Available via license: CC BY-NC 4.0
Content may be subject to copyright.
Review on the plication eas of
cision- king gorithms in art
mes
Lu JIN a,1, Alexander BODEN a
aFraunhofer-Institute for Applied Information Technology FIT, Germany
ORCiD ID: Lu Jin https://orcid.org/0000-0001-6901-4580, Alexander Boden
https://orcid.org/0000-0002-6470-1151
Abstract. Automated decision-making is one of the fundamental functions of smart
home technologies. With the increasing availability of Artificial Intelligence (AI)
and Internet of Things (IoT) technologies, those functions are becoming increas-
ingly sophisticated. While many studies have been conducted on optimizing algo-
rithms to improve the accuracy of predictions, less attention has been paid to how
humans interact with algorithmic systems. This involves questions such as to what
degree humans are involved in the algorithmic decision-making process and how
we can design meaningful interactions between humans and systems relying on
decision-making algorithms. With these questions in mind, our paper presents a lit-
erature review on the current state of decision-making algorithms in smart homes.
Based on an analysis of 49 selected papers, we present a systematic investigation
towards the application areas and the deployment functions that decision-making
algorithms currently take in smart homes. Focusing on two main application ar-
eas – energy management and healthcare, our paper sheds light on the current de-
ployment of decision-making algorithms in smart homes and identifies the current
intentions of involving humans in-the-loop. Within the background of facilitating
human-in-the-loop as an interaction paradigm, we aim to expose the design chal-
lenges for human-in-the-loop decision-making algorithms in smart homes which
can pave the way for developing more effective human-machine hybrid intelligent
systems in smart homes in the future.
Keywords.
smart home, decision-making algorithm, human-centered AI, human and algorithm
interaction
1. Introduction
A smart home is an application area of ubiquitous computing in which the home envi-
ronment is equipped with ambient sensors and actuators to provide context-aware ser-
vices and facilitate remote home control [1], aiming to provide comfort, safety, and save
energy for the inhabitants. With the rapid development of AI (Artificial Intelligence)
1Corresponding Author: Lu Jin, lu.jin@fit.fraunhofer.de
Ap Ar
De Ma Al Sm
Ho
HHAI 2023: Augmenting Human Intellect
P. Lukowicz et al. (Eds.)
© 2023 The Authors.
This article is published online with Open Access by IOS Press and distributed under the terms
of the Creative Commons Attribution Non-Commercial License 4.0 (CC BY-NC 4.0).
doi:10.3233/FAIA230076
74
technology, smart homes are becoming more sophisticated. While early versions mostly
depended on time tables and simple sensor readings, newer designs of smart homes an-
alyze data with the help of machine learning technologies to detect patterns automati-
cally and make more sophisticated decisions for their users. Even though users are still
in control and can override automated decisions of their smart home systems, there is the
question how such forms of human-algorithm interaction can be designed in a way that
ensures both comfort and transparency. Our study seeks to understand the current state
of decision-making algorithms in smart homes, investigating application areas as well
as their deployment situation. Through a literature analysis of current works in the area
of algorithmic decision-making in smart homes, we aim to achieve a bird-eye view and
investigate design opportunities for human-algorithm interaction design.
We can note that lots of studies have focused on improving the quality of decision-
making algorithms [2][3][4] or applying such technologies to broader application areas
[5][6][7]. At the same time, there are few studies on how humans interact with decision-
making algorithms, even though existing studies have investigated the impact of algo-
rithm decisions on human decision making, exposing a need to involve humans into the
automation to reduce biases of the systems [8]. Some research [9] has shown that we
need to put humans in-the-loop to help algorithmic decision-making to overcome their
limitations. It exposes the difficulties of designing for human and algorithm interaction
in various fields of applications and their impacts on our everyday lives.
This paper seeks to provide a systematic review towards decision-making algorithms
in smart homes, an area where such technologies are arguably very impactful on human
life. It aims at investigating the possibilities and challenges for human and algorithm
interaction. Specifically, in this paper, we want to investigate
•RQ1: What are the functions of decision-making algorithms in smart homes that
are currently discussed in the literature?
•RQ2: What are the design challenges of human-in-the-loop decision-making al-
gorithms in this context?
The contributions of this paper are the following:
•We provide a systematic literature review towards functions of decision-making
algorithms in current smart home systems.
•Based on an analysis of the state-of-the-art, we expose design challenges for
human-in-the-loop decision-making algorithms in smart homes.
After the introduction, section 2 introduces the related work on the core issues that
we are investigating. Section 3 describes the methodology that was used in our literature
review. Section 4 provides a comprehensive overview on application areas and functions
that decision-making algorithms provide in smart homes (RQ1). Section 5 goes on to
discuss challenges for human and algorithm interaction, and analyses design challenges
for human-in-the-loop decision-making algorithms in smart homes (RQ2). The paper
closes with a conclusion and outlook on future work.
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms 75
2. Related Work
2.1. IoT in Smart Homes
Internet of Things (IoT) technologies are at the core of smart homes [10]. Generally,
the IoT describes “physical objects (or groups of such objects) with embedded sensors,
processing ability, software, and other technologies that connect and exchange data with
other devices and systems over the Internet or other communications networks”[11].
Hence, the IoT [12] bridges the gap between the virtual world and the physical world.
Based on IoT infrastructure such as networked sensors and actuators, users can control
their house appliances with devices such as tablets, mobile phones or computers. It aims
towards “Home Automation” where users can control all of the devices in the smart
home with all of the appliances connected to each other. Among others, it aims to in-
crease comfort, safety, and also provide means for saving energy for the inhabitants (e.g.
allowing to check from afar if a window is open or the stove still turned on).Apart from
manual control, Artificial Intelligence (AI) increasingly plays a role in the smart home,
for example, Alexa – a smart voice assistant from Amazon – uses speech recognition
and natural language processing technologies to allow users to control appliances with
their voice. Other smart devices try to detect behavior patterns and provide automated
control of appliances based on the user habits. For example, if all persons left the home,
with AI technology, the heater, fan or any device can be turned off and the door locked
automatically. Xiao[13] divides AI functions in smart homes into 6 categories – activ-
ity recognition, data processing, decision-making, image recognition, prediction-making
and voice recognition. In our study, we are concerned with the issue of decision-making
in the context of smart home technologies, which we will further explain in the following
section.
2.2. Decision-making algorithms
The definition of an algorithm is “a process or set of rules to be followed in calculations
or other problem-solving operations, especially by a computer” [14]. Decision-making
[15] is a process in which an agent, no matter if it is a physical entity like a human or
a non-physical entity such as a decision-support system, acts based on data observed
from the environment. With help of algorithms or machine learning, computers can take
decisions based on such observations and also act on them with the help of actuators
(such as turning the light or a heater on or off). Systems that take such decisions on
behalf of users based on algorithms (be it human-designed or trained automatically) is
what we understand as decision-making algorithms in our study. Decision-making al-
gorithms or algorithmic decision-making [16] are widely used in different areas to take
automated decisions or help humans in making decisions. For example, in autonomous
cars, decision-making algorithms can be used to decide where to go next[17], or how to
manage to go to the destinations in the most effective way[17]; in the financial sector,
decision-making algorithms can be used to evaluate the financial risks[18] or help people
invest their resources in a sustainable manner across their lifetime [19]. In the area of
healthcare, decision-making algorithms can be used to support diagnoses of the doctors
[20][21], or help patients to make a decision towards purchasing a specific healthcare
product based on individual parameters [22]. Further examples include aerospace ap-
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms76
plications, where decision-making algorithms can help to prevent midair collisions be-
tween aircraft[23], or other safety-critical domains such as firefighting wildfires, where
decision-making algorithms can support safety and efficiency of emergency response
operations [24][25][26].
Decision-making algorithms also appear in smart home scenarios, where intelligent
buildings increasingly take ever more sophisticated decisions, for example in the domain
of energy management, balancing energy loads based on the status of the energy grid
[27] or control heating/cooling based on predefined values [28] (sometimes set by the
energy providers as a recent example from the USA showed).
3. Methodology
Figure 1. Procedural Visualization
3.1. Databases
Our literature review investigates the role of decision-making algorithms in smart homes.
More specifically, we focus on the use of algorithms that automatically make decisions or
support decision-making in the context of private homes. Relevant papers were identified
by searching the scopus database. Scopus is the largest abstract and citation database
which delivers a broad view on research from several fields, e.g. science, technology,
social science.
3.2. Search Term
The search keywords have been “decision-making algorithms” and “smart home”. We
tried other similar keywords, for example, “decision support tools” and “smart home”,
“decision making” and “smart home”, but discarded papers that did not discuss algo-
rithms as part of the system description. Moreover, because of the rapid development
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms 77
in the AI domain, and to narrow down our search to the latest research, we only col-
lected matching papers from 2018 to 2022. Therefore, based on these criterion, in the
first searching phase, 100 papers were selected.
3.3. Inclusion criteria
The title and abstracts of 100 papers were screened by the first author to identify whether
they are related to our context – smart home and whether they discuss decision-making
algorithms in an appropriate depth. Only full papers were considered for the analysis,
leading to 10 papers being discarded. After reading the title and abstract, 55 papers where
left for us to analyse deeper. Of these, we checked the full text in depth. In the end, we
included 49 papers (see Appendix) in our analysis, after some had to be discarded by
being too far of our research topic. For example, there were several papers that related to
the smart grid, but did not cover the interaction with users [29].
4. Results
4.1. The Application Areas of decision-making algorithms in Smart Homes
RQ1 aims to identify the application areas of decision-making algorithms in smart
homes. According to the literature review, there are two main application areas - energy
management and healthcare. Among all of the 49 selected papers (see Appendix - Table
3), we found that 55% have contributed to the energy management application area, 16%
have contributed to the healthcare area (see Figure 2). For other application areas, it in-
cludes entertainment, security and etc. In the following sections, we describe two main
application areas in more detail.
Figure 2. Application areas of decision-making algorithms in smart homes
4.1.1. Energy management
In the smart home, one of the most prevalent application areas that decision-making algo-
rithms apply to is energy management. Mostly, they are deployed as home energy man-
agement system (HEMS) to control and schedule home appliances. The main objective
of the decision-making algorithms here is to minimize the cost of energy consumption
and maximize the comfort of inhabitants in the smart home. Another objective is to make
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms78
the smart home users “active in the energy market and reduce the pressure of the power
grid including peak shaving, load shifting and etc and achieve better carbon efficient
environment[30]”. According to the literature review, energy management can be done
both on the side of the smart grid and within the smart home. For the smart grid, it mainly
focuses on optimising the supply side of the grid by involving users in the energy allo-
cation decision making process. For the smart home, it mainly focuses on comfort and
energy saving practices within the home, which can be related to smart grid technologies
(but don’t have to). For example, in the smart grid, the decision-making algorithms aim
to allocate energy based on the (measured or predicted) user demands. Furthermore, ap-
plications within the home can be controlled from outside (with user’ agreement) to im-
prove the load balancing of the grid. Before the smart grid, the supply side management
was only focusing on a single-side optimization problem, but with an energy manage-
ment system (HEMS) installed in the smart home, the decision-maker in the supply side
can get responses from the customers and have a “smart” interaction with the demand
side. Current research is proceeding into the bi-level optimization problem with both
the supply side and the demand side. Therefore, an iterative decision-making process
between the upper-level optimization and the low-level optimization [31] is emerging.
On the demand side in smart homes, energy management can “allow the end-users to
communicate with the grid operator so that they can contribute in making decisions and
assist the utilities to reduce the peak power demand through peak periods”. For instance,
Fanlin et al.[32] introduces a two-level decision-making framework between the smart
grid and users. One level is for retails to announce their electricity prices for the next 24
hours, and the other level is for users to schedule their energy usage accordingly.
4.1.2. Healthcare
Healthcare in smart home or smart healthcare is the second major application area where
decision-making algorithms are deployed in smart homes. The aim is usually to reduce
the cost of healthcare and improve people’s quality of life. Healthcare in smart homes is
related to several special concepts such as “Ambient intelligence” or “Ambient assisted
living” [33]. Ambient intelligence can be defined as “a computing paradigm that uses
information technology and its applications to enhance user abilities and performance
through interconnected systems that can sense, anticipate, adapt, predict, and respond to
human behavior and needs” [33]. Ambient assisted living “aims at extending the time
older people can live in their home environment by increasing their autonomy and assist-
ing them in carrying out activities of daily living by the use of intelligent products and the
provision of remote services including care services[34]”. Decision-making algorithms
are one of the fundamental functions in these concepts to help aging people in their daily
life. There are several layers in smart ehealthcare homes [35] – sensors and actuators,
home communication networks, autonomous computing as well as safety and healthcare
services. Decision making is in the layer of autonomous computing, which handles the
data from sensors and actuators in the inhabitant’s environment as well as health data
and takes decisions such as providing diagnoses, presenting reminders, or warnings etc.
4.2. The Functions of decision-making algorithms in Energy Management
In this study, we focus on the two application areas which seem to be the most common
use case (71%) in smart homes: Energy management (see Appendix - Table 4), and
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms 79
healthcare (see Appendix - Table 5). Based on the two application areas, we go forward
to the detail functions that decision-making algorithms have in smart homes.
4.2.1. Energy allocation
We found one of the fundamental functions that decision-making algorithms have is for
energy allocation between smart homes and the smart grid. Its main aim is to minimize
the electrical consumption in the home, while still ensuring user comfort. For energy
allocation between smart homes and the smart grid, the decision-making algorithms are
deployed in the demand side management and consumer side management to distribute
energy. Fanlin et al.[32] “proposed a two-level decision-making framework where the
retailer acting as an upper-level agent firstly announces its electricity prices of next 24
hours and customers acting as lower-level agents subsequently schedule their energy us-
ages accordingly.” Through the feedback from the customers, an upper-level agent can
allocate and distribute energy accordingly. Alisson et al.[36] “present a management sys-
tem of residential loads based on the user’s behavior, climatic variables and possibility of
integration with distributed generation and Smart Grid. The system uses artificial intelli-
gence (AI) techniques to make decisions on each automated load.” Several types of data
are used to make decisions, for example, “in temperature management, the resident’s be-
havior through of the modification of parameters of air conditioners is used for temper-
ature prediction” [37], and [38] introduce the cooperation and distribution between dif-
ferent smart homes and the smart grid in demand-side management to achieve optimized
energy distribution. In [38], M. Hadi Amini et al. use distributed decision making from
agents rather than central decision making on energy distribution.
4.2.2. Energy consumption forecasting
Forecasting is an another function that decision-making algorithms take. It is meant to
support the energy allocation by providing better planning abilities for the energy suppli-
ers. There are several methods used for decision-making in this area, for example, long
short-term memory (LSTM) and fuzzy neural inference systems with genetic algorithms.
In [39], they applied several pre-processing techniques to deal with the diverse nature
of electricity data, followed by an efficient decision-making algorithm for short-term
forecasting and implemented it over resource-constrained devices. Tao et al.[39] provide
an efficient deep learning framework to predict the future energy consumption and also
provide a communication between energy distributors and consumers. Accurate demand
forecasting [40] is important for future strategic planning and scheduling. It can also help
the consumers to minimize the cost of electricity.
4.2.3. Home appliances scheduling and controlling
Decision-making algorithms also provide functions for analyzing and making decisions
for controlling and scheduling the appliances in the houses. Their main aim is to min-
imize the electrical consumption at home and reduce peak-to-average ratio while still
ensuring user comfort. For example, Tengyue et al. [41] deployed decision-making algo-
rithms to change the room temperature, which can adapt to the lifestyle of tenants and
the outdoor temperature. Jordehi [42] introduced a new binary optimisation algorithm for
optimal scheduling of appliances in smart homes. Bilal et al.[43] introduced a new EMS
by scheduling the devices to minimize the electricity bills, alleviate peak-to-average ra-
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms80
tio, and maximize user comfort. In this context, home appliances can be categorized
into deferrable appliances and non-deferrable appliances. For example, dishwashers and
washing machiness are deferrable appliances, while fridges are non-deferrable appli-
ances. In some papers, the categories are named differently, such as shiftable appliances
and non-shiftable appliances, or delay-sensitive appliances and delay-tolerant appliances
[43]. Based on the differences in the appliances, the controlling and scheduling strategies
are different. For example, Diego et al.[44] presented an optimization model to schedule
deferrable appliances in households. They used two algorithm simulation-optimization
approaches and a greedy heuristic for decision-making. The aim is to balance the elec-
tricity costs and user satisfaction.
4.2.4. Anomaly detection / Device Management
The decision-making algorithms also take over several minor functions, such as anomaly
detection and device management. For anomaly detection, they use an algorithm to de-
tect irregular patterns in activities and notify the relevant stakeholders. For example, [45]
introduce a system that can forecast electricity bills and notify users if abnormal energy
consumption of individual home appliances is detected. For device management, in [12]
introduce the common machine learning algorithm – Support Vector Machine – to clas-
sify and handle the the status of devices. Zelin et al.[46] introduce the first component of
activity recognition for a user’s state to support future decisions towards adaptive control
over devices.
4.3. The functions of decision-making algorithms in healthcare
4.3.1. Simple diagnose
Diagnosis is one of the basic functions that decision-making algorithms are deployed on
in the area of healthcare. For some illnesses, it is possible to take an automated diagnosis
based on an algorithm. For example, Mahdi [35] illustrates the CDS (cognitive dynamic
system) algorithm based on a decision-making system (ADMs) that is used in diagnosing
Arrhythmia disease. These are similar to a doctor’s decision process which are often
based on checklists - getting “yes” or “no” answers from patients, and then classifying the
disease. Such simple approaches are the basis for more complex smart home e-healthcare
systems. For example, Prabal et al.[47] illustrates how a wearable device can be used to
remotely monitor a patient’s health in real-time. Based on the daily monitoring data, it
can predict some diseases, such as diabetes.
4.3.2. Simple screening
Simple screening [48] is quite similar to simple diagnosis. Based on sensors, such ap-
proaches can replace some examinations that would normally be taken in a hospital. The
difference between simple screening and simple diagnose is that screening attempts just
to check whether patients are in good health condition, but they don’t attempt to identify
specific diseases. For example, Mahdi et al.[48] introduce a system that can “screen hu-
man health condition automatically between two binary (healthy and unhealthy) states
based on subjects’ single lead ECG traces”. The benefits of simple screening and simple
diagnose are that the underlying algorithms are rather simple to implement.They can be
trained and take decisions very quickly (i.e. recommending to see a doctor) and the result
is simple enough - healthy and unhealthy.
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms 81
4.3.3. Emergency alert
Decision-making algorithms can also help the doctors or caregivers to provide alert func-
tions during emergency situations. Based on the degree of urgency, the system can de-
cide whether to automatically notify family members, doctors or emergency medical
providers. For example, in [47], if the value of THI (temporal health index) is above
the threshold level, patients’ family members are informed by generating warning alerts.
Other projects investigate automatic fall detection, which is a common problem for el-
derly people living at home.
5. Discussion
Decision-making algorithms are a central AI function of smart home applications. Ac-
cording to our literature review, we can see that the functions of decision-making algo-
rithm deployment range from very simple functions to more advanced functions with the
goal of providing a personalized experience. In this progress, we can see that humans
have to be considered in the decision-loops of the systems. However, while we have
found many studies focusing on improving the quality of decision-making algorithms,
we have seen few studies that focus on how humans interact with decision-making al-
gorithms. Therefore, in RQ-2, we asked the question: what are the design challenges for
human-in-the-loop decision-making algorithms in smart homes?
For our analysis, we refer to the framework of “human-centred AI” that has been
suggested by Ben Shneiderman[49]. The goal of human-centred AI is not to replace
humans entirely but to enhance our capabilities by way of intelligent, human-informed
technology. Through its combination of the precision of machine learning with human
input and values, human-centered AI is expected to be able to take more informed deci-
sions. In his definition, human-centered AI aims at bringing AI under human control. It
spans the whole range between full human control to full AI automation). In this defini-
tion, human-centered AI is to build collaboration between humans and AI to augment the
AI system. Such approaches have been more and more recognized lately, as designers
increasingly have to consider human aspects in the design of AI systems [50].
If we look at the functions provided by decision-making algorithms in the smart
home, they all aim to achieve AI automation. For some functions, such as simple diag-
nose, screening, and anomaly detection, the computer will take a decision based on the
data collected by the sensors and take action based on computation. For some functions,
such as home appliances scheduling and controlling, they will not only automatically
schedule the appliances, but also involve user feedback or preference into the decision-
making algorithm process.In order to answer RQ2, we first identified three main efforts
that designers and developers did in our review to involve humans in the loop. Then,
we explain the needs of involving human in the loop. In the end, we illustrate the de-
sign challenges that designers and developers will face in designing human-in-the-loop
decision-making algorithms in smart homes.
5.1. The current state of human-in-the-loop decision-making algorithms in smart homes
5.1.1. Build dialogue between users and algorithms
We can see from the review, that for most of the identified functions, humans are rather
out of the loop of the decision-making process. Zhang et al.[30] mention the challenge
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms82
of how to include human behaviour in the loop of making control decisions. However,
there is some application area – energy allocation between smart home and smart grid
where developers are trying to get users’ feedback from the consumer side to enhance the
results of energy allocation. Another area is home appliances scheduling and controlling.
In [44], the author tries to include user preferences into the decision-making algorithm,
in order to handle the uncertainty of human preferences. Dietvorst et al. [51] show that
allowing algorithmic predictions to be modified by humans can make humans more likely
to use them.
5.1.2. Be transparent about how outputs are generated
Most of the studies we identified do not consider the transparency of the decision-making
process of the algorithm. Only Sara et al. [52] provides a transparent decision-making
procedure with significant predictive performance. For technology oriented works, accu-
racy is usually the biggest concern when building decision-making algorithms. Rene’s
paper [53] discusses how transparency influences the trust towards the algorithm result.
Algorithm transparency makes humans understand the decisions or predictions made by
the AI. It contrasts with the “black box” concept in machine learning where even its
designers cannot explain why an AI arrived at a specific decision. If there is a lack of
transparency of how outputs are generated, users might lose trust towards the results [54].
5.1.3. Override the algorithm result if it takes action
In Ben Shneiderman’s[49] framework, the system can be divided into 4 categories: 1)
high human control with high computer automation, 2) high human control with low
computer automation, 3) low human control with high computer automation and 4) low
human control with low computer automation. For some functions in healthcare, for
example, simple diagnose or simple screening, the functions are fully automated with
low human control, therefore, there is no need to override it because the system does not
take any action. But for some functions, users might want to have the ability to override
the algorithm results. For example, for controlling and scheduling home appliances, users
may want to be the final decision maker even under fully automated systems. Hence,
designers and developers should consider to give users the control towards overriding the
results of automated decisions. From the review, only Nassourous et al. [55] consider to
use a open-loop system which will include dependent control inputs or use a closed-loop
system which will include user feedback for independent control inputs.
5.2. The needs of human-in-the-loop of decision-making algorithms in smart home
We can observe how the deployment of the decision-making algorithms has gone from
simple functions such as energy allocation or simple diagnosis to more advanced func-
tions such as home appliances scheduling and controlling. With the development of AI
systems, we can see several papers attempting to provide a more personalized experi-
ence in energy management and healthcare, which requires more advanced functions.
For example, in the energy management area, Alisson et al.[36] suggest three modes for
the users to select – comfort mode, standard mode and economical mode. In comfort
mode, it is non-intrusive, as the system does not make any restrictions for using residen-
tial loads. Standard mode is the one that makes a little restriction on the use in certain
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms 83
scenarios of tariff and accentuated consumption. The Economical mode is the operation
mode with the highest intrusion of the system in the disconnection of loads, it aims at
users who wish to reduce the use of energy from the network, even in scenarios of low
power generation. In the healthcare area, Zhang et al.[56] mention the concept of lifelong
healthcare data management. Their system can take different decisions such as users’
treatment times, and provide personalized lifestyle guidance based on these data. In the
future, it is expected to provide users with recommendations for exercise actions and get
users’ feedback on recovery feelings to fill the gap in the after-medical-care phase. In
these advanced features, we can see that humans are assumed to be more involved in the
loop of decision-making processes. Similarly, Abdelilah Rochd[57] proposed the ideas
to include human behavior into the control decision to build personalized preference pro-
files. There are two components in their system, one is the environment, and the other is
the agent. Agents learns from the environment and makes decisions for the users. From
this perspective, the author thinks it can involve users into the agent’s decision-making
process, or give rewards to the system. Another example is in HVAC, it gives users in-
put towards how many occupants are in the environment to determine the HVAC control
policy. With human and AI collaboration, such work can be done more efficiently.
5.3. The design challenges of human-in-the-loop of decision-making algorithms in
smart home
As mentioned by Wei Xu[58], history is repeating. When computers first appeared in
the world, the development was very technology-driven by the developers. And when
computers became popular, the concept of “human-centered design” [59] has been put
up to solve the problems of user experience of computers. The same seems to be hap-
pening with AI: a lot of research is focusing on the accuracy of the algorithms, but as AI
technology starts to be adopted on a bigger scale, “human-centered AI” [49] is becoming
much more prominent.
We can see there are efforts from in the literature that we studied to put humans in
the loop of decision-making algorithms, corresponding to the goals mentioned in HCAI
[58], for example,“building dialogue between users and algorithm” and “override the
algorithm results if it takes action” match the goal of providing human-controlled AI
and human-driven decision-making, and “be transparent about how outputs are gener-
ated” corresponds to the goal of providing explainable AI. However, the efforts seem to
be rather underrepresented, often lack systematic design considerations, and also rarely
consider human needs in the beginning.
The goal of designing human-in-the-loop decision-making algorithms would be to
combine human intelligence with machine intelligence to achieve better results, keeping
humans as a part of the automated system [58]. There are two ways to achieve this goal:
one is to involve humans in the decision loop of the system, the other is to involve human
intelligence into the system which has been labelled as cognitive computing [60]. In this
paper, we only focus on the first aspect. Therefore, in Table 1 and 2, we will analyse
some issues of human and decision-making algorithms interaction in different context
under the main issues category of HCAI [58]. With these challenges mentioned above
in different context, we conclude three main challenges of designing human-in-the-loop
decision-making algorithms in smart homes.
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms84
5.3.1. Accessibility to all users
Within the context of smart home, accessibility becomes the main challenge for design-
ing human-in-the-loop decision-making algorithms. It must consider all kinds of users,
no matter their age, their technical expertise and their abilities. In the current situation,
we seldom consider these questions, since the smart home has not yet come into popu-
larity, but in the future, it will become the main challenge.
5.3.2. Different levels of user control
We can see that the level of control from users’ needs differs depending on the context
of use. If the functions are not directly related to the needs of the users, such as the issue
of energy allocation and anomaly detection, the control needs from users are not high.
In the function of home appliances scheduling and controlling, users have the option to
customize their usage of smart homes or allow the algorithms to make decisions auto-
matically based on the training dataset results. Additionally the algorithms can evolve
continuously with the collection of more data and feedback from the users, theoretically
leading to ever more adapted systems based on the user’s previous habits. Still, as we
have discussed above, there is a need for users to have some sort of control over their
algorithmic decision to some extent and that they can intervene into the process.
5.3.3. Personalization with acceptable privacy
As we mentioned above, the future goal of decision-making algorithms in smart homes
should be personalized to the needs and preferences of individual users. Especially in the
light of ongoing debates about “surveillance captialism”, this will pose the question of
how to collect and analyze user data in a way that respects user privacy? In different con-
texts, for example, in healthcare, the closeness between humans and machines means the
privacy issue will be the top concern in designing human-in-the-loop decision-making
algorithms. The relationship is two-way: the machines need continuous input and feed-
back from users and the more data the users provide, the more privacy concerns the users
might have. Therefore, in order to reach personalization, we should put the privacy issue
at the top.
6. Conclusion
Based on the literature review and its analysis, we can see the ongoing development of
AI systems with decision-making algorithms in two major application areas of smart
homes – energy management and healthcare. We analyze what the functions of decision-
making algorithms are: they are mainly used for energy allocation between smart grid
and smart home, home appliances scheduling and controlling, energy consumption fore-
casting, simple diagnosis in healthcare, simple screening, real-time monitoring and diag-
nosis. In our analysis, we first identified the efforts of implementing human-centered AI
in the literature review. On that basis, we illustrated challenges of implementing human-
centered AI in different contexts in detail based on the umbrella of the main issues of
HCAI. In the end, we conclude the three main challenges of implementing human-in-
the-loop decision-making algorithms in smart homes. In the future, we will investigate
how algorithmic decision-making and human-decision can cooperate together (hybrid
intelligence) seamlessly in smart homes in more depth.
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms 85
7. Acknowledgements
This work has been funded by the EU Horizon 2020 Marie Skłodowska-Curie Interna-
tional Training Network GECKO, Grant number 955422 (https://gecko-project.eu/).
References
[1] Alam MR, Reaz MBI, Ali MAM. A review of smart homes—Past, present, and future. IEEE transactions
on systems, man, and cybernetics, part C (applications and reviews). 2012;42(6):1190-203. Available
from: https://doi.org/10.1109/TSMCC.2012.2189204.
[2] Teoh CC, Tan CE. A neural network approach towards reinforcing smart home security. In: 8th Asia-
Pacific Symposium on Information and Telecommunication Technologies. IEEE; 2010. p. 1-5.
[3] Li D, Jayaweera SK. Reinforcement learning aided smart-home decision-making in an interactive smart
grid. In: 2014 IEEE Green Energy and Systems Conference (IGESC). IEEE; 2014. p. 1-6. Available
from: https://doi.org/10.1109/IGESC.2014.7018632.
[4] Yu L, Xie W, Xie D, Zou Y, Zhang D, Sun Z, et al. Deep reinforcement learning for smart home
energy management. IEEE Internet of Things Journal. 2019;7(4):2751-62. Available from: https:
//doi.org/10.1109/JIOT.2019.2957289.
[5] Habal F, Huang V. a decision-making algorithm for the management of pregnancy in the inflammatory
bowel disease patient. Alimentary pharmacology & therapeutics. 2012;35(5):501-15. Available from:
https://doi.org/10.1111/j.1365-2036.2011.04967.x.
[6] Chouldechova A, Benavides-Prado D, Fialko O, Vaithianathan R. A case study of algorithm-assisted
decision making in child maltreatment hotline screening decisions. In: Conference on Fairness, Ac-
countability and Transparency. PMLR; 2018. p. 134-48.
[7] Yang Z, Ouyang T, Fu X, Peng X. A decision-making algorithm for online shopping using deep-
learning–based opinion pairs mining and q-rung orthopair fuzzy interaction Heronian mean oper-
ators. International journal of intelligent systems. 2020;35(5):783-825. Available from: https:
//doi.org/10.1002/int.22225.
[8] De-Arteaga M, Fogliato R, Chouldechova A. A case for humans-in-the-loop: Decisions in the pres-
ence of erroneous algorithmic scores. In: Proceedings of the 2020 CHI Conference on Human Fac-
tors in Computing Systems; 2020. p. 1-12. Available from: https://doi.org/10.1145/3313831.
3376638.
[9] Calero Valdez A, Ziefle M. Human factors in the age of algorithms. understanding the human-in-the-
loop using agent-based modeling. In: International Conference on Social Computing and Social Media.
Springer; 2018. p. 357-71. Available from: https://doi.org/10.1007/978-3-319- 91485-5_27.
[10] Alaa M, Zaidan AA, Zaidan BB, Talal M, Kiah MLM. A review of smart home applications based on
Internet of Things. Journal of Network and Computer Applications. 2017;97:48-65. Available from:
https://doi.org/10.1016/j.jnca.2017.08.017.
[11] wikipedia. Internet of Things; 2020. Available from: https://en.wikipedia.org/wiki/
Internet_of_things.
[12] Majeed R, Abdullah NA, Ashraf I, Zikria YB, Mushtaq MF, Umer M. An intelligent, secure, and smart
home automation system. Scientific Programming. 2020;2020. Available from: https://doi.org/
10.1155/2020/4579291.
[13] Guo X, Shen Z, Zhang Y, Wu T. Review on the application of artificial intelligence in
smart homes. Smart Cities. 2019;2(3):402-20. Available from: https://doi.org/10.3390/
smartcities2030025.
[14] Lee MK. Understanding perception of algorithmic decisions: Fairness, trust, and emotion in response to
algorithmic management. Big Data & Society. 2018;5(1):2053951718756684. Available from: https:
//doi.org/10.1177/2053951718756684.
[15] Mykel Kochenderfer TW, Wray K. Algorithms for Decision Making. MIT PRESS; 2022.
[16] Wang R, Harper FM, Zhu H. Factors Influencing Perceived Fairness in Algorithmic Decision-Making:
Algorithm Outcomes, Development Procedures, and Individual Differences. In: Proceedings of the 2020
CHI Conference on Human Factors in Computing Systems. CHI ’20. New York, NY, USA: Association
for Computing Machinery; 2020. p. 1–14. Available from: https://doi.org/10.1145/3313831.
3376813.
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms86
[17] Schwarting W, Alonso-Mora J, Rus D. Planning and decision-making for autonomous vehicles. Annual
Review of Control, Robotics, and Autonomous Systems. 2018;1:187-210. Available from: https:
//doi.org/10.1146/annurev-control- 060117-105157.
[18] Srinivasan S, Kamalakannan T. Multi criteria decision making in financial risk management with a
multi-objective genetic algorithm. Computational Economics. 2018;52(2):443-57. Available from:
https://doi.org/10.1007/s10614-017- 9683-7.
[19] Lakhno V, Malyukov V, Gerasymchuk N, Shtuler I. Development of the decision making support system
to control a procedure of financial investment. - . 2017;6(3):35-41. Available from: https://doi.
org/0.15587/1729-4061.2020.194531.
[20] L´
egar´
e F, Adekpedjou R, Stacey D, Turcotte S, Kryworuchko J, Graham ID, et al. Interventions for in-
creasing the use of shared decision making by healthcare professionals. Cochrane database of systematic
reviews. 2018;(7). Available from: https://doi.org/10.1002/14651858.CD006732.pub4.
[21] Lysaght T, Lim HY, Xafis V, Ngiam KY. AI-assisted decision-making in healthcare. Asian Bioethics
Review. 2019;11(3):299-314. Available from: https://doi.org/10.1007/s41649-019-00096- 0.
[22] Yang Z, Garg H, Peng R, Wu S, Huang L. Group decision algorithm for aged healthcare product
purchase under q-rung picture normal fuzzy environment using Heronian mean operator. International
Journal of Computational Intelligence Systems. 2020;13(1):1176-97.
[23] Kochenderfer MJ, Wheeler TA, Wray KH. Algorithms for decision making. MIT press; 2022.
[24] Cardil A, Monedero S, Schag G, de Miguel S, Tapia M, Stoof CR, et al. Fire behavior modeling for
operational decision-making. Current Opinion in Environmental Science & Health. 2021;23:100291.
Available from: https://doi.org/10.1016/j.coesh.2021.100291.
[25] Noonan-Wright EK, Opperman TS, Finney MA, Zimmerman GT, Seli RC, Elenz LM, et al. Developing
the US wildland fire decision support system. Journal of Combustion. 2011;2011. Available from:
https://doi.org/0.1155/2011/168473.
[26] Calkin DE, Thompson MP, Finney MA, Hyde KD. A real-time risk assessment tool supporting wildland
fire decisionmaking. Journal of Forestry. 2011;109(5):274-80.
[27] Javaid N, Ullah I, Akbar M, Iqbal Z, Khan FA, Alrajeh N, et al. An intelligent load management system
with renewable energy integration for smart homes. IEEE access. 2017;5:13587-600. Available from:
https://doi.org/10.1109/ACCESS.2017.2715225.
[28] Ahn J, Cho S. Anti-logic or common sense that can hinder machine’s energy performance: Energy and
comfort control models based on artificial intelligence responding to abnormal indoor environments. Ap-
plied Energy. 2017;204:117-30. Available from: https://doi.org/10.1016/j.apenergy.2017.
06.079.
[29] Parvez I, Sarwat A, Debnath A, Olowu T, Dastgir MG, Riggs H. Multi-layer perceptron based photo-
voltaic forecasting for rooftop pv applications in smart grid. In: 2020 SoutheastCon. IEEE; 2020. p. 1-6.
Available from: https://doi.org/10.1109/SoutheastCon44009.2020.9249681.
[30] Zhang H, Wu D, Boulet B. A review of recent advances on reinforcement learning for smart home
energy management. In: 2020 IEEE Electric Power and Energy Conference (EPEC). IEEE; 2020. p.
1-6. Available from: https://doi.org/10.1109/EPEC48502.2020.9320042.
[31] Huang Y, Wang K, Gao K, Qu T, Liu H. Jointly optimizing microgrid configuration and energy con-
sumption scheduling of smart homes. Swarm and Evolutionary Computation. 2019;48:251-61. Avail-
able from: https://doi.org/10.1016/j.swevo.2019.04.007.
[32] Meng F, Zeng XJ, Zhang Y, Dent CJ, Gong D. An integrated optimization+ learning approach to op-
timal dynamic pricing for the retailer with multi-type customers in smart grids. Information Sciences.
2018;448:215-32. Available from: https://doi.org/10.1016/j.ins.2018.03.039.
[33] Thakur N, Han CY. An ambient intelligence-based human behavior monitoring framework for ubiq-
uitous environments. Information. 2021;12(2):81. Available from: https://doi.org/10.3390/
info12020081.
[34] Sun H, De Florio V, Gui N, Blondia C. Promises and challenges of ambient assisted living systems.
In: 2009 Sixth International Conference on Information Technology: New Generations. Ieee; 2009. p.
1201-7. Available from: https://doi.org/10.1109/ITNG.2009.169.
[35] Naghshvarianjahromi M, Kumar S, Deen MJ. Brain-inspired intelligence for real-time health situation
understanding in smart e-health home applications. IEEE Access. 2019;7:180106-26. Available from:
https://doi.org/10.1109/ACCESS.2019.2958827.
[36] Souza AT, Canha LN, Milbradt RG, Lemos CL, Michels C, Santana TA. Intelligent demand response
system integrated with micro generation and energy storage using machine learning and Internet of
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms 87
Things concepts. In: 2019 IEEE PES Innovative Smart Grid Technologies Conference-Latin America
(ISGT Latin America). IEEE; 2019. p. 1-6. Available from: https://doi.org/10.1109/ISGT-LA.
2019.8895479.
[37] Garroussi Z, Ellaia R, Lucas Jy, et al. A matheuristic for a bi-objective demand-side optimization
for cooperative smart homes. Electrical Engineering. 2020;102(4):1913-30. Available from: https:
//doi.org/10.1007/s00202-020- 00997-6.
[38] Amini MH, Imteaj A, Mohammadi J. Distributed machine learning for resilient operation of electric
systems. In: 2020 International Conference on Smart Energy Systems and Technologies (SEST). IEEE;
2020. p. 1-6. Available from: https://doi.org/10.1109/SEST48500.2020.9203368.
[39] Han T, Muhammad K, Hussain T, Lloret J, Baik SW. An efficient deep learning framework for intelligent
energy management in IoT networks. IEEE Internet of Things Journal. 2020;8(5):3170-9. Available
from: 10.1109/JIOT.2020.3013306.
[40] Vasudevan N, Venkatraman V, Ramkumar A, Sheela A. Real-time day ahead energy manage-
ment for smart home using machine learning algorithm. Journal of Intelligent & Fuzzy Systems.
2021;41(5):5665-76. Available from: https://doi.org/10.3233/JIFS-189886.
[41] Li T, Fong S, Li X, Lu Z, Gandomi AH. Swarm decision table and ensemble search methods in fog
computing environment: Case of day-ahead prediction of building energy demands using IoT sensors.
IEEE Internet of Things Journal. 2019;7(3):2321-42. Available from: https://doi.org/10.1109/
JIOT.2019.2958523.
[42] Jordehi AR. Binary particle swarm optimisation with quadratic transfer function: A new binary op-
timisation algorithm for optimal scheduling of appliances in smart homes. Applied Soft Computing.
2019;78:465-80. Available from: https://doi.org/10.1016/j.asoc.2019.03.002.
[43] Alhasnawi BN, Jasim BH, Esteban MD, Guerrero JM. A novel smart energy management as a service
over a cloud computing platform for nanogrid appliances. Sustainability. 2020;12(22):9686. Available
from: https://doi.org/10.3390/su12229686.
[44] Rossit DG, Nesmachnow S, Toutouh J, Luna F. Scheduling deferrable electric appliances in smart
homes: a bi-objective stochastic optimization approach. 2021. Available from: https://doi.org/10.
3934/mbe.2022002.
[45] Rashid RA, Chin L, Sarijari MA, Sudirman R, Ide T. Machine learning for smart energy monitoring
of home appliances using IoT. In: 2019 Eleventh International Conference on Ubiquitous and Future
Networks (ICUFN). IEEE; 2019. p. 66-71. Available from: https://doi.org/10.1109/ICUFN.
2019.8806026.
[46] Wang Z, Dogan S, Stolikj M, Kondoz AM. Towards adaptive control in smart homes: Overall system de-
sign and initial evaluation of activity recognition. In: 2017 Intelligent Systems Conference (IntelliSys).
IEEE; 2017. p. 129-36. Available from: https://doi.org/10.1109/IntelliSys.2017.8324274.
[47] Verma P, Sood SK. Fog assisted-IoT enabled patient health monitoring in smart homes. IEEE Internet
of Things Journal. 2018;5(3):1789-96. Available from: https://doi.org/10.1109/JIOT.2018.
2803201.
[48] Naghshvarianjahromi M, Majumder S, Kumar S, Naghshvarianjahromi N, Deen MJ. Natural Brain-
Inspired Intelligence for Screening in Healthcare Applications. IEEE Access. 2021;9:67957-73.
[49] Shneiderman B. Human-centered artificial intelligence: Reliable, safe & trustworthy. International
Journal of Human–Computer Interaction. 2020;36(6):495-504. Available from: https://doi.org/
10.1080/10447318.2020.1741118.
[50] Xu W. Toward human-centered AI: a perspective from human-computer interaction. interactions.
2019;26(4):42-6. Available from: https://doi.org/10.1145/3328485.
[51] Dietvorst BJ, Simmons JP, Massey C. Overcoming algorithm aversion: People will use imperfect algo-
rithms if they can (even slightly) modify them. Management Science. 2018;64(3):1155-70. Available
from: https://doi.org/10.1287/mnsc.2016.2643.
[52] Mohammadi S, Eliassen F, Zhang Y, Jacobsen HA. Detecting false data injection attacks in peer to
peer energy trading using machine learning. IEEE Transactions on Dependable and Secure Computing.
2021. Available from: https://doi.org/10.1109/TDSC.2021.3096213.
[53] Kizilcec RF. How much information? Effects of transparency on trust in an algorithmic interface. In:
Proceedings of the 2016 CHI conference on human factors in computing systems; 2016. p. 2390-5.
Available from: https://doi.org/10.1145/2858036.2858402.
[54] Doˇ
silovi´
c FK, Brˇ
ci´
c M, Hlupi´
c N. Explainable artificial intelligence: A survey. In: 2018 41st In-
ternational convention on information and communication technology, electronics and microelectron-
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms88
ics (MIPRO). IEEE; 2018. p. 0210-5. Available from: https://doi.org/10.23919/MIPRO.2018.
8400040.
[55] Nassourou M, Blesa J, Puig V. Robust economic model predictive control based on a zonotope and
local feedback controller for energy dispatch in smart-grids considering demand uncertainty. Energies.
2020;13(3):696. Available from: https://doi.org/10.3390/en13030696.
[56] Zhang L, Ren J, Yuan H, Yang Z, Wang W, Guo M, et al. Evaluation of smart healthcare systems
and novel uv-oriented solution for integration, resilience, inclusiveness and sustainability. In: 2020 5th
International Conference on Universal Village (UV). IEEE; 2020. p. 1-28. Available from: https:
//doi.org/10.1109/UV50937.2020.9426210.
[57] Rochd A, Benazzouz A, Abdelmoula IA, Raihani A, Ghennioui A, Naimi Z, et al. Design and im-
plementation of an AI-based & IoT-enabled Home Energy Management System: A case study in
Benguerir—Morocco. Energy Reports. 2021;7:699-719. Available from: https://doi.org/10.
1016/j.egyr.2021.07.084.
[58] Maadi M, Akbarzadeh Khorshidi H, Aickelin U. A review on human–ai interaction in machine learning
and insights for medical applications. International journal of environmental research and public health.
2021;18(4):2121. Available from: https://doi.org/10.3390/ijerph18042121.
[59] Boy GA. The handbook of human-machine interaction: a human-centered design approach. CRC Press;
2017.
[60] Zheng Nn, Liu Zy, Ren Pj, Ma Yq, Chen St, Yu Sy, et al. Hybrid-augmented intelligence: collabora-
tion and cognition. Frontiers of Information Technology & Electronic Engineering. 2017;18(2):153-79.
Available from: https://doi.org/10.1631/FITEE.1700053.
[61] Xiao G. Machine learning in smart home energy monitoring system. In: IOP Conference Series: Earth
and Environmental Science. vol. 769. IOP Publishing; 2021. p. 042035. Available from: https://doi.
org/10.1088/1755-1315/769/4/042035.
[62] Divya V, Leena Sri R. Intelligent real-time multimodal fall detection in fog infrastructure using ensemble
learning. In: Challenges and Trends in Multimodal Fall Detection for Healthcare. Springer; 2020. p.
53-79. Available from: https://doi.org/10.1007/978-3-030- 38748-8_3.
[63] Roy PN, Armin M, Kamruzzaman S, Hoque ME. A Supervisory Control of Home Appliances using
Internet of Things. In: 2019 International Conference on Electrical, Computer and Communication En-
gineering (ECCE). IEEE; 2019. p. 1-6. Available from: https://doi.org/10.1109/ECACE.2019.
8679314.
[64] Khir NAM, Abd Malik NNN, Latiff NMA, Said DM, Hamzah SA. Smart Micro-Grid Delivery Response
System via Internet of Things. In: 2019 IEEE 14th Malaysia International Conference on Communi-
cation (MICC). IEEE; 2019. p. 75-80. Available from: https://doi.org/10.1109/MICC48337.
2019.9037590.
[65] Hussain SM, Zafar A, Khalid R, Abid S, Qasim U, Khan ZA, et al. An efficient scheduling of electrical
appliance in micro grid based on heuristic techniques. In: Conference on Complex, Intelligent, and
Software Intensive Systems. Springer; 2017. p. 164-73. Available from: https://doi.org/10.1007/
978-3- 319-61566-0_15.
[66] Rojek L, Islam S, Hartmann M, Creutzburg R. IoT-based real-time monitoring system for a smart energy
house. Electronic Imaging. 2021;2021(3):38-1.
[67] Wang J, Srikantha P. Stealthy black-box Attacks on deep learning non-intrusive load monitoring mod-
els. IEEE Transactions on Smart Grid. 2021;12(4):3479-92. Available from: https://doi.org/10.
1109/TSG.2021.3062722.
[68] Demirezen G, Fung AS, Deprez M. Development and optimization of artificial neural network algo-
rithms for the prediction of building specific local temperature for HVAC control. International Journal
of Energy Research. 2020;44(11):8513-31. Available from: https://doi.org/10.1002/er.5537.
[69] Alves MJ, Antunes CH. A semivectorial bilevel programming approach to optimize electricity dynamic
time-of-use retail pricing. Computers & Operations Research. 2018;92:130-44. Available from: https:
//doi.org/10.1016/j.cor.2017.12.014.
[70] Mpawenimana I, Pegatoquet A, Roy V, Rodriguez L, Belleudy C. A comparative study of LSTM
and ARIMA for energy load prediction with enhanced data preprocessing. In: 2020 IEEE Sensors
Applications Symposium (SAS). IEEE; 2020. p. 1-6. Available from: https://doi.org/10.1109/
SAS48726.2020.9220021.
[71] Chatrati SP, Hossain G, Goyal A, Bhan A, Bhattacharya S, Gaurav D, et al. Smart home health monitor-
ing system for predicting type 2 diabetes and hypertension. Journal of King Saud University-Computer
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms 89
and Information Sciences. 2020. Available from: https://doi.org/10.1016/j.jksuci.2020.01.
010.
[72] van Berkel N, Skov MB, Kjeldskov J. Human-AI interaction: intermittent, continuous, and proactive.
Interactions. 2021;28(6):67-71. Available from: https://doi.org/10.1145/3486941.
[73] Oyeleke RO, Yu CY, Chang CK. Situ-centric reinforcement learning for recommendation of tasks in ac-
tivities of daily living in smart homes. In: 2018 IEEE 42nd Annual Computer Software and Applications
Conference (COMPSAC). vol. 2. IEEE; 2018. p. 317-22.
[74] Igwe OM, Wang Y, Giakos GC. Activity learning and recognition using margin setting algorithm in
smart homes. In: 2018 9th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communica-
tion Conference (UEMCON). IEEE; 2018. p. 653-8. Available from: https://doi.org/10.1109/
UEMCON.2018.8796766.
[75] Qu W, Tang J, Niu W. Research on fire detection based on multi-source sensor data fusion. In: Interna-
tional Conference on Mechatronics and Intelligent Robotics. Springer; 2019. p. 629-35. Available from:
https://doi.org/10.3390/s22124550.
[76] Smart Firefighting System for Smart Cities Adopting Fog/Edge Computing. VOLUME-8 ISSUE-10,
AUGUST 2019, REGULAR ISSUE. 2019. Available from: https://doi.org/10.35940/ijitee.
j9920.0881019.
[77] Imteaj A, Amini MH. Distributed sensing using smart end-user devices: Pathway to federated learn-
ing for autonomous IoT. In: 2019 International conference on computational science and computa-
tional intelligence (CSCI). IEEE; 2019. p. 1156-61. Available from: https://doi.org/10.1109/
CSCI49370.2019.00218.
[78] Klaus P, Zaichkowsky JL. The convenience of shopping via voice AI: Introducing AIDM. Journal of
Retailing and Consumer Services. 2022;65:102490. Available from: https://doi.org/10.1016/j.
jretconser.2021.102490.
[79] Jha DN, Alwasel K, Alshoshan A, Huang X, Naha RK, Battula SK, et al. IoTSim-Edge: a simulation
framework for modeling the behavior of Internet of Things and edge computing environments. Software:
Practice and Experience. 2020;50(6):844-67. Available from: https://doi.org/10.48550/arXiv.
1910.03026.
[80] Kashef R. Enhancing the Role of Large-Scale Recommendation Systems in the IoT Context. IEEE
Access. 2020;8:178248-57. Available from: https://doi.org/10.1109/ACCESS.2020.3026310.
[81] Djalaei B, Salehi ME. Optimizing embedded neural networks based on saturated states. In: 2018
Real-Time and Embedded Systems and Technologies (RTEST). IEEE; 2018. p. 24-30. Available from:
https://doi.org/10.1109/RTEST.2018.8397075.
[82] Kumar C, Mahdi HF. A real time algorithm for the machine to machine communication-smart warrior
system. In: 2017 IEEE 13th Malaysia International Conference on Communications (MICC). IEEE;
2017. p. 213-7. Available from: https://doi.org/10.1109/MICC.2017.8311761.
[83] Wolf C, Blomberg J. Evaluating the promise of human-algorithm collaborations in everyday work
practices. Proceedings of the ACM on Human-Computer Interaction. 2019;3(CSCW):1-23. Available
from: https://doi.org/10.1145/3359245.
[84] Salvi S, Geetha V, Kamath SS. Jamura: a conversational smart home assistant built on Telegram and
Google Dialogflow. In: TENCON 2019-2019 IEEE Region 10 Conference (TENCON). IEEE; 2019. p.
1564-71. Available from: https://doi.org/10.1109/TENCON.2019.8929316.
[85] Dahlgren K, Pink S, Strengers Y, Nicholls L, Sadowski J. Personalization and the Smart Home:
questioning techno-hedonist imaginaries. Convergence. 2021;27(5):1155-69. Available from: https:
//doi.org/10.1177/13548565211036801.
[86] Ashenov N, Myrzaliyeva M, Mussakhanova M, Nunna HK. Dynamic Cloud and ANN based Home
Energy Management System for End-Users with Smart-Plugs and PV Generation. In: 2021 IEEE Texas
Power and Energy Conference (TPEC). IEEE; 2021. p. 1-6. Available from: https://doi.org/10.
1109/TPEC51183.2021.9384980.
[87] Zhu J. Development Strategy of Artificial Intelligence Technology in the Field of Ice and Snow Sports.
In: International Conference on Application of Intelligent Systems in Multi-modal Information Analyt-
ics. Springer; 2021. p. 619-26. Available from: https://doi.org/10.1007/978-3-030-74811-1_
89.
[88] Zhou Z, Yu H, Shi H. Human activity recognition based on improved Bayesian convolution network
to analyze health care data using wearable IoT device. IEEE Access. 2020;8:86411-8. Available from:
https://doi.org/10.1109/ACCESS.2020.2992584.
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms90
A. Appendix
Table 1.: Summary of main challenges for human interaction with
decision-making algorithms in energy management
Main Issues Energy Allocation/Energy consumption forecasting
/Home Appliances scheduling and controlling /Anomaly
detection/Device management
Machine Behaviour Handle unexpected outcome within system.
Human-machine collabo-
ration
Investigate what kinds of collaboration energy manage-
ment system needs to improve its results.
Machine intelligence Design different engagements to involve human in the
loop, for example, for energy allocation, users might ex-
pect low-level engagement, for home appliances schedul-
ing and controlling, users might expect high-level engage-
ment.
Explainability of machine
output
Set different kinds of explainability towards different
kinds of function.
Autonomous characteris-
tics of machines
Handle some situation when the operation is not as ex-
pected
User interface System may set the feedback invisible to avoid interrup-
tion.
Ethical design Set different kinds of decision-making authority towards
different functions.
Table 2.: Summary of main challenges for human interaction with
decision-making algorithm in e-healthcare
Main Issues Simple diagnose / Simple screening / Emergency alert
Machine Behaviour Support unexpected outcome for elder person.
Human-machine collabo-
ration
Form the long-term collaboration between machine and
human to build human’s trust and get continually human
data input.
Machine intelligence Involve users into the system to make users final control
towards decisions that machines make.
Explainability of machine
output
Explain to the end users of the process of machine deci-
sion - diagnose/screening/alert
Autonomous characteris-
tics of machines
Handle with unanticipated situation especially emergency
situation.
User interface Handle with different devices - mobile, smartwatch, tablet
with different interaction modes - voice, text, picture,
even gesture in a system.
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms 91
Ethical design Privacy issues are significant in healthcare and also the
authority of decision-making.
Table 3.: The overview of decision-making algorithms application area
in smart home
Application Areas Papers
Energy management [61] [46] [62] [44] [42] [32] [30] [63] [40] [57] [37] [45]
[57] [44] [36] [64] [39] [41] [43] [65] [66] [67] [68] [69]
[52] [70] [38]
Healthcare [33] [71] [72] [56] [73] [48] [47] [35]
Others [74] [75] [76] [77] [78] [4] [79] [80] [62] [81] [82] [83]
[84] [85]
Table 4.: The functions of decision-making algorithms in energy man-
agement
Functions Papers
Energy allocation [62] [38] [37] [76] [32]
Home appliances schedul-
ing and controlling
[39] [69] [44] [42] [57] [86] [44] [43] [13] [63] [41] [45]
[87]
Energy forecasting [82] [39] [70] [66]
Energy forecasting and
anomaly detection
[61]
Anomaly detection [52] [67]
Device Management [50] [62]
Table 5.: The functions of decision-making algorithms in healthcare
Functions Papers
Simple diagnose [47] [83] [71] [88] [73] [35]
Simple screening [48]
Emergency alert [33]
L. Jin and A. Boden / Review on the Application Areas of Decision-Making Algorithms92
