PreprintPDF Available

Beyond Mobile Apps: A Survey of Technologies for Mental Well-being

Authors:

Abstract and Figures

Mental health problems are on the rise globally and strain national health systems worldwide. Mental disorders are closely associated with fear of stigma, structural barriers such as financial burden, and lack of available services and resources which often prohibit the delivery of frequent clinical advice and monitoring. Technologies for mental well-being exhibit a range of attractive properties which facilitate the delivery of state of the art clinical monitoring. This review article provides an overview of traditional techniques followed by their technological alternatives, sensing devices, behaviour changing tools, and feedback interfaces. The challenges presented by these technologies are then discussed with data collection, privacy and battery life being some of the key issues which need to be carefully considered for the successful deployment of mental health tool-kits. Finally, the opportunities this growing research area presents are discussed including the use of portable tangible interfaces combining sensing and feedback technologies. Capitalising on the captured data these ubiquitous devices offer, state of the art machine learning algorithms can lead to the develop
Content may be subject to copyright.
1
Beyond Mobile Apps: A Survey of Technologies
for Mental Well-being
Kieran Woodward, Eiman Kanjo, David Brown, T.M. McGinnity, Becky Inkster, Donald J Macintyre &
Athanasios Tsanas
Abstract—Mental health problems are on the rise globally and strain national health systems worldwide. Mental disorders are closely
associated with fear of stigma, structural barriers such as financial burden, and lack of available services and resources which often
prohibit the delivery of frequent clinical advice and monitoring. Technologies for mental well-being exhibit a range of attractive
properties which facilitate the delivery of state of the art clinical monitoring. This review article provides an overview of traditional
techniques followed by their technological alternatives, sensing devices, behaviour changing tools, and feedback interfaces. The
challenges presented by these technologies are then discussed with data collection, privacy and battery life being some of the key
issues which need to be carefully considered for the successful deployment of mental health tool-kits. Finally, the opportunities this
growing research area presents are discussed including the use of portable tangible interfaces combining sensing and feedback
technologies. Capitalising on the captured data these ubiquitous devices offer, state of the art machine learning algorithms can lead to
the development of a robust clinical decision support tools towards diagnosis and improvement of mental well-being delivery in
real-time.
Index Terms—Pervasive computing, Machine learning, Ubiquitous computing, Physiological Measures, Diagnosis or assessment
F
1INTRODUCTION
Mental health problems constitute a global challenge that af-
fects a large number of people of all ages and socioeconomic
backgrounds. The World Health Organisation [1] defines
the well-being of an individual as being encompassed in
the realisation of their abilities, coping with the normal
stresses of life, productive work and contribution to their
community. Hectic modern lifestyles contribute to daily
stress and a general decline in mental health, as 59% of
UK adults currently experience work-related stress [2]. This
makes stress the leading cause of sickness absences from
work, with about 70 million days lost each year at an esti-
mated cost of £2.4 billion [2]. Furthermore, the Physiological
Society [3] reported 18-24 year-olds were the most stressed
age group with students studying higher degrees exhibiting
considerable stress levels, where the majority (60.9%) of
the high-risk undergraduate students rated their mental
health as poor or very poor [4] showing the negative impact
modern lifestyles are having on mental well-being.
Traditionally clinical visits are infrequent and intermit-
tent, representing a very small time window into patients’
lives, where clinicians are challenged to decipher the possi-
ble manifestation of symptoms and disease trajectory. Fur-
ther problems are often encountered with patients’ recall
bias, when they are asked to provide details regarding
certain symptoms in detail in retrospect. In many clinical
fields patients are encouraged to use standardized clinical
questionnaires, typically in the form of Patient Reported
Outcome Measures (PROMs) or experience sampling [5],
[6] to understand the longitudinal variability of mental
health symptom trajectory over months in-between clini-
cal visits. A common problem encountered during clinical
psychiatric assessments is that the questions asked about
patients’ mood and pain can be impacted by an unreliable
autobiographical memory [7]. An alternative to traditional
methods involves smartphone applications that can provide
a variety of tasks including symptom assessment, talking
therapies, psycho-education and monitoring the efficiency
of treatment [8].
Mental disorders and poor mental well-being often lead
to physiological changes. For example, stress is defined
as the non-specific response of the body to any demand
for change, resulting in reduced heart rate variability [9],
lower skin temperature [10] and increased galvanic skin
resistance [11], [12]. Technological advances have led to
smartphones and tangible devices which are user interfaces
in which a person interacts with digital information through
the physical environment; these can incorporate sensors
to measure physiological changes and help alleviate the
stress people experience. This provides new opportunities to
utilise non-invasive technology for behavioural health care
in order to assess and aid mental health conditions such as
anxiety and stress accurately in real-time. With the rise of
pervasive computing tangible user interfaces are increasing
in popularity as they combine the use of software and
hardware to provide interfaces or systems that can be ma-
nipulated. Multimodal interactions are currently used for a
wide variety of purposes such as improving communication
but mental well-being is an area where these interactions
could have a profound impact [13], [14].
This article provides a literature survey and taxonomy
that aims to explore the use of innovative interfaces that go
beyond mobile applications to assess the potential of new
technologies and how they can be utilised to improve men-
tal well-being. It first examines traditional methods to assess
and improve mental well-being and then the technological
alternatives are explored aiming to address the following
2
research questions:
1) Can technology supplement traditional mental well-
being assessment techniques?
2) Can mHealth apps be used to accurately infer mental
well-being in real-time?
3) Can machine learning be paired with sensors to improve
mental well-being state classification?
4) How can behaviour changing tools be used to help
improve mental well-being?
5) Is it possible to teach people using technology how to
improve their mental well-being?
6) Can a combination of sensing and feedback technologies
be used to improve mental well-being in real-time?
After these six highlighted areas have been reviewed, the
challenges, tools, and opportunities modern technological
advancements present for mental well-being are discussed.
2ATAXONOMY OF MENTAL WELL-BEING TECH-
NOLOGIES RESEARCH
2.1 Traditional Assessment tools and techniques
Traditional methods used to assess mental well-being often
utilise self-reporting for example, when people record their
emotions and stresses in a diary that can be assessed and
monitored to help establish stressful triggers [15] [16] or
the use of validated questionnaires to measure daily life
stresses, symptoms, etc. Examples of questionnaires include
the Positive and Negative Affect Schedule (PANAS) [17],
Brief Job Stress Questionnaire [18], Quick Inventory of
Depressive Symptomatology (QIDS) [19] and the validated
Patient Health Questionnaire (PHQ-9) [20].
Diagnostic interviews are performed by psychia-
trists/care professionals by asking service users and their
friends or family about their symptoms, experiences,
thoughts, feelings and the impact they are having. Diag-
nostic interviews allow for a diagnosis to be made accord-
ing to standard classification systems such as ICD-10 [21]
and DSM-5 [22] and these are used in conjunction with
a biopsychosocial formulation to construct a management
plan, which can include talking therapies which teach peo-
ple to learn new behaviours, and develop greater resilience
(e.g. to cope with stressful events) [23] [17]. Discussions with
trained experts leads to potentially identifying underlying
problems and can be used as treatment by teaching people
new behaviours (e.g., to cope with stressful events).
Self-reporting diaries can take considerable time to as-
sess as they must be completed over a long period to gain
useful insights [24]. Symptom self-reporting is not always
accurate due to poor recall; for example a study investigated
how accurately individuals self-reported the number of fruit
and vegetables eaten, with accuracies ranging from 40.4%
to 58% [25]. Additionally, all of the traditional assessment
methods require people to be aware of their mental health
and actively seek help which often many forego due to
fear of social stigma and lack of available resources [26],
[27]. A technological alternative that could actively mon-
itor patients’ mental health state and provide methods to
improve their mental well-being would be beneficial as it
could improve accessibility to mental health tools [28].
2.2 Technological supplements to traditional assess-
ment techniques
Can technology supplement traditional mental well-being
assessment techniques?
Research reviewed in this category reports on the devel-
opment and evaluation of mental well-being technologies
to modernise traditional techniques such as self-reporting
diaries using mobile apps and tangible interfaces.
2.2.1 Overview of mHealth apps
With the high prevalence of smartphone ownership [29]
access to treatment which is flexible and fits in with people’s
lifestyles is greatly enhanced [30]. Those at risk of men-
tal health problems often have difficulty accessing quality
mental health care [31], especially when symptoms first
manifest [32], demonstrating the need for more accessible
help. An Australian survey found that 76% of people would
be interested in using mobile phone apps for mental health
monitoring and self-management [33] illustrating the high
demand for mHealth apps because of their convenience and
accessibility.
Many apps have been developed to modernise and
advance existing practices of recording mental well-being.
Numerous mental health diary apps are available to down-
load, although these are effectively digital representations
of existing self-reporting diaries using new techniques such
as the touchscreen, volume buttons and monitoring notifi-
cations [34], [35], [36]. However, using a phone in public
is more socially acceptable than completing a paper form
allowing monitoring to be completed discreetly in real-
time, unlike paper forms which are often completed after
the event occurred resulting in less accurate data being
recorded [24]. A problem many apps face is the frequency
for eliciting PROMs which may under-represent the true
symptom fluctuation. Given that mood is highlt variable,
clinically useful information is likely in the daily fluctu-
ations of mood for many cohorts suffering from mental
disorders. Previous research demonstrates the possibility of
eliciting daily responses to assess mental health with very
good adherence over a 1 year period [37], demonstrating the
feasibility of longitudinal daily PROMs engagements by two
cohorts diagnosed with bipolar disorders and borderline
personality disorders.
More recently, chatbot apps are being developed to as-
sess mental well-being by, in some cases mimicking conver-
sation with users via a chat interface [38], thus removing
the requirement to continuously self-report. A survey given
to 5,141 participants in the age range 16-24 years showed
nearly two thirds would be comfortable with a chatbot
giving them a diagnosis [39]. Chatbots can utilise artificial
intelligence to reduce their reliance on predefined scripts
and deliver individualised therapy suggestions based on
linguistic analysis and enhance user engagement [40]. Fur-
thermore, chatbots can generate emotional responses by
using context sensitive advanced natural language-based
computational models to detect user state and emotions and
continuously provide personalised responses [41]. However,
fully generative models for chatbots can result in hurtful
comments on sensitive topics such as race [42] and mental
health [43] [44] [45] which cannot be permitted in the
3
Fig. 1: Structure of mental health tools and technologies reviewed
domain of mental well-being as in this field, we must
go beyond striving to pass the Turing test to additionally
prioritise safety. It is of central importance that ethics and
safety are constantly considered in this field, especially
when working with young and vulnerable populations [46].
Text-based conversational Chatbots can go beyond as-
sessing mental well-being with some actively aiming to
improve users’ well-being. Wysa [47] and Woebot are two
such chatbots that participants have found to be helpful and
encouraging resulting in mood improvements [48]. Other
mental well-being chatbots show positive reception of the
intervention but also demonstrate the importance of an intu-
itive user interface and the potential for artificial intelligence
to understand the meaning of sentences without relying on
pre-programmed keywords as this is a common criticism
of chatbots [49]. People are showing a growing interest in
this type of bot-based interactive support as Wysa has been
downloaded over 500,000 on the Google Play store alone
[50].
iOS and Android app stores allow any developer to pub-
lish mental health apps without any precautionary checks
or safeguards that go beyond standard malicious program
assessment, such as also verifying whether apps have been
scientifically evaluated. Figure 3 reviews and compares five
of the most popular mental well-being apps from the iOS
and Android app stores. The five apps have been developed
by a wide range of organisations with varying levels of
features and effectiveness. These apps show that while app
stores allow for feature rich intuitive apps such as ‘Daylio’,
‘Wellmind’ and ‘Moodpath’ they also allow for untested,
unintuitive apps such as ‘Calm’ and ‘What’s up?’ which
could have a negative impact on users’ mental well-being.
To prevent this, app stores could be more rigorous in their
testing and approval of mental well-being apps to prevent
possible detrimental impact on people’s mental health.
Mental health apps are getting increasing attention and
become profitable businesses. For example, Calm, a medi-
tation app which is free to download and use has recently
been valued at $1 billion [51] even though there have been
no clinical trials or evaluation to confirm the mental well-
being benefits of using the app. More worryingly Apple
and Google have endorsed Calm by making it the 2017 app
of the year and the 2018 editor’s choice respectively [52],
which could create a strong impact on people’s uptake of
the app despite the lack of scientific evidence supporting its
use. There are studies showing the benefits of mindfulness
technology interventions [53] [54] but hitherto no evaluation
has proved the benefits of Calm over evaluatedcompeting
apps (some of which have been scientifically validated).
Similarly, Calm Harm an app designed to prevent self
harm is featured on the NHS digital library [55] and
while the app has been developed by a psychologist using
principles of practice there have been no clinical trials or
evaluation to confirm this. The presence of Calm Harm on
the NHS digital library implies its effectiveness although
the NHS digital library operates three distinct badges for
apps, approved, being tested and no badge, Calm Harm has
received no badge meaning it meets NHS quality standards
for safety, usability and accessibility but it is not currently
being tested by the NHS for clinical effectiveness. The badge
system used by the NHS allows any app meeting their
unpublished standards to be prominently displayed and
easily misrepresented as clinically tested.
Headspace currently has over ten million downloads on
the Android Play store alone, undelining the immense pop-
ularity of mobile well-being apps. Unlike Calm, Headspace
has published research findings demonstrating it can help
reduce stress by 14% [56], increase compassion by 23% [57],
reduce aggression by 57% [58] and improve focus by 14%
[59]. However, most of these studies were small scale with
the longest period people were followed being just thirty
days.Another research study reported that using the app
over a six week period resulted in no improvements in
critical thinking performance [60]. Additionally, there has
been no follow-up after the initial studies and as some
studies lasted as little as ten days, raising some concerns that
the positive outcomes from the app may only be apparent
during an individual’s initial period of use.
Figure 2 shows the six most popular mental health apps
on Android with each of them achieving an overall rating
of at least 4 out of 5. However, the number of downloads
vastly varies as ’Headspace’, ’Calm’ and ’Daylio’ make up
the vast majority of downloads with a combined total of 25
million whereas next most popular apps only amass 500,000
4
downloads each, showing that receiving favourable reviews
does not necessarily lead to mass downloads. Evaluated
apps developed by respected organisations also do not
necessarily result in popularity as ’Wellmind’ developed by
the NHS has only been downloaded around 10000 times and
received an average rating of 3.4 out of 5, displaying users’
preference of usability and functionality.
Fig. 2: Comparison of the six most popular mental health
apps on Android
Both the iOS app store and Google play store do not
have a dedicated category for mental well-being apps mean-
ing they are combined with other health and fitness apps
increasing the difficulty of finding relevant apps. Figure 4
below shows the subcategories of the top 100 free and paid
health and fitness apps on the iOS App store in September
2018. The majority of smartphone apps within the health
and fitness category are dedicated to exercising with only
a small proportion of apps for stress or mood monitoring
and these apps were generally lower in the charts obscuring
them from users. App stores could improve the visibility
of tested mental health apps through a dedicated mental
health category which may facilitate the uptake of well-
established smartphone apps which have received positive
feedback from users.
Additional apps have been developed by researchers
that actively aim to improve mental health and well-being
such as mobile stress management apps that use stress
inoculation training to prepare people to better handle
stressful events. Studies show stress inoculation apps were
consistently successful in reducing stress in participants and
increasing their active coping skills [61] [62] [63]. The study
by Grassi et al. demonstrated that mHealth apps are not
only capable of augmenting traditional techniques to help
monitor conditions but they can also be used to educate
users on techniques to actively improve their mental well-
being.
Fig. 4: Comparison of the six most popular mental health
apps on Android
A smartphone app, FOCUS, has been developed to
proactively ask users with schizophrenia about their mood,
feelings and well-being multiple times each day to provide
relevant coping strategies [64]. This allows the app to go be-
yond traditional self-reporting as it educates users on meth-
ods to help immediately after an issue has been reported
which is only possible using technology that people have
continuous access to such as smartphones. Focus demon-
strated a reduction of positive symptoms of schizophrenia,
and depression, when trialled by 33 participants over 4
weeks. A common issue with mental well-being apps is low
user engagement. However, Focus was used by participants
on 86.5% of days averaging 5.2 times each day over 30 days
and Oiva, a mental well-being training app [65] was on
average used every third day for 12 minutes over a 30 day
period demonstrating the possibility for mental well-being
technologies to be highly engaging.
While apps could be considered as an alternative to
seeking professional help some apps have been designed to
work in conjunction with clinicians such as Post-Traumatic
Stress Disorder (PTSD) coach. The app allows users to learn
more about PTSD, track symptoms, set up a support net-
work and provides strategies for coping with overwhelming
emotions. 10 US veterans with PTSD were assigned to use
PTSD Coach independently while another 10 used the app
with the support of their primary-care providers [30]. At
the end of the trial, seven of the ten patients using the
app with support showed a reduction in PTSD symptoms,
compared with just three of the patients who used the app
independently. Apps used with care providers show more
potential for effective treatment in the small sample trials
although this still requires users to actively seek help [66].
Pairing apps with psychiatrists’ and psychologists’ sup-
port has been shown to be successful resulting in a range of
apps using content explicitly created by psychiatrists such
as Rizvi et al. [67] who developed the app DBT Field Coach
to provide instructions, exercises, reminders, games, videos
and messages to help people cope with emotional crises. The
results of that study demonstrate the 22 participants used
the app frequently over at least 10 days and it was successful
in reducing intense emotions, reducing substance use crav-
ings and improving symptoms of depression without the
need to visit a clinician [67]. This app again shows the suc-
cess of apps utilising psychiatrists and clinicians although
as this app only used content created by psychiatrists, it
5
Fig. 3: Comparison of popular mental health apps in the Android and iOS app stores
negates the need to visit clinicians increasing accessibility.
Mobile health apps provide many advantages over tradi-
tional techniques including improved accessibility, real-time
symptom monitoring, reduced cost and reduced barriers to
access [68]. The apps reviewed demonstrated the potential
for mHealth apps to improve mental well-being however
majority of the apps have been tested in limited number of
clinical trials. One of the main shortcomings of available
smartphone apps is the lack of personalised features as
many treatments and strategies have to be individually
tailored [69].
2.2.2 Tangible interfaces
An alternative method to enhance existing techniques is
through the use of tangible interfaces which are user in-
terfaces in which a person interacts with digital information
through the physical environment. This presents new op-
portunities as Matthews and Doherty [70] and Niemantsver-
driet and Versteeg [71] have found that people are more
likely to create stronger emotional attachments with physi-
cal devices rather than digital interfaces such as apps.
Many existing tangible devices have been utilised to
promote communication and provide an easy method to
express emotions and mental well-being state that can often
be difficult to communicate. These devices provide a tech-
nological alternative to traditional self-reporting allowing
users to report their current mental well-being in real-
time. Emoball [72] is one such device that allows users to
record their mood by squeezing an electronic ball making
users conscious of their current mood. While this device
only allows users to report a limited number of emotions
participants did believe mental well-being and education
were the areas where devices to report emotions could be
of most use. A smaller, portable device that works similarly
is Keppi [73], which allows users to squeeze to record low,
medium or high pain.
Another tangible approach to self-report is the mood TUI
[74] which as well as allowing users to record their emotions
also collected relevant data from the user’s smartphone
including location data and physiological data such as heart
rate. Participants found the use of a tangible interface very
exciting, although when the device was tested with users,
they felt the device was too large and they would lose
motivation to continue using it for an extended period.
This feedback shows the use of tangible user interfaces
excites users but the design and functionality must be
6
prioritised. Mood sprite [75] is another handheld device
developed to help people suffering from anxiety and stress
by using coloured lights and an infinity mirror to assist
with relaxation. The device records the time users create
new sprites allowing them to be revisited much like a
diary again showing ways in which tangible interfaces
can accompany traditional techniques to make treatment
more accessible and user-centric. The device educates users
similarly to traditional self-reporting diaries by allowing
users to recall their emotions but is more engaging with
different coloured lights representing different times and
moods promoting continued use. These devices allow users
to educate themselves of their emotional state over time as
well as share this information with other people such as
clinicians and family members. However, a common issue
with mental health tangible interfaces is that they remain
largely unproven and even those that have been trialled
with users such as Mood sprite have been done so in small-
scale trails that lack statistical power.
A problem often encountered with sharing mental health
state is privacy as people only wish to share this information
with people they trust and those who act responsibly with
the user’s best interests in mind. Subtle Stone [76]; is a
tangible device that allows users to express their current
emotion through a unique colour displayed on a stone,
limiting the number of people to whom users expose their
emotions. Subtle stone was tested with eight high school
students in their language class with the teacher able to
view the data in real-time using an app. The study showed
the use of colours to represent emotions was well received
with students liking the anonymity it provided along with
finding it easier to use than words. Subtle Stone both allows
users to communicate their emotions privately and monitor
their own emotions over time proving clear advantages over
traditional self-reporting methods.
A tangible interface used to detect stress in real-time
without the need to self-report is Grasp, which was tested
with anxious participants in a dentist’s office [77]. Par-
ticipants were able to squeeze Grasp whenever they felt
stressed and the device detected how much pressure was
exhorted and displayed this data on a mobile app. This
device allowed users to quickly and easily record their
anxiety in real time which could be useful for monitoring
stress over long periods as it does not rely on participants
recording stressful events in a diary. Force sensors have also
been used to create a tactile ball that allows for the ma-
nipulation of music by squeezing different areas of the ball
along with movement detected by an accelerometer [78].
The research concluded squeeze music could successfully be
used for music therapy with children as it promoted positive
emotions through tactile input and music. Sensors such
as force sensors have been shown to provide an intuitive
method of interaction for tangible user interfaces and show
the possibility for additional sensors to be utilised when
educating, detecting and improving mental well-being that
is not possible when using smartphones or traditional tech-
niques.
Overall, there are multiple tangible interfaces that go be-
yond mobile apps to provide a variety of purposes including
self-reporting of emotions, relaxation and communication.
Tangible interfaces have been shown to allow people to
successfully report and monitor their emotions over time
as well as communicate them which can often be difficult
for people experiencing mental well-being challenges. When
developing tangible mental well-being devices the design
needs to be carefully considered to ensure it is effective
and not damaging. Guidelines [79] have been produced to
ensure mental health technologies are successfully devel-
oped which include designing for client and therapist users,
making the system adaptable and sustainable and provid-
ing flexibility in the delivery of support. Overall tangible
interfaces and mobile apps provide new opportunities to
enhance existing assessment methods as the convenience
and additional functionality lead these technological alter-
natives to improve the reporting and communicating of
mental well-being state.
2.3 Sensing mental well-being state using mHealth
apps
Can mHealth apps be used to accurately infer mental health
well-being in real-time?
Apps have been shown to enhance traditional assess-
ment techniques but by utilising sensors within phones the
capability of apps is greatly enhanced enabling them to pos-
sibly better detect mental well-being arguably without the
need to self-report. Smartphones are capable of collecting a
vast amount of data such as location, motion and phone use
which can result in many features being extracted to train
machine learning classifiers. It is possible to use the data
collected from smartphones to determine emotions with a
70% accuracy utilising machine learning to process the data
[80]. The possibility to automatically infer emotion based
on smartphone use is extremely valuable in determining
mental well-being as the emotions people feel can give clear
insights into their mental health.
In addition to using phones’ sensors to detect mental
well-being, it may be possible to use the phone’s touch-
screen to sense stress. Using an infrared touchscreen to
measure photoplethysmograph (PPG) it was possible to
recognise stress with accuracies of 87% and 96% across two
tests, a vast improvement upon previous touchscreen based
stress detection [81]. However, infrared touchscreens are
rarely used especially within smartphones, the possibility
of measuring stress through capacitive touchscreens could
have a more significant impact.
Smartphone apps have also been paired with wrist-worn
sensors to infer mental well-being by allowing for a high
magnitude of data to be collected [82]. The collected data
was expressed using 15 multimodal features ranging from
physiological data such as skin conductance to phone usage
data such as screen time duration. The 15 sets of features
were then trained with a variety of classifiers and the
accuracy of the different features were examined for each
classifier. The system was capable of detecting stress with a
75% accuracy, with some of the features such as increased
acceleration during sleep and high evening phone use being
more beneficial than others in determining stress. Similarly,
a wrist sensor along with a mobile app and a self-reported
PHQ-8 and PHQ-4 depression scores were used to quantify
depression symptoms in 83 undergraduate college students
across two 9-week periods by measuring phone use, heart
7
rate, sleep and location [83]. That study concluded students
who reported they were depressed were more likely to
use their phone at study locations, have irregular sleep,
spend more time being stationary and visit fewer places.
They demonstrated that they could automatically detect
depression with a 69.1% precision when evaluated against
the PHQ-4 depression subscale [84], this could be improved
if additional physiological sensors were included such as
skin conductance or skin temperature. In addition to phys-
iological sensors location could be used to assess mental
well-being as movement patterns and uncertainty in visits
has been shown to be predictive of the Quick Inventory
of Depressive Symptomatology (QIDS) [85]. These studies
demonstrate the potentially powerful combination machine
learning, sensors and mobile apps provide when tested in
high quality trials to automatically determine stress levels.
BreathWell, [86] which has been developed for Android
Wear smartwatches has been designed to assist users in
practising deep breathing to reduce stress from PTSD al-
though the app has limited functionality to determine stress
as it only uses the user’s heart rate. Despite the limited
functionality, all seven participants believed the app could
help them and preferred the app being incorporated into a
wearable device making it more convenient to use although
the extent of the trial was extremely limited. Apps for smart-
watches have great potential to sense mental well-being as
they already contain relevant sensors and developers can
use existing software and APIs to develop the apps.
Figure 5 shows widely used sensors contained within
smartphones and smartwatches and how apps could fur-
ther capitalize on the data collected from these sensors to
assess mental well-being more accurately. Some sensors are
already widely utilised such as heart rate as this can be di-
rectly associated with mental state, but other commonplace
sensors such as the camera, GPS and accelerometer could be
used more effectively within mHealth apps.
Overall, the use of mobile and wearable apps to de-
tect mental well-being could have considerable impact on
people being able to monitor their mental well-being un-
obtrusively in real-time although additional sensors and
the extraction of more robust, meaningful and personalised
features from the raw data is required to more accurately
assess mental well-being.
Tangible interfaces present a more significant oppor-
tunity than mobile apps to sense mental well-being as a
variety of sensors can be incorporated along with machine
learning techniques to classify the data. Machine learning
is vital to accurately infer mental well-being. There are
numerous sensors that when combined with sufficiently
trained machine learning classifiers can be used to assess
mental well-being in real-time.
Non-invasive physiological sensors present the most sig-
nificant opportunity to assess mental well-being. The main
measures for stress are brain wave activity, Galvanic Skin
Response (GSR) and Heart Rate Variability (HRV) [87]. GSR
is often used to detect mental well-being as it directly corre-
lates to the sympathetic nervous system [88]. Near-Infrared
Spectroscopy is a non-invasive sensor that measures oxyhe-
moglobin and deoxyhemoglobin, research has shown this
can be used to detect mental stress similar to GSR [89] but
is more challenging to use outside of laboratories due to its
large size and placement on the forehead. Stress can also be
detected from brain activity using ElectroEncephaloGrams
[90] (EEG) as Khosrowabadi et al. demonstrates using eight
channels to classify students’ stress during exams with over
90% accuracy [91].
A wearable device that aimed to detect stress mea-
sured ElectroCardioGram (ECG), GSR and ElectroMyoGra-
phy (EMG) of the trapezius muscles [92]. Principal compo-
nent analysis reduced 9 features from the sensor data to 7
principal components. 18 participants completed three dif-
ferent stressors; a calculation task, a puzzle and a memory
task with a perceived stress scale questionnaire completed
before and after each task. The principal components and
different classifiers were used to detect stressed and non-
stressed states with an average of almost 80% classification
accuracy across the three tests compared with the question-
naire results. However, this study only detected two states;
stressed and non-stressed and was conducted in a controlled
environment so it is not known how accurate it is in daily
life as physiological signals can be affected by factors other
than mental well-being.
HRV is commonly used to measure stress as this is the
variation in time between heartbeats meaning the lower the
HRV, the more likely the user is to be stressed [93]. It is
possible to measure HRV using electrocardiograms [94] but
in 1997 it was found that finger pulse amplitude decreased
significantly during mental tasks [95] leading to HRV being
accurately measured using PhotoPlethysmoGgraphy (PPG)
which are easier to use and cheaper then ECGs as they
only 1 contact point. There are 3 types of PPG; transmitted,
reflected and remote. Transmitted signals are often used
in medical monitoring [96], remote signals use cameras to
detect changes to measure HRV by monitoring skin colour
changes [97] [98] and reflected which measures the signal
reflected from an LED using light sensing photodiodes to
measure HRV, making this the smallest and most convenient
method to use in tangible interfaces [99].
Both GSR and HRV were used in a wearable device to
measure stress during driving [100]. The wearable device
took measurements over a 5-minute period to detect stress
levels with an accuracy of 97.4% and found that HRV
and skin conductance are highly relatable making them
extremely useful in detecting mental state. The ability to use
sensors to measure HRV and skin conductance allows for
small wearable devices to accurately determine stress levels
in real-time and should be further utilised to detect stress,
anxiety and mental well-being. However, physiological sig-
nals do not account for the context in which the devices are
used as the context can play a significant role in the users’
perceived stress levels meaning additional environmental
sensors may also be required [101].
Another non-invasive sensor that has previously been
used to detect stress is a skin temperature as it can indicate
acute stressor intensity as stress often results in skin tem-
perature changes [102]. One study [103] used a wearable
device that contained multiple sensors including skin con-
ductance, skin temperature and motion and provided it to
6 people with dementia and 30 staff in a nursing home for
2 months. The device aimed to automatically detect stress
and categorise it into one of five levels, the accuracy for
each of these levels varied from 9.9% to 89.4% showing an
8
Fig. 5: Possible uses of smartphone and smartwatch sensors in relation to mental well-being
extremely wide variation as when the threshold was raised
fewer events were classified as stress because of the harder
criteria, in turn, increasing precision. ”Accurately assessing
stress levels is extremely useful as it allows for only the
required stress to be recorded depending on whether all
data or a higher accuracy is required.
Deep learning possess benefits beyond those of machine
learning including the capability to classify the raw sensory
data without the need for manually designing features to
be extracted from the raw data before feeding in a classifier
or regressor. Recurrent Neural Networks (RNN) relying on
Long Short-Term Memory are especially valuable for use
with sensor data as they are fundamental in distinguishing
similar data which differ only by the ordering of the samples
which can often dictate differences in mental health [104].
RNNs helped increase the accuracy of classifying daily
activities by 4% [104] and have also been used to accurately
classify raw ECG signals [105] and physical movement from
a wearable [106] but little research has been conducted in
using recurrent neural networks to classify mental well-
being.
Convolutional Neural Networks (CNN) also have the
capability to expand the accuracy of mental well-being
classification. CNNs have traditionally been used to classify
images and speech due to their ability to scale invariance of
a signal but recently have been used to classify raw sensor
data. The inputs in a convolutional layer connect to the
subregions of the layers instead of being fully-connected as
in traditional neural networks. These inputs share the same
weights, therefore the inputs of a CNN produce spatially-
correlated outputs. Data recorded from activity trackers
was classified using a CNN with relaxed weight sharing
as unlike in images the same pattern appearing in different
frames may be behaving differently. The CNN provided an
accuracy of up to 96.88% outperforming the previous best
algorithm [107]. CNNs have been shown to be useful in the
classification of raw sensor data but care needs to be taken
with the level of weight sharing.
A CNN and an RNN have been combined to allow
raw data to be classified more accurately [108] [109]. This
deep learning approach is capable of using raw data to
automate the feature extraction and selection. This approach
to classifying emotions from physiological, environmental
and location data outperformed traditional multilayer per-
ceptrons by over 20%. The ad-hoc feature extraction by the
CNN matched or outperformed models with the features
already extracted showing the clear advantages of using
deep learning to both extract features and classify data.
These new techniques to classify data create the possibility
to vastly increase the speed at which classifiers are trained
9
and improve the accuracy compared with traditional ma-
chine learning approaches often used with physiological
data such as Support Vector Machines. A combination of
a CNN and an RNN have previously been used to improve
the accuracy of detecting activity recognition by extracting
simple and complex activities from sensors directly[110].
This resulted in comparative performance to previous work
but was able to separate simple and complex activities
as well as use the raw data collected from the sensors.
Furthermore, deep learning has been used to recognise not
only the activity being performed but also the context by
combining multi-modal data such as audio and barometric
pressure [111]. Fusing the audio and text data required
learning both intra-modality and cross-modality resulting
in improved performance compared to previous general
solutions and outperformed task-specific sensor-tuned so-
lutions. The context in which data is collected has already
been shown to have a major impact in improving inference
and the ability to combine multi-model data and then run
the model on wearables or smartphones reaffirms the new
opportunities deep learning presents.
While tangible interfaces paired with machine learning
have shown the ability to infer mental well-being state in
limited trials, the new computational advancements dis-
cussed have demonstrated high accuracy when classify-
ing data and can be successfully ran from wearables and
smartphones providing opportunities to more accurately
detect mental well-being in real-time. Combining all these
data streams along with intelligent algorithms may greatly
advance the field of digital psychiatry and mental health.
2.4 Tools to promote positive behaviour change
How can behaviour changing tools be used to help improve
mental well-being?
2.4.1 Virtual reality and augmented reality
Numerous studies have shown Virtual Reality (VR) to help
improve many psychological disorders including PTSD and
anxiety by allowing patients to be exposed to stressful or
feared situations in a safe environment [112] [113]. When
using VR people are aware the situation is artificial allowing
them to temporarily suspend their disbelief and be more
confident in trying different approaches educating them of
the best approach to take in reality. VR has the potential to
help people overcome mental well-being challenges if high
levels of presence are achieved for situations that trouble
them and they are educated on new approaches to adopt.
A pilot study at the University of Oxford demonstrated
that virtual reality tools might reduce the delusional beliefs
that come with schizophrenia and severe paranoia [114].
Participants experienced a lift or train simulation with an
increasing number of people to increase the difficulty at a
manageable pace. One group of participants was told to
practice their normal defence behaviours such as avoiding
eye contact while the second group was told to drop their
defences to learn that the situation was safe by holding long
stares and standing close to the avatars. The second group
showed substantial reductions in their paranoid delusions
with over 50% no longer having severe paranoia for the
situation. The group who continued to practice their defence
behaviours also showed improvements as 20% of this group
no longer had severe paranoia. VR has the potential to allow
people to learn and reinforce new techniques and prac-
tices in a safe environment [115]. This goes beyond mobile
apps as it allows people to learn new approaches, helping
improve their mental well-being in real-world situations
although further research is needed to see if the benefits are
maintained beyond VR for more than the specific scenarios
trialled [116].
Augmented reality has the capability to assist people in
the real world by using glasses to overlay digital informa-
tion over the real world allowing people to get feedback
and learn in real-time rather than learning new approaches
beforehand such as with VR. Autism lends itself to AR
as it can often lead to mental well-being challenges such
as stress and anxiety as people with autism often fail to
recognise basic facial emotions, which make social inter-
actions and developing friendships challenging to sustain.
Researchers at Stanford University [117] are exploring the
use of augmented reality glasses to help children with
autism understand emotions. Machine learning classifies
camera data in real-time to infer and inform the wearer
of the nearby people’s emotions. These augmented reality
glasses could greatly help children with autism reduce the
daily stress they experience although the machine learning
classifier must be improved to recognise faces other than
those it has been trained on if it is to be used by the wider
population.
There are numerous challenges facing the mainstream
use of VR as mental well-being treatment including the lack
of training with only 17% of surveyed licensed psycholo-
gists trained to use VR and 38%–46% of those not using
VR exposure therapy [118]. If bad virtual reality exposure
therapy is received, it may negatively impact the patient’s
mental well-being making training more psychologists vi-
tal to VR’s success [119]. Additionally, some patients may
prefer traditional tried and tested methods rather than new
technological alternatives further reducing VR’s use [120].
Previous research shows VR trials often have a high success
rate but there are two main factors contributing to this;
the age of the participants as younger participants result
in higher success rates possibly because of their greater
exposure and consumption of technology and the quality
of randomised trials as low-quality studies show increased
differences. This shows that to improve VR’s success in
improving mental well-being more representative samples
and high quality randomised trials are required to ensure
results generalise well in new settings and more psycholo-
gists should be trained to use VR exposure therapy.
Virtual reality is now affordable with the tools and
technologies required already developed yet its potential to
educate people on different coping skills to use in stressful
situations has not been fully realised. A potentially contro-
versial topic which raises some concerns is that the recent
appearance of VR app stores will allow for VR software to
be released without being clinically evaluated, similar to
the majority of mental health mobile apps that have been
released, this issue should be addressed before VR software
to assist mental well-being becomes mainstream [68]. While
VR may not be able to replace other technologies and tools
used to monitor mental well-being it should be further
10
utilised as a behaviour changing tool to educate people
suffering from mental health conditions.
2.5 Biofeedback interventions
2.5.1 Biofeedback Therapy
Is it possible to teach people using technology how to
improve their mental well-being?
One method to improve mental well-being is biofeed-
back therapy; this involves monitoring a normal automatic
bodily function and then training people to acquire volun-
tary control of that function. Biofeedback is often completed
in a lab but the techniques learned can be applied to any
real-world situation. Nolan et al. [121] measured HRV in
patients with coronary heart disease as cardiac death is
more likely in these patients when stressed. The study
involved 46 patients half of which undertook HRV biofeed-
back involving training patients in paced breathing in order
to improve their HRV and stress management. The study
resulted in patients showing reduced symptoms of psycho-
logical stress and depression proving the positive effect of
biofeedback training and controlled breathing. However, all
evaluation was completed in an artificial lab setting where
the biofeedback training had been received. It is not possible
to know whether the techniques learned could be applied in
real world situations or whether more training is required
outside the lab.
Another study [122] used biofeedback for general stress
management; this biofeedback used a game to encourage
users to improve their heart rate and cerebral blood flow
control. This study used stress focused questionnaires, a
stress marker and a voxel-based morphometric analysis to
determine stress allowing the study to conclude that the
biofeedback helped reduce daily stress due to the increase
in regional grey matter. HRV biofeedback has also been used
during the postpartum period after the birth of a child. The
study [123] showed the biofeedback helped improve HRV
and improve sleep over the 1 month period it was used by
25 mothers. However, the lack of a control group means the
study does not definitively show the improvements were
due to the biofeedback training.
Biofeedback has been shown to have a significant im-
pact in reducing stress during trials although its effective-
ness in real-world stressful situations has not been proven
[124]. The possibility of pairing biofeedback training with
VR would allow users to practice the techniques learned
through biofeedback to reduce stress in a setting they
find stressful which would demonstrate the effectiveness of
biofeedback. Furthermore, biofeedback requires people to
have an understanding, willingness and time to train their
body to acquire voluntary control which many people do
not possess. Tangible interfaces may solve many of these
problems by using sensors to analyse mental state similar to
biofeedback but additionally provide feedback to improve
mental well-being in real-time.
2.5.2 Real-time tangible feedback interfaces
Can a combination of sensing and feedback technologies be
used to improve mental well-being in real-time?
An area of application still in its infancy is technologies
that go beyond sensing to additionally provide feedback
helping to improve mental well-being. Devices that sense
and provide feedback ranging from tangible interfaces to
robotics have the possibility to positively impact the broader
population who may temporarily experience mental well-
being challenges but do not seek professional help. Re-
searchers have developed tangible devices that actively aim
to improve mental well-being, these are often paired with
sensors and real-world feedback [125] to be automatically
provided when required.
A variety of tangible mental well-being devices have
been produced by Vaucelle, Bonanni, and Ishii [126], these
include touch me which contains multiple vibrotactile motors
to provide the sensation of touch, squeeze me; a vest to
simulate therapeutic holding, hurt me; a wearable device
that applies a moderated painful stimuli to ground people’s
senses and cool me down a device that heats up to ground
people’s senses. From the devices developed clinicians be-
lieved hurt me had the most potential as it could allow for
the patient and therapist to better relate to one another, by
having the therapist working with the class of pain the
patient is experiencing psychologically and externalising
viscerally. All of these interfaces have specific purposes such
as hurt me which may be beneficial for people considering
self-harming but not for people suffering from other mental
health challenges, a more general mental well-being device
is required for people who may experience temporary men-
tal well-being challenges.
It is possible to help improve general mental well-being
using small devices with real-time intervention; one such
device is Squeeze, Rock and Roll [127]. This device allowed
users to simulate rolling behaviours as many people do with
a pen when stressed but the device gradually guides the
user to reduce their movements and their stress through
dynamic tactile feedback. However, while people acknowl-
edged the device helped them relax no stress reduction
was found possibly because the device offered very little
feedback. Guiding users behaviours is a novel approach to
improve mental well-being although possibly less effective
as some people may find the action of rolling or twisting
objects relaxing by providing a distraction which can result
in mood improvements [128] and is often used as a coping
strategy for people suffering from mental health conditions
[129].
Haptic feedback is a method of providing feedback that
recreates the sense of touch through the use of motors and
vibrations; this allows people to experience real sensations
which can significantly affect emotional well-being and has
been shown to successfully improve mental well-being [130]
[131] [132] [133]. Good vibes [134] used a haptic sleeve to
provide varying feedback dependent on heart rate readings.
A stress test was conducted while the sleeve used dynamic
vibrations to help reduce the heart rates of the participants
by 4.34% and 8.31% in the two tests compared to the control
group. Doppel [135] also used haptic feedback in a wearable
device that aimed to reduce stress before public speaking
measuring users’ heart rates and skin conductance to de-
termine stress. The speed of the vibration was dependant
on the user’s heart rate providing personalised real-time
feedback. When users were told they were to present a
speech the skin conductance data showed users wearing the
Doppel remained less stressed than the control group. This
11
research shows that haptic feedback can have a substantial
positive impact in improving mental well-being and is more
successful than guiding user interactions. The advantage of
personalised haptic feedback is clear, but more research needs to
be conducted to establish the best rate of feedback for individual
users.
An alternative to haptic feedback uses deep breathing to
improve mental well-being. BioFidget [136] is a selfcontained
device that uses a heart rate monitor to detect HRV and allows
users to train their breathing by blowing on the fidget spinner to
reduce stress. Twenty participants stated BioFidget helped them
feel relaxed and overall it helped the majority of users improve
their HRV showing they were less stressed due to the positive
effect of deep breathing practices.
A headband has also been developed that uses EEG combined
with machine learning to assess stress by analysing alpha and beta
waves as alpha waves decrease when stressed [137] and then uses
two low powered massage motors to reduce stress using massage
therapy to provide significant reductions in physiological stress
[138]. The massage motors were tested on 4 participants with 3
of these responding well to the feedback and becoming less
stressed showing the possibility for massage therapy to be further
utilised in stress reduction devices. However, as the device was
only used by 4 participants with a 75% success rate, much more
research will need to be conducted to prove it can be as effective
as haptic feedback.
A different approach to provide real-time feedback is to alert
the user of their current mental state allowing them to take
appropriate measures such as reducing workload or taking time to
relax. MoodWings [139] aimed to reduce stress through wing
actuations informing users of their current stress levels.
Participants wore the device on their arm while ECG and
Electrodermal activity (EDA) readings were taken to determine
stress. A simulated driving experience was undertaken by
participants and once stress was detected the wing movement
was manually activated. The results show that MoodWings
improve the participants’ awareness of their stress but their
awareness further increased their stress as shown by EDA data
resulting in the device having a negative effect on users’ mental
well-being due to its alerting nature. Overall this study
demonstrated that sharing data with users needs to be carefully
considered
[139].
Communicating with others has a positive mental impact
leading to research that remotely connects people through
biofeedback. Shared breathing experiences through Breeze using
tactile, visual and audio feedback helped to increase the feeling of
belonging between connected participants [140]. EmoEcho [141]
similarly allowed users to share motion, touch and pulse through
haptic feedback with trusted partners to create a remote tangible
connection with the aim of improving mental well-being. Stress
levels have also been inferred through personal encounters
measured using Bluetooth although measuring encounters was
shown to not predict stress as accurately as physiological sensors
[142]. Communication with others is vital to positive mental well-
being and while feedback devices that aim to remotely connect
individuals appear to improve mental well-being in the current
limited trials.
Table ?? summarises the different feedback devices that aim
to both detect mental well-being and help improve mental well-
being. Some devices reviewed require manual feedback activation
and are not portable preventing them from being used outside of
the experiment and were not sufficiently validated such as
MoodWings and the headband. Other devices such as Doppel and
BioFidget show high potential for portable devices to both
automatically detect mental state and provide relevant feedback
although neither utilise machine learning to improve mental
wellbeing classification.
TABLE 1
Summary of Tangible Feedback Devices
Device
Sensors
Features
Squeeze
rock
a
nd
roll
Force,
movement
Dynamic tactile
feedback
MoodWings
EKG,
EDA,
GSM
Moving wings
Good
vibes
HR
Vibrotactile
feedback
Doppel
HRV, skin
conductance
Vibrotactile
feedback
BioFidget
HRV
Deep
breat
hing
Headband
EEG
Massage mo-
tors
A novel approach to provide feedback is through the use of
robotics such as therapy animals which are most commonly used
to reduce loneliness. One example of a robot used for therapy is
Paro; a robotic seal that was designed as an easy to use robotic
animal that encourages user interaction with its large eyes and
soft fur [143]. Tactile sensors allow Paro to understand the
location and force of users’ touch allowing for the response’s
magnitude to be relevant to the input. Studies show Paro provided
extremely effective therapy as it helped reduce stress in a day
service centre for elderly adults [144], increased user’s social
interactions and improved their reactions to stress in a care home
[143]. Paro has been shown to have a great impact in helping
reduce stress in elderly adults even with its limited sensors and
responses and has the potential to have a wider positive impact
on people’s mental well-being.
Although most therapy animals such as Paro target the elderly,
a robotic teddy aimed at reducing stress in young children at the
hospital has been developed [145]. Rather than relying upon
tactile interaction like Paro, this teddy uses vocal interactions
which children preferred. The children who used the robotic teddy
spent more time playing with it than the comparative virtual or
traditional plush teddy, they also had more meaningful
interactions and their behaviours conveyed they were emotionally
attached to the bear and not stressed. Robotic interactions can
12
automatically detect mental well-being in real-time allowing
for more personalised responses to be produced which may
have a significant positive impact on mental well-being.
Overall a variety of technologies that both sense mental
well-being and provide real-time feedback have been devel-
oped using a variety of different approaches. The feedback
incorporated in a device requires careful consideration and
evaluation to ensure it is effective in improving mental well-
being with machine learning being utilised to accurately
determine when feedback should be provided.
3REFLECTION AND CHALLENGES OF MENTAL
HEALTH TECHNOLOGIES
This section reflects on the finding of the survey; evaluating
the opportunities for mental well-being technologies along
with the challenges faced.
3.1 Discussion of existing research
A number of systems to support mental well-being using
apps, sensors, tangible interfaces, robotics and biofeedback
have been reviewed. A large number of mental well-being
apps already exist providing a range of features and func-
tionality with many existing apps aiming to improve tra-
ditional self-reporting tools and experience sampling. Apps
designed to elicit PROMs provide additional convenience
over traditional methods as they can be used anywhere
discreetly but self reporting is subjective and people may
fail to report [7] or be less truthful [25] when recording their
mental state, showing the benefits of appropriately using
objective measurements from sensors even if they are more
obtrusive. Recent developments in mHealth apps utilise
sensors within smartphones and wearable devices to mea-
sure physiological activity allowing mental well-being to be
automatically inferred. Currently, this is limited due to the
small number of sensors incorporated into such devices but
presents a much larger opportunity for continuous mobile
mental well-being monitoring [146] [147]. Mobile apps reaf-
firm the increasing popularity of people wishing to monitor
and improve their mental well-being using technological
alternatives to traditional techniques. However, currently
most mental well-being apps published in the Google Play
store and Apple app store have not been evaluated possibly
resulting in these apps having unforeseen consequences.
Sensing devices are also increasing in popularity with
advancements in physiological and environmental sensors
resulting in cheaper and smaller devices promoting exten-
sive use. A range of psychological sensors have been used
to detect mental well-being including pulse, HRV, GSR and
skin temperature, pairing these with environmental sensors
including accelerometer, gyroscope and magnetometer for
motion and force sensitive resistors to detect touch enables a
wide range of data to be collected to train machine learning
models. The ability to pair machine learning classifiers with
sensors presents an enormous opportunity allowing for
mental well-being to be detected with high accuracies of
over 90% [91] [100]. Pairing sensors with machine learning
in a portable interface enables well-being to be continuously
monitored without the need to continuously self-report as
deep learning models are able to infer mental well-being
from the raw data collected. While artificial intelligence
is extremely beneficial in classifying mental state it does
present its own set of challenges as a large amount of
labelled data is required to train the model accurately and
can struggle with predicting future outcomes related to
mental illness [148].
Feedback devices aim to advance upon sensing devices
by actively improving mental well-being in real-time using
varying feedback mechanisms including haptic, visual and
auditory [149]. Haptic feedback has been used in multiple
devices and often resulted in improved mental well-being
especially when the feedback was personalised. Machine
learning could again be utilised to control the feedback
based on the mental state inferred; this would allow for
automated personalised feedback to be provided in real-
time to improve mental well-being. Other feedback inter-
faces aimed to reduce stress using existing techniques such
as deep breathing [136] [140] or massage therapy [150];
all these techniques proved to be beneficial in improving
mental well-being displaying the need for more widespread
adoption of such devices. While some feedback devices
incorporated sensors to monitor the impact the feedback
had, very little research has been conducted pairing physio-
logical sensors, feedback mechanisms and machine learning
into devices that aim to both sense and improve mental
well-being in real-time. The effectiveness of the tangible
interfaces reviewed drastically varied in mostly small-scale
trials, or in some cases no current evaluation showing more
evaluation especially real-world trials are required.
3.2 Challenges
Applying therapies that work well when delivered in per-
son and translating them into digital or mobile versions is
not straightforward. There are many challenges associated
with mental well-being technologies, some specific issues re-
lated to tangible interfaces and mobile apps and additional
general challenges.
Fig. 6: Challenges associated with developing mental well-
being devices
13
Privacy is a significant issue as the majority of users want
to keep their mental health information private [72]. Users
are more cautious regarding sharing their health data mak-
ing integrating the data with established e-health systems
challenging [151]. Ideally data processing should be com-
pleted locally although on-device inference is only currently
feasible for very limited applications [152]. Furthermore,
care needs to be exercised regarding users’ privacy with the
data collected; ethical guidelines should be abided and users
should be made aware of the data being collected and how
it is being processed.
Given the stigma associated with mental illness, security
has to be a high priority for anyone thinking of developing
or using mental well-being tools. Concerns about how apps
respect privacy and use patient data remain rife, with many
mental well-being apps still lacking even basic privacy
policies or covertly selling users’ mental health information
to data brokers. Efforts such as the General Data Protection
Regulation (GDPR) in the EU and EEA have attempted to
give control to citizens over their personal data by ensuring
they are able to access their data and understand how
it is being processed [153]. Additionally, the EU Medical
Device Regulation (MDR) [154] will require all digital health
technologies to pass a conformity assessment and meet
safety and performance requirements by 2020. GDPR and
MDR may help regain people’s trust in mental well-being
applications and interfaces as it allows people to understand
how their personal data is stored and that the devices have
been assessed but additional steps should also be taken to
minimise any security flaws.
An issue with some of the discussed devices is users’
digital competence as elderly adults generally lack a high
level of digital skills which may be required to operate
these devices. One study [155] found elderly users preferred
wearable devices over mobile phones to report emotions.
However, Emoball [72] was a self-contained device rather
than a wearable and there was no evidence of digital com-
petence affecting user interactions showing devices to aid
mental well-being can be widely adopted.
User adherence and engagement is another crucial prob-
lem for mental well-being devices as users may not im-
mediately see the benefits of such solutions, preventing
continued use. Making the devices as small and portable
as possible should encourage engagement as it allows them
to be used anywhere users may experience poor mental
well-being [74]. The design of the devices must also be
carefully considered if it is going to be used in public as
it must be aesthetically pleasing to ensure the promotion of
continuous engagement [156]. However, there should also
be considerable debate around how much engagement is
necessary to best serve individual’s particular needs.
Recruiting users to test and use such devices is another
key challenge as it can be challenging to recruit users
willing to trial new technologies especially when it might
impact their mental well-being. Users will be required to
trial devices to ensure their effectiveness but also to collect
data enabling machine learning models to be trained. A vast
amount of data is required to train machine learning models
meaning a large number of users will be required to use new
devices to collect a sufficiently large dataset to accurately
train the model.
An issue with much of the existing research is the lack of
control groups and small sample sizes when trialling well-
being technologies. Little research contains enough scientific
analysis with sufficient statistical power to confirm its effec-
tiveness with many trials using fewer than 15 participants.
Furthermore, very few trials collect or test using real-world
data as people becoming artificially stressed in trials may
not exhibit the same patterns when stressed or suffer from
other mental well-being challenges in real-world situations.
One of the largest technical challenges when using deep
learning to classify data from sensors in real-time is that it
must first be trained with large amounts of labelled data
with issues arising if the model is trained with mislabelled
or insufficient data. Mental well-being can widely vary
depending on people’s characteristics making it essential to
collect data from a large and diverse group of people. On
the diagnostic side, one of the biggest issues is mental state
sensing as sensing people’s mental state is a challenging
proposition due to it often being subjective and it may be
difficult to infer through sensor data alone [157]. Machine
learning models could be trained on an individual basis
to allow for subjectivity to be taken into account but this
would initially require a vast amount of time and data to be
collected from each user before the device could accurately
infer well-being which may not be possible if an off the
shelf devices is to be developed. Furthermore, the ability to
provide personalised feedback may also require the model
to be trained on an individual basis to ensure the most
effective feedback for each user is provided. Data collection
poses one of the greatest challenges to tangible devices
utilising machine learning to sense and improve mental
well-being due to the vast amount of data required and the
subjectivity of mental well-being. As deep learning models
require thousands of samples to be sufficiently trained it is
difficult to develop a robust deep learning approach for the
classification of mental well-being without first developing
more accessible data collection tools.
Furthermore, traditional machine learning and feature
engineering algorithms may not be sufficiently efficient
enough to extract the complex and non-linear patterns gen-
erally observed in time series datasets such as those from
sensory data. Deep learning can help resolve this issue as the
use of a CNN and RNN combined has shown that features
can be extracted and classified automatically with LSTM
being fundamental in distinguishing time series data.
Sensing mental well-being not only requires accurate
machine learning models but also accurate sensors as if the
data recorded from the sensors is not reliable the classifica-
tion from the machine learning model will not be accurate.
However, when machine learning classifiers were paired
with off the shelf sensors stress was detected with similar
accuracy to clinical grade sensors that are expensive and
custom-made [158]. Tangible interfaces require sensors to
be cheap and as small as possible to reduce costs and the
overall footprint of the device demonstrating the need for
small yet accurate sensors and machine learning.
Assuming patients are willing to use instruments used in
the domain of assessing mental well-being, the underlying
issue of battery life still needs to be addressed. Often IoT
devices need to remain small and contain the necessary mi-
crocontroller and sensors leaving little room for the battery
14
meaning it will need to be recharged regularly. A possible
solution to this would be to only enable specific sensors after
other actions have been performed, this means high pow-
ered sensors will not have to be continually powered but
an additional step is required to collect data. Until batteries
with considerably longer battery life are developed, it will
remain impractical to continually collect vast amounts of
activity and behavioural data. Instead, pragmatic solutions
such as activating and deactivating sensors to reduce battery
consumption are necessary.
If tangible devices are to improve mental well-being
then they must also contain the relevant feedback. There
are many challenges to overcome when using sensors and
feedback actuators in tangible interfaces to improve mental
well-being. One issue is the size of the device as it must
contain sensors, a battery and feedback mechanisms such as
vibration motors for haptic feedback which make the device
large. There are new approaches to provide feedback includ-
ing Visio-Tactile feedback, that moves liquid metal drops in
real-time between electrodes allowing for the feedback to be
dynamic and smaller [159]. However, this is very early in
development and it may not yet be possible to incorporate
it into wearable devices.
Another general challenge is the business opportunity,
it will be critical to develop business models based on
responsible impact and socially-driven outcomes.. There
is the possibility of national health systems funding such
devices to ease the increasing pressure mental well-being
challenges have on health care but a lack of government
funding may prevent this. Alternatively, people suffering
with poor mental well-being could purchase the devices to
assist themselves independently if new technology enables
the devices to be made affordable while remaining accurate.
Overall there are many challenges to overcome when
developing tangible mental well-being devices ranging from
privacy issues to technological problems but new regula-
tions such as GDPR along with advancements in electronics
and machine learning should help reduce the difficulties
these challenges impose.
3.3 Opportunities
3.3.1 User feedback
The opportunities new technologies present to monitor and
improve mental well-being were explored during focus
groups at a school for students with severe, profound and
complex learning and physical disabilities in Nottingham,
UK. Mobile well-being apps were discussed although not
used by the participants due to their complexity as many
participants had fine and gross motor control issues making
touchscreens challenging to use, demonstrating the need to
develop tools to target specific sub-categories. Alternatives
to mobile apps such as tangible interfaces and virtual reality
show more potential for this user group as they are easier to
handle and operate.
Existing examples of mental well-being tangible inter-
faces were discussed to explore the opportunities they
present. Participants liked the portability of tangible de-
vices and the different ways they can be interacted with
compared with smartphones. Participants were excited by
the concept of devices being able to infer their mental well-
being as many had trouble recording their emotions. The
possibility for devices to improve mental well-being was
also intriguing as the participants had not used such de-
vices, demonstrating the requirement for tangible interfaces
to sense and improve mental well-being.
Wearable devices were considered to be useful as they
remove any requirement for fine motor control which some
participants did not possess. Different motor control levels
were examined in a separate group which showed some
participants’ inability to tightly grip objects while others had
difficulty relaxing their muscles. This demonstrates it may
not be possible to develop a single tangible device aimed at
all people suffering mental well-being challenges; separate
interfaces may need to be developed targeting different
groups of people.
Cost was a key factor discussed during the focus group
as the school and individuals would need the device to
be inexpensive if it was to become adopted into practice.
Durability was another issue raised as devices can often be
used in unintended ways which must be considered during
design and development. This focus group demonstrates the
need for a range of technological solutions to address mental
well-being issues as a one-size-fits-all solution could not
feasibly address all mental well-being issues for all potential
users especially those whose mental well-being issues are
often diagnostically overshadowed. The session concluded
that mental well-being tangible interfaces demonstrate the
most potential to both express feelings as well as actively
improve mental well-being but cost, durability and er-
gonomics need to be prioritised.
3.3.2 Advancements to enable real-time intervention
Recently there have been many developments in the tools
required to develop devices to sense and improve mental
well-being in real-time including the required microproces-
sors and sensors. Numerous System on Chip (SoC) devices
are now available that are capable of reading data from
sensors as well as processing the data in real-time. Micro-
controllers such as the Arduino platform are also available
but are incapable of complex data processing, however the
popularity of mobile phones allows for microcontrollers to
export the data to be processed externally.
New software is also available to program SoCs such
as Android things [160] which is a variation of the mobile
operating system Android. Android things allows for much
of the same functionality already available through other
software but provides additional access to libraries and
features that have already been developed for Android such
as TensorFlow lite to process data.
Additionally, advances in mobile phones and microcon-
trollers allow for machine learning to classify the data col-
lected from sensors locally. Many machine learning frame-
works have been developed to run on low powered devices
including TensorFlow lite which displayed high perfor-
mance in both single inference latency and CPU-optimized
continuous throughput when tested on Android phones
[144] [160]. It is now possible to run TensorFlow models on
smartphones and Raspberry Pis enabling tangible interfaces
powered by these devices to use deep learning to infer
mental well-being in real-time. These advancements allow
15
for small, portable, unobtrusive devices to be developed which
can utilise deep learning to improve people’s mental well-being in
real-time.
4 CONCLUSION
Different methods to sense and improve mental well-being have
been considered including apps, sensing devices, behaviour
changing tools and real-time intervention devices. Tangible
interfaces present a substantial opportunity for mental well-being
devices as they have the capability to both sense mental well-
being and provide interventional feedback. Sensors to detect well-
being can now be incorporated into small devices and advances in
deep learning allow for the raw data to be classified accurately on-
device allowing for real-time personalised feedback.
Given mental well-being challenges affect a large number of
people over the lifespan from young children to the elderly, it is
essential a range of tangible interfaces are developed to cater for
all users. Pairing toys for children with sensors and feedback will
make the interfaces more approachable and engaging as the toys
could react to the physiological or environmental data. Toys also
provide great opportunities to present feedback in engaging and
fun ways for children such as different coloured lights or haptic
feedback patterns to represent varying mental states or user
interactions. Alternatively, robotics can incorporate sensors and
feedback to create engaging devices for all ages, like toys robotics
can contain all of the required sensors and feedback mechanisms
in an approachable form promoting use.
One of the largest opportunities lies within personalising the
feedback tangible interfaces can provide. Personalising feedback
has multiple benefits including the possibility of quicker, better-
targeted treatment as the feedback provided can be continuously
adjusted automatically through machine learning models to
provide the most effective feedback on an individual basis.
Personalised feedback also removes the assumption many
existing tangible interfaces have made by creating a one size fits
all device as different people suffering from poor mental well-
being may prefer and respond better to different interventions.
There are numerous challenges associated with mental well-
being technologies such as the size of the device, data collection,
privacy and battery life but recent technological advancements
are increasing the opportunity and capability for small devices to
monitor and improve mental wellbeing. Wearable devices would
enable easier collection of physiological data. However, ensuring
all of the electronics are small enough to be contained within a
wrist-worn device may reduce battery life and increase costs
significantly.
Tangible user interfaces go beyond the capabilities that mobile
apps can offer, but have not yet been fully explored. There is
relatively little research conducted in the use of tangible devices
to both infer and improve mental wellbeing in real-time. Many
existing studies rely on small sample trials conducted over a short
period of time, making it challenging to evaluate their long-term
effectiveness. More rigorous studies need to be conducted to
provide robust evidence for the alleged capabilities tangible
interfaces possess to enable such technology to be modified,
scaled, and culturally adapted to serve the global population.
REFERENCES
[1] World Health Organisation, WHO Mental health: a state of well-
being, 2014. [Online]. Available:
https://www.who.int/features/factfiles/mental health/ en/.
[2] Perkbox, “THE 2O18 UK WORKPLACE STRESS
SURVEY,” 2018.
[3] The Physiological Society, “Stress in modern Britain Making Sense
of Stress,” 2017.
[4] B. Sheaves, K. Porcheret, A. Tsanas, C. A. Espie, D. Nuffield, R. G.
Foster, D. Freeman, P. J. Harrison DM, K. Wulff, and G. M. Goodwin,
“Insomnia, nightmares, and chronotype as markers of risk for severe
mental illness: results from a student population,” Sleep, vol. 39, no.
1, pp. 173181, 2016.
[5] I. Myin-Germeys, Z. Kasanova, T. Vaessen, H. Vachon, O. Kirtley,
W. Viechtbauer, and U. Reininghaus, “Experience sampling
methodology in mental health research: new insights and technical
developments,” World Psychiatry, 2018.
[6] I. Myin-Germeys, M. Oorschot, D. Collip, J. Lataster, P. Delespaul,
and J. Van Os, “Experience sampling research in psychopathology:
Opening the black box of daily life,” Psychological Medicine, 2009.
[7] S. Shiffman, A. A. Stone, and M. R. Hufford, “Ecological momentary
assessment.,” Annual review of clinical psychology, vol. 4, pp. 132,
2008.
[8] M. J. Hutchesson, M. E. Rollo, R. Callister, and C. E. Collins, “Self-
Monitoring of Dietary Intake by Young Women: Online Food
Records Completed on Computer or Smartphone Are as Accurate as
PaperBased Food Records but More Acceptable,” Journal of the
Academy of Nutrition and Dietetics, vol. 115, no. 1, pp. 8794, Jan.
2015.
[9] H.-G. Kim, E.-J. Cheon, D.-S. Bai, Y. H. Lee, and B.-H. Koo, “Stress
and Heart Rate Variability: A MetaAnalysis and Review of the
Literature.,” Psychiatry investigation, vol. 15, no. 3, pp. 235245, Mar.
2018.
[10] K. A. Herborn, J. L. Graves, P. Jerem, N. P. Evans, R. Nager, D. J.
McCafferty, and D. E. F. McKeegan, “Skin temperature reveals the
intensity of acute stress.,” Physiology & behavior, vol. 152, no. Pt A,
pp. 22530, Dec. 2015.
[11] R. Zangroniz,´ A. Mart´ınez-Rodrigo, J. M. Pastor, M. T. Lopez,´ and
A. Fernandez-Caballero,´ “Electrodermal Activity Sensor for
Classification of Calm/Distress Condition.,” Sensors (Basel,
Switzerland), vol. 17, no. 10, Oct. 2017.
[12] M. V. Villarejo, B. G. Zapirain, and A. M. Zorrilla, “A stress sensor
based on Galvanic Skin Response (GSR) controlled by ZigBee.,”
Sensors (Basel, Switzerland), vol. 12, no. 5, pp. 6075101, 2012.
[13] L. Al-barrak, E. Kanjo, and E. M. G. Younis, “NeuroPlace:
Categorizing urban places according to mental states,” PLOS ONE,
vol. 12, no. 9, B. Podobnik, Ed., Sep. 2017.
[14] N. El, M. Ieee, and E. Kanjo, “A Supermarket Stress Map,” vol. 13,
2013.
[15] A. A. Stone, J. S. Turkkan, C. A. Bachrach, J. B. Jobe,H. S. Kurtzman,
and V. S. Cain, Eds., The science of self-report: Implications for
research and practice. Mahwah, NJ, US: Lawrence Erlbaum
Associates Publishers, 2000, pp. xi, 380xi, 380.
[16] J. G. Gustavson and A. R., “The Science of SelfReport,” APS
Observer, vol. 10, no. 1, Jan. 1997.
[17] J. R. Weisz, I. N. Sandler, J. A. Durlak, and B. S. Anton, “Promoting
and Protecting Youth Mental
Health Through Evidence-Based Prevention and Treatment.,”
American Psychologist, vol. 60, no. 6, pp. 628648, 2005.
[18] I. McDowell, Measuring health : a guide to rating scales and
questionnaires. Oxford University Press, 2006, p. 748.
[19] A. J. Rush, M. H. Trivedi, H. M. Ibrahim, T. J. Carmody, B. Arnow,
D. N. Klein, J. C. Markowitz, P. T. Ninan, S. Kornstein, R. Manber,
16
M. E. Thase, J. H. Kocsis, and M. B. Keller, “The 16-Item Quick
Inventory of Depressive Symptomatology (QIDS), clinician rating
(QIDS-C), and self-report (QIDS-SR): a psychometric evaluation in
patients with chronic major depression.,” Biological psychiatry, vol.
54, no. 5, pp. 57383, Sep. 2003.
[20] A. J. Mitchell, M. Yadegarfar, J. Gill, and B. Stubbs, “Case finding
and screening clinical utility of the Patient Health Questionnaire
(PHQ-9 and PHQ-2) for depression in primary care: a diagnostic
metaanalysis of 40 studies,” BJPsych Open, vol. 2, no. 02, pp. 127
138, Mar. 2016.
[21] T. Gonzalez and C. Chiodo, “ICD 10,” Foot & Ankle
International, 2015.
[22] W. J. Earle, “DSM-5,” Philosophical Forum, 2014.
[23] I. Elkin, M. T. Shea, J. T. Watkins, S. D. Imber, S. M. Sotsky, J. F.
Collins, D. R. Glass, P. A. Pilkonis, W. R. Leber, J. P. Docherty, S. J.
Fiester, and M. B. Parloff, “National Institute of Mental Health
Treatment of Depression Collaborative Research Program,” Archives
of General Psychiatry, vol. 46, no. 11, p. 971, Nov. 1989.
[24] A. A. Stone, S. Shiffman, J. E. Schwartz, J. E. Broderick, and M. R.
Hufford, “Patient non-compliance with paper diaries.,” BMJ (Clinical
research ed.), vol. 324, no. 7347, pp. 11931194, May 2002.
[25] M. M. Graziose, “On the Accuracy of Self-Report Instruments for
Measuring Food Consumption in the School Setting,” Advances in
Nutrition, vol. 8, no. 4, pp. 635636, Jul. 2017.
[26] O. F. Wahl, “Mental Health Consumers’ Experience of Stigma,”
Schizophrenia Bulletin, vol. 25, no. 3, pp. 467478, Jan. 1999.
[27] S. Clement, O. Schauman, T. Graham, F. Maggioni, S. Evans-Lacko,
N. Bezborodovs, C. Morgan, N. Rusch, J. S. L. Brown, and G.
Thornicroft, “What is¨ the impact of mental health-related stigma on
helpseeking? A systematic review of quantitative and qualitative
studies,” Psychological Medicine, vol. 45, no. 01, pp. 1127, Jan.
2015.
[28] D. Swendeman, W. S. Comulada, N. Ramanathan, M. Lazar, and D.
Estrin, “Reliability and Validity of Daily Self-Monitoring by
Smartphone Application for Health-Related Quality-of-Life,
Antiretroviral Adherence, Substance Use, and Sexual Behaviors
Among People Living with HIV,” AIDS and Behavior, vol. 19, no. 2,
pp. 330340, Feb. 2015.
[29] J. Poushter and R. Stewart, “Smartphone Ownership and Internet
Usage Continues to Climb in Emerging Economies But advanced
economies still have higher rates of technology use,” Tech. Rep., 2016.
[30] E. Anthes, “Pocket psychiatry: mobile mental-health apps have
exploded onto the market, but few have been thoroughly tested,”
Nature, vol. 532, no. 7597, pp. 2024, Apr. 2016.
[31] P. M. Burgess, J. E. Pirkis, T. N. Slade, A. K. Johnston, G. N.
Meadows, and J. M. Gunn, “Service Use for Mental Health Problems:
Findings from the 2007 National Survey of Mental Health and
Wellbeing,” Australian & New Zealand Journal of Psychiatry, vol. 43,
no. 7, pp. 615623, Jul. 2009.
[32] S. G. Trusz, A. W. Wagner, J. Russo, J. Love, and D. F. Zatzick,
“Assessing Barriers to Care and Readiness for Cognitive Behavioral
Therapy in Early Acute Care PTSD Interventions,” Psychiatry:
Interpersonal and Biological Processes, vol. 74, no. 3, pp. 207223,
Sep. 2011.
[33] J. Proudfoot, G. Parker, D. Hadzi Pavlovic, V. Manicavasagar, E.
Adler, and A. Whitton, “Community attitudes to the appropriation of
mobile phones for monitoring and managing depression, anxiety, and
stress.,” Journal of medical Internet research, vol. 12, no. 5, e64, Dec.
2010.
[34] E. M. G. Younis, E. Kanjo, and A. Chamberlain, “Designing and
evaluating mobile self-reporting techniques: crowdsourcing for
citizen science,” Personal and Ubiquitous Computing, pp. 110, Mar.
2019.
[35] E. Kanjo, D. J. Kuss, and C. S. Ang, “NotiMind: Utilizing Responses
to Smart Phone Notifications as Affective Sensors,” IEEE Access, vol.
5, pp. 22 02322 035, 2017.
[36] J. Torous, R. Friedman, and M. Keshavan, “Smartphone ownership
and interest in mobile applications to monitor symptoms of mental
health conditions.,” JMIR mHealth and uHealth, vol. 2, no. 1, e2, Jan.
2014.
[37] A. Tsanas, K. Saunders, A. Bilderbeck, N. Palmius, M. Osipov, G.
Clifford, G. Goodwin, and M. De Vos, “Daily longitudinal self-
monitoring of mood variability in bipolar disorder and borderline
personality disorder,” Journal of Affective Disorders, vol. 205, pp.
225233, Nov. 2016.
[38] S. Abdul-Kade and J. Woods, “Survey on Chatbot Design Techniques
in Speech Conversation Systems,” International Journal of Advanced
Computer Science and Applications, vol. 6, no. 7, 2015.
[39] Roche, “THE NHS AT 100,” Tech. Rep., 2018. [Online]. Available:
http://roche-images.co.uk/design/ corporate/rxukcomm01930b(1)a(1)
nhs100 roche report.pdf.
[40] S. D’Alfonso, O. Santesteban-Echarri, S. Rice, G.Wadley,
R.Lederman, C. Miles, J. Gleeson, and M. Alvarez-Jimenez,
“Artificial Intelligence-Assisted Online Social Therapy for Youth
Mental Health,” Frontiers in Psychology, vol. 8, p. 796, Jun. 2017.
[41] Dongkeon Lee, Kyo-Joong Oh, and Ho-Jin Choi, “The chatbot feels
you - a counseling service using emotional response generation,” in
2017 IEEE International Conference on Big Data and Smart
Computing (BigComp), IEEE, Feb. 2017, pp. 437440.
[42] A. Schlesinger, K. P. O’Hara, and A. S. Taylor, “Let’s Talk About
Race,” in Proceedings of the 2018 CHI Conference on Human Factors
in Computing Systems CHI ’18, New York, New York, USA: ACM
Press,
2018, pp. 114.
[43] B. Whitby, “The Ethical Implications of Non-human Agency in
Health Care The Ethical Implications of Non-Human Agency in
Health Care Ethical Problems in System-Patient Interaction,”
Proceedings of MEMCA-14:(Machine ethics in the context of medical
and care agents), 2014.
[44] P. Balthazar, P. Harri, A. Prater, and N. M. Safdar, “Protecting Your
Patients’ Interests in the Era of Big Data, Artificial Intelligence, and
Predictive Analytics.,” Journal of the American College of
Radiology : JACR, vol. 15, no. 3 Pt B, pp. 580586, Mar. 2018.
[45] M. Bauer, T. Glenn, S. Monteith, R. Bauer, P. C. Whybrow, and J.
Geddes, “Ethical perspectives on recommending digital technology
for patients with mental illness,” International Journal of Bipolar
Disorders, vol. 5, no. 1, p. 6, Dec. 2017.
[46] K. Kretzschmar, H. Tyroll, G. Pavarini, A. Manzini, I. Singh, and N.
Y. P. A. Group, “Can Your Phone Be Your Therapist? Young
People’s Ethical Perspectives on the Use of Fully Automated
Conversational Agents (Chatbots) in Mental Health Support,”
Biomedical Informatics Insights, vol. 11, Jan. 2019.
[47] B. Inkster, S. Sarda, and V. Subramanian, “An Empathy-Driven,
Conversational Artificial Intelligence Agent (Wysa) for Digital
Mental WellBeing: Real-World Data Evaluation Mixed-Methods
Study.,” JMIR mHealth and uHealth, vol. 6, no. 11, Nov. 2018.
[48] K. K. Fitzpatrick, A. Darcy, and M. Vierhile, “Delivering Cognitive
Behavior Therapy to Young Adults With Symptoms of Depression
and Anxiety Using a
Fully Automated Conversational Agent (Woebot): A Randomized
Controlled Trial.,” JMIR mental health, vol. 4, no. 2, e19, Jun. 2017.
[49] D. Elmasri and A. Maeder, “A Conversational Agent for an Online
Mental Health Intervention,” in, Springer, Cham, Oct. 2016, pp. 243
251.
[50] Google Play, Wysa: stress therapy and mindfulness chatbot Apps on
Google Play, 2019. [Online]. Available:
https://play.google.com/store/apps/details?id= bot.touchkin.
[51] CNBC, Relaxation app Calm raises $88 million, valuing it $1 billion,
2019.
[52] Calm, press Calm Blog, 2019. [Online]. Available:
https://blog.calm.com/press.
[53] K. Cavanagh, A. Churchard, P. O’Hanlon, T. Mundy, P. Votolato, F.
Jones, J. Gu, and C. Strauss, “A Randomised Controlled Trial of a
Brief Online
17
Mindfulness-Based Intervention in a Non-clinical Population:
Replication and Extension,” Mindfulness, vol. 9, no. 4, pp. 11911205,
Aug. 2018.
[54] J. Boettcher, V. Astr˚ om, D. P¨ ahlsson, O. Schen-˚ strom, G.
Andersson, and P. Carlbring, “Internet-Based Mindfulness Treatment
for Anxiety Disorders: A Randomized Controlled Trial,” Behavior
Therapy, vol. 45, no. 2, pp. 241253, Mar. 2014.
[55] NHS, NHS Apps Library - NHS, 2019. [Online]. Available:
https://www.nhs.uk/apps-library/.
[56] M. Economides, J. Martman, M. J. Bell, and B.Sanderson,
“Improvements in Stress, Affect, and Irritability Following Brief Use
of a Mindfulness-based Smartphone App: A Randomized Controlled
Trial,” Mindfulness, vol. 9, no. 5, pp. 15841593, Oct. 2018.
[57] D. Lim, P. Condon, and D. DeSteno, “Mindfulness and Compassion:
An Examination of Mechanism and Scalability,” PLOS ONE, vol. 10,
no. 2, J. D. Creswell, Ed., e0118221, Feb. 2015.
[58] D. DeSteno, D. Lim, F. Duong, and P. Condon, “Meditation Inhibits
Aggressive Responses to Provocations,” Mindfulness, vol. 9, no. 4, pp.
11171122, Aug. 2018.
[59] I. H. Bennike, A. Wieghorst, and U. Kirk, “Onlinebased Mindfulness
Training Reduces Behavioral Markers of Mind Wandering,” 2017.
[60] C. Noone and M. J. Hogan, “A randomised activecontrolled trial to
examine the effects of an online mindfulness intervention on
executive control, critical thinking and key thinking dispositions in a
university student sample,” BMC Psychology, vol. 6, no. 1, p. 13, Dec.
2018.
[61] D. Villani, A. Grassi, C. Cognetta, D. Toniolo, P. Cipresso, and G.
Riva, “Self-help stress management training through mobile phones:
An experience with oncology nurses.,” Psychological Services, vol.
10, no. 3, pp. 315322, 2013.
[62] D. Villani, A. Grassi, C. Cognetta, P. Cipresso, D. Toniolo, and G.
Riva, “The effects of a mobile stress management protocol on nurses
working with cancer patients: a preliminary controlled study.,” in
MMVR, 2012, pp. 524528.
[63] A. Grassi, A. Gaggioli, and G. Riva, New technologies to manage
exam anxiety. 2011.
[64] D. Ben-Zeev, S. M. Kaiser, C. J. Brenner, M. Begale, J. Duffecy, and
D. C. Mohr, “Development and usability testing of FOCUS: a
smartphone system for self-management of schizophrenia.,”
Psychiatric rehabilitation journal, 2013.
[65] A. Ahtinen, E. Mattila, P. Valkkynen, K. Kaipainen,Vanhala, M.
Ermes, E. Sairanen, T. Myllymaki, and¨ R. Lappalainen, “Mobile
mental wellness training for stress management: feasibility and design
implications based on a one-month field study.,” JMIR mHealth and
uHealth, vol. 1, no. 2, e11, Jul. 2013.
[66] J. Marley and S. Farooq, “Mobile telephone apps in mental health
practice: uses, opportunities and challenges,” BJPsych Bulletin, vol.
39, no. 6, pp. 288290, Dec. 2015.
[67] S. L. Rizvi, L. A. Dimeff, J. Skutch, D. Carroll, and M. M. Linehan,
“A Pilot Study of the DBT Coach: An Interactive Mobile Phone
Application for Individuals With Borderline Personality Disorder and
Substance Use Disorder,” Behavior Therapy, vol. 42, no. 4, pp. 589
600, Dec. 2011.
[68] T. Donker, K. Petrie, J. Proudfoot, J. Clarke, M.-R. Birch, and H.
Christensen, “Smartphones for smarter delivery of mental health
programs: a systematic review.,” Journal of medical Internet research,
vol. 15, no. 11, e247, Nov. 2013.
[69] V. Harrison, J. Proudfoot, P. P. Wee, G. Parker, D. H. Pavlovic, and
V. Manicavasagar, “Mobile mental health: Review of the emerging
field and proof of concept study,” Journal of Mental Health, vol. 20,
no. 6, pp. 509524, Dec. 2011.
[70] M. Matthews and G. Doherty, “In the Mood: Engaging Teenagers in
Psychotherapy Using Mobile Phones,” Proceedings of the 2011
annual conference on Human factors in computing systems - CHI ’11,
2011.
[71] K. Niemantsverdriet and M. Versteeg, “Interactive Jewellery as
Memory Cue,” in Proceedings of the TEI ’16: Tenth International
Conference on Tangible, Embedded, and Embodied Interaction -
TEI ’16, New York, New York, USA: ACM Press, 2016, pp. 532538.
[72] J. Bravo, R. Hervas, and V. Villarreal, “Ambient intel-´ ligence for
health first international conference, AmIHEALTH 2015 Puerto
Varas, Chile, December 14, 2015 proceedings,” Lecture Notes in
Computer Science (including subseries Lecture Notes in Artificial
Intelligence and Lecture Notes in Bioinformatics), vol. 9456, pp. 189
200, 2015.
[73] A. T. Adams, E. L. Murnane, P. Adams, M. Elfenbein, P. F. Chang,
S. Sannon, G. Gay, and T. Choudhury,
“Keppi: A Tangible User Interface for Self-Reporting Pain,”
Proceedings of the 2018 CHI Conference on Human Factors in
Computing Systems - CHI ’18, 2018.
[74] F. Sarzotti, “Self-Monitoring of Emotions and Mood Using a Tangible
Approach,” Computers, vol. 7, no. 1, p. 7, Jan. 2018.
[75] M. Barker and LindenJanet, “A sprite in the dark: supporting
conventional mental healthcare practices with a tangible device,”
Proceedings of the Tenth International Conference on Tangible,
Embedded, and Embodied Interaction - TEI ’17, 2016.
[76] M. Balaam, G. Fitzpatrick, J. Good, and R. Luckin, “Exploring
Affective Technologies for the Classroom with the Subtle Stone,”
Proceedings of the 28th international conference on Human factors
in computing systems - CHI ’10, p. 1623, 2009.
[77] F. Guribye and T. Gjøsæter, “Tangible Interaction in the Dentist
Office,” in Proceedings of the Twelfth International Conference on
Tangible, Embedded, and Embodied Interaction - TEI ’18, New York,
New York, USA: ACM Press, 2018, pp. 123130.
[78] D. Beattie, SqueezeMusic- HCI & Audio Interaction Research, 2017.
[79] G. Doherty, D. Coyle, and M. Matthews, “Design and evaluation
guidelines for mental health technologies,” Interacting with
Computers, vol. 22, no. 4, pp. 243252, Jul. 2010.
[80] X. Zhang, W. Li, X. Chen, and S. Lu, “MoodExplorer: Towards
Compound Emotion Detection via Smartphone Sensing,” Proc. ACM
Interact. Mob. Wearable Ubiquitous Technol. Article, vol. 1, no. 176,
2017.
[81] X. Zhang, Y. Lyu, X. Luo, J. Zhang, C. Yu, H. Yin, and Y. Shi,
“Touch Sense,” Proceedings of the ACM on Interactive, Mobile,
Wearable and Ubiquitous Technologies, vol. 2, no. 2, pp. 118, 2018.
[82] A. Sano and R. W. Picard, “Stress Recognition using Wearable
Sensors and Mobile Phones,” 2013 Humaine Association Conference
on Affective Computing and Intelligent Interaction, 2013.
[83] R. Wang, W. Wang, A. DaSilva, J. F. Huckins, W. M. Kelley, T. F.
Heatherton, and A. T. Campbell, “Tracking Depression Dynamics in
College Students Using Mobile Phone and Wearable Sensing,”
Proceedings of the ACM on Interactive, Mobile, Wearable and
Ubiquitous Technologies, vol. 2, no. 1, pp. 126, 2018.
[84] K. Kroenke, R. L. Spitzer, J. B. Williams, and B. Lowe, “An Ultra-
Brief Screening Scale for Anxiety¨ and Depression: The PHQ4,”
Psychosomatics, vol. 50, no. 6, pp. 613621, Nov. 2009.
[85] N. Palmius, A. Tsanas, K. E. A. Saunders, A. C. Bilderbeck, J. R.
Geddes, G. M. Goodwin, and M. De Vos, “Detecting Bipolar
Depression From Geographic Location Data,” IEEE Transactions on
Biomedical Engineering, vol. 64, no. 8, pp. 17611771, Aug. 2017.
[86] T. Wallace, J. T. Morris, S. Bradshaw, and C. Bayer, “BreatheWell:
Developing a Stress Management App on Wearables for TBI &
PTSD,” Journal on Technology and Persons with Disabilities
Santiago, J, 2017.
[87] N. Sharma and T. Gedeon, “Objective measures, sensors and
computational techniques for stress recognition and classification: A
survey,” Computer Methods and Programs in Biomedicine, vol. 108,
no. 3, pp. 12871301, Dec. 2012.
[88] J. Schumm, E. Zurich, U. Ehlert, C. Setz, B. Arnrich, R. L. Marca, and
G. Troster, “Discriminating stress¨ from cognitive load using a
wearable EDA device. Discriminating Stress From Cognitive Load
Using a Wearable EDA Device,” IEEE Transactions on Information
Technology in Biomedicine, vol. 14, no. 2, 2010.
[89] M. Tanida, M. Katsuyama, and K. Sakatani, “Relation between mental
stress-induced prefrontal cortex activity and skin conditions: A near-
18
infrared spectroscopy study,” Brain Research, vol. 1184, pp. 210216,
Dec. 2007.
[90] Y. Choi and M. Kim, “Measurement of occupants’ stress based on
electroencephalograms (EEG) in twelve combined environments,”
Building and Environment, vol. 88, pp. 6572, Jun. 2015.
[91] R. Khosrowabadi, C. Quek, K. K. Ang, S. W. Tung, and M. Heijnen,
“A Brain-Computer Interface for classifying EEG correlates of
chronic mental stress,” in The 2011 International Joint Conference on
Neural Networks, IEEE, Jul. 2011, pp. 757762.
[92] J. Wijsman, B. Grundlehner, Hao Liu, H. Hermens, and J. Penders,
“Towards mental stress detection using wearable physiological
sensors,” in 2011 Annual International Conference of the IEEE
Engineering in Medicine and Biology Society, IEEE, Aug. 2011, pp.
17981801.
[93] U. Rajendra Acharya, K. Paul Joseph, N. Kannathal, C. M. Lim, and
J. S. Suri, “Heart rate variability: a review,” Medical & Biological
Engineering & Computing, vol. 44, no. 12, pp. 10311051, Dec. 2006.
[94] H.-S. Chiang, “ECG-based Mental Stress Assessment Using Fuzzy
Computing and Associative Petri Net,” Journal of Medical and
Biological Engineering, vol. 35, no. 6, pp. 833844, Dec. 2015.
[95] H. S. GOLDSTEIN and R. EDELBERG, “A plethysmographic
method for demonstrating the response specificity of the oral vascular
bed,” Psychophysiology, vol. 34, no. 1, pp. 124128, Jan. 1997.
[96] J. Hashimoto, K. Chonan, Y. Aoki, T. Nishimura,
Ohkubo, A. Hozawa, M. Suzuki, M. Matsubara,M. Michimata, T.
Araki, and Y. Imai, “Pulse wave velocity and the second derivative of
the finger photoplethysmogram in treated hypertensive patients: their
relationship and associating factors.,” Journal of hypertension, vol. 20,
no. 12, pp. 241522, Dec. 2002.
[97] D. McDuff, S. Gontarek, and R. Picard, “Remote measurement of
cognitive stress via heart rate variability,” in 2014 36th Annual
International Conference of the IEEE Engineering in Medicine and
Biology Society, vol. 2014, IEEE, Aug. 2014, pp. 29572960.
[98] M.-Z. Poh, D. J. McDuff, and R. W. Picard, “Advancements in
Noncontact, Multiparameter Physiological Measurements Using a
Webcam,” IEEE Transactions on Biomedical Engineering, vol. 58, no.
1, pp. 711, Jan. 2011.
[99] Y. Maeda, M. Sekine, and T. Tamura, “The Advantages of Wearable
Green Reflected Photoplethysmography,” Journal of Medical Systems,
vol. 35, no. 5, pp. 829834, Oct. 2011.
[100] J. Healey and R. Picard, “Detecting Stress During Real-World Driving
Tasks Using Physiological Sensors,” IEEE Transactions on
Intelligent Transportation Systems, vol. 6, no. 2, pp. 156166, Jun.
2005.
[101] S. Sun, M. J. Ball, and C.-h. Chen, “Investigating the Role of Context
in Perceived Stress Detection in the,” Proceedings of the ACM
International Joint Conference on Pervasive and Ubiquitous
Computing, no. July 2017, pp. 17081716, 2018.
[102] K. A. Herborn, J. L. Graves, P. Jerem, N. P. Evans, R. Nager, D. J.
McCafferty, and D. E. F. McKeegan, “Skin temperature reveals the
intensity of acute stress,” Physiology & Behavior, vol. 152, pp. 225
230, Dec. 2015.
[103] B. Kikhia, T. G. Stavropoulos, S. Andreadis, N. Karvonen, I.
Kompatsiaris, S. Savenstedt, M. Pijl, and C.¨ Melander, “Utilizing a
Wristband Sensor to Measure the Stress Level for People with
Dementia.,” Sensors (Basel, Switzerland), vol. 16, no. 12, Nov. 2016.
[104] F. Ordo´nez and D. Roggen, “Deep Convolutional and˜ LSTM
Recurrent Neural Networks for Multimodal Wearable Activity
Recognition,” Sensors, vol. 16, no. 1, p. 115, Jan. 2016.
[105] W. G. Hatcher and W. Yu, “A Survey of Deep Learning: Platforms,
Applications and Emerging Research Trends,” IEEE Access, vol. 6,
pp. 24 41124 432, 2018.
[106] A. Murad and J.-Y. Pyun, “Deep Recurrent Neural Networks for
Human Activity Recognition,” Sensors, vol. 17, no. 11, p. 2556, Nov.
2017.
[107] M. Zeng, L. T. Nguyen, B. Yu, O. J. Mengshoel, J. Zhu, P. Wu, and
J. Zhang, “Convolutional Neural Networks for Human Activity
Recognition using Mobile Sensors,” in Proceedings of the 6th
International Conference on Mobile Computing, Applications and
Services, ICST, 2014.
[108] E. Kanjo, E. M. G. Younis, and C. S. Ang, “Deep
Learning Analysis of Mobile Physiological, Environmental and
Location Sensor Data for Emotion Detection,” Journal of Information
Fusion, pp. 133, 2018.
[109] E. Kanjo, E. M. Younis, and N. Sherkat, “Towards unravelling the
relationship between on-body, environmental and emotion data using
sensor information fusion approach,” Information Fusion, vol. 40, pp.
1831, Mar. 2018.
[110] L. Peng, L. Chen, Z. Ye, and Y. Zhang, “AROMA: A Deep Multi-
Task Learning Based Simple and Complex Human Activity
Recognition Method Using Wearable Sensors,” Proc. ACM Interact.
Mob. Wearable Ubiquitous Technol. Proc. ACM Interact. Mob.
Wearable Ubiquitous Technol, vol. 2, no. 74, 2018.
[111] V. Radu, N. D. Lane, C. Tong, S. Bhattacharya, C. Mascolo, M. K.
Marina, and F. Kawsar, “Multimodal Deep Learning for Activity and
Context Recognition,” Proc. ACM Interact. Mob. Wearable
Ubiquitous Technol, vol. 1, no. 4, pp. 127157, 2017.
[112] D. Opris¸, S. Pintea, A. Garc´ıa-Palacios, C. Botella, S¸. Szamoskozi,
and D. David, “Virtual reality ex-¨ posure therapy in anxiety
disorders: a quantitative meta-analysis,” Depression and Anxiety, vol.
29, no. 2, pp. 8593, Feb. 2012.
[113] K. Meyerbroker and P. M. G. Emmelkamp, “Vir-¨ tual reality
exposure therapy in anxiety disorders: a systematic review of process-
and-outcome studies,” Depression and Anxiety, vol. 27, no. 10, pp.
933944, Aug. 2010.
[114] D. Freeman, J. Bradley, A. Antley, E. Bourke, N. DeWeever, N.
Evans, E. Cernis, B. Sheaves, F. Waite, G.ˇ Dunn, M. Slater, and D.
M. Clark, “Virtual reality in the treatment of persecutory delusions:
Randomised controlled experimental study testing how to reduce
delusional conviction,” British Journal of Psychiatry, vol. 209, no. 01,
pp. 6267, Jul. 2016.
[115] M. B. Powers and P. M. G. Emmelkamp, “Virtual reality exposure
therapy for anxiety disorders: A meta-analysis,” 2008.
[116] N. Morina, H. Ijntema, K. Meyerbroker, and P. M. G.¨ Emmelkamp,
“Can virtual reality exposure therapy gains be generalized to real-life?
A meta-analysis of studies applying behavioral assessments,”
Behaviour Research and Therapy, vol. 74, pp. 1824, Nov. 2015.
[117] P. Washington, C. Voss, N. Haber, S. Tanaka, J. Daniels, C. Feinstein,
T. Winograd, and D. Wall, “A
Wearable Social Interaction Aid for Children with Autism,”
Proceedings of the 2016 CHI Conference Extended Abstracts on
Human Factors in Computing Systems - CHI EA ’16, 2016.
[118] E. B. Foa, E. Hembree, and B. Rothbaum, Prolonged Exposure
Therapy for PTSD: Therapist Guide. Oxford University Press, Mar.
2007.
[119] M. Gerardi, J. Cukor, J. Difede, A. Rizzo, and B. O. Rothbaum,
“Virtual Reality Exposure Therapy for Post-Traumatic Stress
Disorder and Other Anxiety Disorders,” 2010.
[120] L. Gregg and N. Tarrier, “Virtual reality in mental health,” Social
Psychiatry and Psychiatric Epidemiology, vol. 42, no. 5, pp. 343354,
May 2007.
[121] R. P. Nolan, M. V. Kamath, J. S. Floras, J. Stanley, C. Pang, P. Picton,
and Q. R. Young, “Heart rate variability biofeedback as a behavioral
neurocardiac intervention to enhance vagal heart rate control,”
American Heart Journal, vol. 149, no. 6, pp. 11137, Jun. 2005.
[122] Y. Kotozaki, H. Takeuchi, A. Sekiguchi, Y. Yamamoto, T.Shinada, T.
Araki, K. Takahashi, Y. Taki, T. Ogino,M. Kiguchi, and R.
Kawashima, “Biofeedback-based training for stress management in
daily hassles: an intervention study.,” Brain and behavior, vol. 4, no.
4, pp. 566579, Jul. 2014.
[123] N. Kudo, @. Hitomi, S. @bullet, and H. Kodama, “Heart Rate
Variability Biofeedback Intervention for Reduction of Psychological
Stress During the Early Postpartum Period,” Applied
Psychophysiology and Biofeedback, vol. 39, no. 3-4, pp. 203211,
2014.
19
[124] A. L. Wheat and K. T. Larkin, “Biofeedback of Heart Rate Variability
and Related Physiology: A Critical Review,” Applied
Psychophysiology and Biofeedback, vol. 35, no. 3, pp. 229242, Sep.
2010.
[125] K. Woodward and E. Kanjo, “Things of the Internet (ToI),” in
Proceedings of the 2018 ACM International Joint Conference and
2018 International Symposium on Pervasive and Ubiquitous
Computing and Wearable Computers - UbiComp ’18, 2018, pp. 1228
1233.
[126] C. Vaucelle, L. Bonanni, and H. Ishii, “Design of haptic interfaces for
therapy,” in Proceedings of the 27th international conference on
Human factors in computing systems - CHI 09, New York, New York,
USA: ACM
Press, 2009, p. 467.
[127] M. Bruns, A. David, V. Keyson, and C. C. M. Hummels, “Squeeze,
Rock, and Roll; Can Tangible Interaction with Affective Products
Support Stress Reduction?” Proceedings of the 2nd international
conference on Tangible and embedded interaction - TEI ’08, 2008.
[128] J. Joormann, M. Siemer, and I. H. Gotlib, “Mood Regulation in
Depression: Differential Effects of Distraction and Recall of Happy
Memories on Sad Mood,” 2007.
[129] V. Carr, “Patients’ techniques for coping with schizophrenia: An
exploratory study,” British Journal of Medical Psychology, vol. 61,
no. 4, pp. 339352, Dec. 1988.
[130] C. L. Walters, “The Psychological and Physiological Effects of
Vibrotactile Stimulation, Via a Somatron, on Patients Awaiting
Scheduled Gynecological Surgery,” Journal of Music Therapy, vol.
33, no. 4, pp. 261287, Dec. 1996.
[131] J. Klamet, D. J. C. Matthies, and M. Minge, “WeaRelaxAble,” in
Proceedings of the 3rd International Workshop on Sensor-based
Activity Recognition and Interaction - iWOAR ’16, New York, New
York, USA: ACM
Press, 2016, pp. 16.
[132] B. Corbett, C. S. Nam, and T. Yamaguchi, “The Effects of Haptic
Feedback and Visual Distraction on Pointing Task Performance,”
International Journal of Human-Computer Interaction, vol. 32, no. 2,
pp. 89102, Feb. 2016.
[133] P. Paredes and M. Chan, “CalmMeNow,” in Proceedings of the 2011
annual conference extended abstracts on Human factors in computing
systems - CHI EA ’11, New York, New York, USA: ACM Press, 2011,
p. 1699.
[134] C. Kelling, D. Pitaro, and J. Rantala, “Good vibes,” in Proceedings of
the 20th International Academic Mindtrek Conference on -
AcademicMindtrek ’16, New York, New York, USA: ACM Press,
2016, pp. 130136.
[135] R. T. Azevedo, N. Bennett, A. Bilicki, J. Hooper, F. Markopoulou,
and M. Tsakiris, “The calming effect of a new wearable device during
the anticipation of public speech,” Scientific Reports, vol. 7, no. 1, p.
2285, 2017.
[136] R.-H. Liang, B. Yu, M. Xue, J. Hu, and L. M. G.
Feijs, “BioFidget: Biofeedback for Respiration Training Using an
Augmented Fidget Spinner,” Extended Abstracts of the 2018 CHI
Conference on Human Factors in Computing Systems - CHI ’18, 2018.
[137] S.-H. Seo and J.-T. Lee, “Stress and EEG,” in Convergence and
Hybrid Information Technologies, 2012.
[138] S. H. Cady and G. E. Jones, “Massage therapy as a workplace
intervention for reduction of stress.,” O Perceptual and Motor Skills,
vol. 84, pp. 157158, 1997.
[139] D.MacLean, A.Roseway, and M.Czerwinski, “MoodWings,” in
Proceedings of the 6th International Conference on PErvasive
Technologies Related to Assistive Environments - PETRA ’13, New
York, New York, USA: ACM Press, 2013, pp. 18.
[140] J. Frey and I. D. C. Herzliya, “Remote Biofeedback Sharing,
Opportunities and Challenges.,” Proceedings of the ACM
International Joint Conference on Pervasive and Ubiquitous
Computing, pp. 730733, 2018.
[141] K. Woodward, E. Kanjo, S. Burton, and A. Oikonomou, “EmoEcho :
A Tangible Interface to Convey and Communicate Emotions,”
Proceedings of the ACM International Joint Conference on Pervasive
and Ubiquitous Computing, 2018.
[142] C. Wu, “Vector Space Representation of Bluetooth Encounters for
Mental Health Inference,” Proceedings of the ACM International
Joint Conference on Pervasive and Ubiquitous Computing, pp. 1691
1699, 2018.
[143] T. Shibata and K. Wada, “Robot Therapy: A New Approach for
Mental Healthcare of the Elderly A Mini-Review,” Gerontology, vol.
57, no. 4, pp. 378386, 2011.
[144] B. D. Argall and A. G. Billard, A survey of Tactile HumanRobot
Interactions, 2010.
[145] S. Jeong, M. Zisook, L. Plummer, C. Breazeal, P. Weinstock, D. E.
Logan, M. S. Goodwin, S. Graca, B. O’Connell, H. Goodenough, L.
Anderson, N. Stenquist, and K. Fitzpatrick, “A Social Robot to
Mitigate Stress, Anxiety, and Pain in Hospital Pediatric Care,” in
Proceedings of the Tenth Annual ACM/IEEE International
Conference on Human-Robot Interaction Extended Abstracts -
HRI’15 Extended Abstracts, New York, New York, USA: ACM Press,
2015, pp. 103104.
[146] D. Ben-Zeev, E. A. Scherer, R. Wang, H. Xie, and A. T. Campbell,
“Next-generation psychiatric assessment: Using smartphone sensors
to monitor behaviour and mental health.,” Psychiatric Rehabilitation
Journal, vol. 38, no. 3, pp. 218226, 2015.
[147] A Pantelopoulos and N. Bourbakis, “A Survey on Wearable Sensor-
Based Systems for Health Monitoring and Prognosis, IEEE
Transactions on Systems, Man, and Cybernetics, Part C (Applications
and Reviews), vol. 40, no. 1, pp. 112, Jan. 2010. .
[148] B. E. Belsher, D. J. Smolenski, L. D. Pruitt, N. E. Bush, E. H. Beech,
D. E. Workman, R. L. Morgan, D. P. Evatt, J. Tucker, and N. A. Skopp,
“Prediction Models for Suicide Attempts and Deaths,” JAMA
Psychiatry, Mar. 2019.
[149] J. Frey and J. Cauchard, “Remote Biofeedback Sharing, Opportunities
and Challenges,” Proceedings of the ACM International Joint
Conference on Pervasive and Ubiquitous Computing: Adjunct, Aug.
2018.
[150] N. Chaitanya M., S. Jayakkumar, E. Chong, and C. Yeow, “A
wearable, EEG-based massage headband for anxiety alleviation,” in
2017 39th Annual International Conference of the IEEE Engineering
in Medicine and Biology Society (EMBC), vol. 2017, IEEE, Jul. 2017,
pp. 35573560.
[151] R. Williams, “Why is it difficult to achieve e-health systems at scale?”
Information, Communication & Society, vol. 19, no. 4, pp. 540550,
Apr. 2016.
[152] T. Guo, “Cloud-Based or On-Device: An Empirical Study of Mobile
Deep Inference,” in 2018 IEEE International Conference on Cloud
Engineering (IC2E), IEEE, Apr. 2018, pp. 184190.
[153] European Comission, Data protection European Commission,
2018.
[154] Gov.uk, Medical devices: EU regulations for MDR and IVDR -
GOV.UK, 2019. [Online]. Available:
https://www.gov.uk/guidance/medical-devices-euregulations-for-
mdr-and-ivdr.
[155] I. Rodr´ıguez, G. Cajamarca, V. Herskovic, C. Fuentes, and M.
Campos, “Helping Elderly Users Report Pain Levels: A Study of User
Experience with Mobile and Wearable Interfaces,” Mobile
Information Systems, vol. 2017, pp. 112, Nov. 2017.
[156] M. Pakanen, A. Colley, J. Hakkil¨ a, J. Kildal, and V.¨ Lantz,
“Squeezy bracelet,” in Proceedings of the 8th Nordic Conference on
Human-Computer Interaction Fun, Fast, Foundational -
NordiCHI ’14, New York, New York, USA: ACM Press, 2014, pp.
305314.
[157] F.-T. Sun, C. Kuo, H.-T. Cheng, S. Buthpitiya, P. Collins, and M.
Griss, “Activity-Aware Mental Stress Detection Using Physiological
Sensors,” in, Springer, Berlin, Heidelberg, Oct. 2012, pp. 211230.
[158] V. Mishra, “The Case for a Commodity Hardware Solution for Stress
Detection,” Proceedings of the ACM International Joint Conference
on Pervasive and Ubiquitous Computing: Adjunct, pp. 17171728,
2018.
20
[159] D. R. Sahoo, T. Neate, Y. Tokuda, J. Pearson, S. Robinson, S.
Subramanian, M. Jones, Deepak, R. Sahoo, and J. Pear-Son,
“Tangible Drops: A Visio-Tactile Display Using Actuated Liquid-
Metal Droplets,” Proceedings of the 2018 CHI Conference on Human
Factors in Computing Systems - CHI ’18, 2018.
[160] Google, Overview Android Things Android Developers, 2018.
[Online]. Available: https://developer. android.com/things/get-
started/.
ResearchGate has not been able to resolve any citations for this publication.
Conference Paper
Full-text available
In this work, we would like to discuss our findings obtained from the newly proposed hardware design for anxiety detection and its mitigation where a subject's state of mind is identified from the neurological data retrieved. A feedback-based network implemented for stress alleviation mainly comprises of a pair of massage motors and RGB light-emitting diode (LED) that activate during conditions of stress detected, a custom-made Electroencephalography (EEG)-sensor and a massage motor circuit both functioning on an Arduino driven platform. The skin electrodes facilitate a hassle-free retrieval of beta waves from the frontal areas that are transmitted wirelessly by a Bluetooth console to a computer post signal amplification and filtration. Rising amplitudes of beta signals that are associated to anxiety have been successfully tackled in three out of four subjects by suppressing the high values due to the massage motor therapy introduced. The motors on sensing high amplitude values exceeding the pre-set threshold limits during the three experiment trials rotate smoothly thus helping one to relax and guaranteeing a higher work performance.
Article
Full-text available
In recent years, mobile phone technology has taken tremendous leaps and bounds to enable all types of sensing applications and interaction methods, including mobile journaling and self-reporting to add metadata and to label sensor data streams. Mobile self-report techniques are used to record user ratings of their experiences during structured studies, instead of traditional paper-based surveys. These techniques can be timely and convenient when data are collected Bin the wild^. This paper proposes three new viable methods for mobile self-reporting projects and in real-life settings such as recording weather information or urban noise mapping. These techniques are Volume Buttons control, NFC-on-Body, and NFC-on-Wall. This work also provides an experimental and comparative analysis of various self-report techniques regarding user preferences and submission rates based on a series of user experiments. The statistical analysis of our data showed that pressing screen buttons and screen touch allowed for higher labelling rates, while Volume Buttons proved to be more valuable when users engaged in other activities, e.g. while walking. Similarly, based on participants' preferences, we found that NFC labelling was also an easy and intuitive technique when used in the context of self-reporting and place-tagging. Our hope is that by reviewing current self-reporting interfaces and user requirements, we will be able to enable new forms of self-reporting technologies that were not possible before.
Article
Full-text available
Over the last decade, there has been an explosion of digital interventions that aim to either supplement or replace face-to-face mental health services. More recently, a number of automated conversational agents have also been made available, which respond to users in ways that mirror a real-life interaction. What are the social and ethical concerns that arise from these advances? In this article, we discuss, from a young person’s perspective, the strengths and limitations of using chatbots in mental health support. We also outline what we consider to be minimum ethical standards for these platforms, including issues surrounding privacy and confidentiality, efficacy, and safety, and review three existing platforms (Woebot, Joy, and Wysa) according to our proposed framework. It is our hope that this article will stimulate ethical debate among app developers, practitioners, young people, and other stakeholders, and inspire ethically responsible practice in digital mental health.
Article
Full-text available
Background: A World Health Organization 2017 report stated that major depression affects almost 5% of the human population. Major depression is associated with impaired psychosocial functioning and reduced quality of life. Challenges such as shortage of mental health personnel, long waiting times, perceived stigma, and lower government spends pose barriers to the alleviation of mental health problems. Face-to-face psychotherapy alone provides only point-in-time support and cannot scale quickly enough to address this growing global public health challenge. Artificial intelligence (AI)-enabled, empathetic, and evidence-driven conversational mobile app technologies could play an active role in filling this gap by increasing adoption and enabling reach. Although such a technology can help manage these barriers, they should never replace time with a health care professional for more severe mental health problems. However, app technologies could act as a supplementary or intermediate support system. Mobile mental well-being apps need to uphold privacy and foster both short- and long-term positive outcomes. Objective: This study aimed to present a preliminary real-world data evaluation of the effectiveness and engagement levels of an AI-enabled, empathetic, text-based conversational mobile mental well-being app, Wysa, on users with self-reported symptoms of depression. Methods: In the study, a group of anonymous global users were observed who voluntarily installed the Wysa app, engaged in text-based messaging, and self-reported symptoms of depression using the Patient Health Questionnaire-9. On the basis of the extent of app usage on and between 2 consecutive screening time points, 2 distinct groups of users (high users and low users) emerged. The study used mixed-methods approach to evaluate the impact and engagement levels among these users. The quantitative analysis measured the app impact by comparing the average improvement in symptoms of depression between high and low users. The qualitative analysis measured the app engagement and experience by analyzing in-app user feedback and evaluated the performance of a machine learning classifier to detect user objections during conversations. Results: The average mood improvement (ie, difference in pre- and post-self-reported depression scores) between the groups (ie, high vs low users; n=108 and n=21, respectively) revealed that the high users group had significantly higher average improvement (mean 5.84 [SD 6.66]) compared with the low users group (mean 3.52 [SD 6.15]); Mann-Whitney P=.03 and with a moderate effect size of 0.63. Moreover, 67.7% of user-provided feedback responses found the app experience helpful and encouraging. Conclusions: The real-world data evaluation findings on the effectiveness and engagement levels of Wysa app on users with self-reported symptoms of depression show promise. However, further work is required to validate these initial findings in much larger samples and across longer periods.
Article
Full-text available
The detection and monitoring of emotions are important in various applications, e.g. to enable naturalistic and personalised human-robot interaction. Emotion detection often require modelling of various data inputs from multiple modalities, including physiological signals (e.g.EEG and GSR), environmental data (e.g. audio and weather), videos (e.g. for capturing facial expressions and gestures) and more recently motion and location data. Many traditional machine learning algorithms have been utilised to capture the diversity of multimodal data at the sensors and features levels for human emotion classification. While the feature engineering processes often embedded in these algorithms are beneficial for emotion modelling, they inherit some critical limitations which may hinder the development of reliable and accurate models. In this work, we adopt a deep learning approach for emotion classification through an iterative process by adding and removing large number of sensor signals from different modalities. Our dataset was collected in a real-world study from smart-phones and wearable devices. It merges local interaction of three sensor modalities: on-body, environmental and location into global model that represents signal dynamics along with the temporal relationships of each modality. Our ap proach employs a series of learning algorithms including a hybrid approach using Convolutional Neural Network and Long Short-term Memory Recurrent Neural Network (CNN-LSTM) on the raw sensor data, eliminating the needs for manual feature extraction and engineering. The results show that the adoption of deep-learning approaches is effective in human emotion classification when large number of sensors input is utilised (average accuracy 95% and F Measure=%95) and the hybrid models outperform traditional fully connected deep neural network (average accuracy 73% and F-Measure=73%). Further more, the hybrid models outperform previously developed Ensemble algorithms that utilise feature engineering to train the model average accuracy 83% and F-Measure=82%)
Conference Paper
Full-text available
Social interactions have multifaceted effects on individu-als' mental health statuses, including mood and stress. As a proxy for the social environment, Bluetooth encounters detected by personal mobile devices have been used to improve mental health prediction and have shown preliminary success. In this paper, we propose a vector space model representation of Bluetooth encounters in which we convert encounters into spatiotemporal tokens within a multidimen-sional feature space. We discuss multiple token designs and feature value schemes and evaluate the predictive power of the resulting features for stress recognition tasks using the StudentLife and Friends & Family datasets. Our findings motivate further discussion and research on bag-of-words approaches for representing raw mobile sensing signals for health outcome inference.
Conference Paper
Full-text available
An interactive tangible interface has been developed to capture and communicate emotions between people who are missing and longing for loved ones. EmoEcho measures the wearer’s pulse, touch and movement to provide varying vibration patterns on the partner device. During an informal evaluation of two prototype devices users acknowledged how EmoEcho could help counter the negative feeling of missing someone and liked the range of feedback offered. In general, we believe, tangible interfaces appear to offer a non-obtrusive means towards interpreting and communicating emotions to others.
Conference Paper
The advances in mobile and wearable sensing have led to a myriad of approaches for stress detection in both laboratory and free-living settings. Most of these methods, however, rely on the usage of some combination of physiological signals measured by the sensors to detect stress. While these solutions work great in a lab or a controlled environment, the performance in free-living situations leaves much to be desired. In this work, we explore the role of context of the user in free-living conditions, and how that affects users' perceived stress levels. To this end, we conducted an 'in-the-wild' study with 23 participants, where we collected physiological data from the users, along with 'high-level' contextual labels, and perceived stress levels. Our analysis shows that context plays a significant role in the users' perceived stress levels, and when used in conjunction with physiological signals leads to much higher stress detection results, as compared to relying on just physiological data.
Conference Paper
Timely detection of an individual's stress level has the potential to expedite and improve stress management, thereby reducing the risk of adverse health consequences that may arise due to unawareness or mismanagement of stress. Recent advances in wearable sensing have resulted in multiple approaches to detect and monitor stress with varying levels of accuracy. The most accurate methods, however, rely on clinical grade sensors strapped to the user. These sensors measure physiological signals of a person and are often bulky, custom-made, expensive, and/or in limited supply, hence limiting their large-scale adoption by researchers and the general public. In this paper, we explore the viability of commercially available off-the-shelf sensors for stress monitoring. The idea is to be able to use cheap, non-clinical sensors to capture physiological signals, and make inferences about the wearer's stress level based on that data. In this paper, we describe a system involving a popular off-the-shelf heart-rate monitor, the Polar H7; we evaluated our system in a lab setting with three well-validated stress-inducing stimuli with 26 participants. Our analysis shows that using the off-the-shelf sensor alone, we were able to detect stressful events with an F1 score of 0.81, on par with clinical-grade sensors.