. 90–91, 2016. Towards the Empathic Building
13
[FGCL16] Frešer, Martin; Gradišek, Anton; Cvetković, Božidara; Luštrek, Mitja: An Intelligent
System to Improve T-H-C Parameters at the Workplace. In: Proceedings of the 2016
ACM International Joint Conference on Pervasive and Ubiquitous Computing:
Adjunct. UbiComp ’16, ACM, New York, NY, USA, pp. 61–64, 2016.
[IO05] International Organization for Standardization: ISO 7730 2005-11-15 Ergonomics of
the Thermal Environment: Analytical Determination and Interpretation of Thermal
Towards the Empathic Building 57
Comfort Using Calculation of the PMV and PPD Indices and Local Thermal Comfort
Criteria. International standards. ISO, 2005.
[He16] Heine, Thomas; Buhr, Daniel; Eschweiler, Gerhard; Weimar, Udo: Cross-generational
Smart Environment Acceptance: Experiences of Experiencing Smart Environments at
the LebensPhasenHaus. In: Proceedings of the 1st Workshop on Embodied Interaction
with Smart Environments. EISE ’16, ACM, New York, NY, USA, 2016.
[HR17] Harfield, Antony; Rattanongphisat, Waraporn: Towards an open monitoring platform
for improving energy efficiency and thermal comfort in public buildings. In: 9th
International Conference on Knowledge and Smart Technology (KST), 2017. pp. 150–
155.
[Ki13] Kim, Yongsung; Eberle, Julien; Hanninen, Riikka; Un, Erol Can; Aberer, Karl: Mobile
Observatory: An Exploratory Study of Mobile Air Quality Monitoring Application. In:
Proceedings of the 2013 ACM Conference on Pervasive and Ubiquitous Computing
Adjunct Publication. UbiComp ’13 Adjunct, ACM, New York, NY, USA, pp. 733–
736, 2013.
[KP09] Kim, Sunyoung; Paulos, Eric: inAir: Measuring and Visualizing Indoor Air Quality.
In: Proceedings of the 11th International Conference on Ubiquitous Computing.
UbiComp ’09, ACM, New York, NY, USA, pp. 81–84, 2009.
[LB99] Leaman, Adrian; Bordass, Bill: Productivity in buildings: The ‘killer’ variables.
Building Research & Information, 27(1):4–19, 1999.
[LLLZ13] LiKamWa, Robert; Liu, Yunxin; Lane, Nicholas D.; Zhong, Lin: MoodScope. In (Chu,
Hao-Hua; Huang, Polly; Choudhury, Romit Roy; Zhao, Feng, eds): Proceeding of the
11th annual international conference on Mobile systems, applications, and services.
Taiwan 2013 p. 389.
[MFS14] Min, Kyeong T.; Forys, Andrzej; Schmid, Thomas: Demonstration Abstract: Airfeed:
Indoor Real Time Interactive Air Quality Monitoring System. In: Proceedings of the
13th International Symposium on Information Processing in Sensor Networks. IPSN
’14, IEEE Press, Piscataway, NJ, USA, pp. 325–326, 2014.
[NEL15] Niforatos, Evangelos; Elhart, Ivan; Langheinrich, Marc: Public displays for monitoring
and improving community wellbeing. In (Mase, Kenji; Langheinrich, Marc; Gatica-
Perez, Daniel; Gellersen, Hans; Choudhury, Tanzeem; Yatani, Koji, eds): the 2015
ACM International Joint Conference. pp. 775–778.
[RUU10a] Raab, Gerhard; Unger, Alexander; Unger, Fritz: Die Theorie psychologischer
Reaktanz. In (Raab, Gerhard; Unger, Alexander; Unger, Fritz, eds): Marktpsychologie,
pp. 65–76. Gabler, Wiesbaden, 2010.
[RUU10b] Raab, Gerhard; Unger, Alexander; Unger, Fritz, eds. Marktpsychologie. Gabler,
Wiesbaden, 2010.
58 Jan Budke, Mario Cichonczyk and Dominic Becking
[SSCS16] Snow, Stephen; Soska, Anna; Chatterjee, Shre Kumar; schraefel, m.c.: Keep Calm and
Carry On: Exploring the Social Determinants of Indoor Environment Quality. In:
Proceedings of the 2016 CHI Conference Extended Abstracts on Human Factors in
Computing Systems. CHI EA ’16, ACM, New York, NY, USA, pp. 1476–1482, 2016.
[WAL79] Wyon, D. P.; Andersen, I.; Lundqvist, G. R.: The effects of moderate heat stress on
mental performance. Scandinavian journal of work, environment & health, 5(4):352–
361, 1979.
[WHHB15] Wilson, Charlie; Hargreaves, Tom; Hauxwell-Baldwin, Richard: Smart homes and
their users: a systematic analysis and key challenges. Personal and Ubiquitous
Computing, 19(2):463–476, 2015.
[ZLW14] Zhang, Liang; Lam, Abraham Hang-yat; Wang, Dan: Strategy-proof Thermal Comfort
Voting in Buildings. In: Proceedings of the 1st ACM Conference on Embedded
Systems for Energy-Efficient Buildings. BuildSys ’14, ACM, New York, NY, USA,
pp. 160–163, 2014.