Ijerph 18 06868

International Journal of
Environmental Research
and Public Health
Article
User Centered Virtual Coaching for Older Adults at Home
Using SMART Goal Plans and I-Change Model
Andoni Beristain Iraola 1,2,3, * and Roberto Álvarez Sánchez 1,2, *
1 Vicomtech Foundation, Basque Research and Technology Alliance (BRTA), Mikeletegi 57,
20009 San Sebastián, Spain
2 Biodonostia Health Research Institute, Paseo Dr Begiristain s/n, 20014 San Sebastián, Spain
3 Computational Intelligence Group, Computer Science Faculty, University of the Basque Country, UPV/EHU,
20018 San Sebastian, Spain
* Correspondence: [email protected] (A.B.I.); [email protected] (R.Á.S.);
Tel.: +34-943-309230 (A.B.I.)
Abstract: Preventive care and telemedicine are expected to play an important role in reducing the
impact of an increasingly aging global population while increasing the number of healthy years.
Virtual coaching is a promising research area to support this process. This paper presents a user-
centered virtual coach for older adults at home to promote active and healthy aging and independent
living. It supports behavior change processes for improving on cognitive, physical, social interaction
and nutrition areas using specific, measurable, achievable, relevant, and time-limited (SMART)
goal plans, following the I-Change behavioral change model. Older adults select and personalize
which goal plans to join from a catalog designed by domain experts. Intervention delivery adapts
to user preferences and minimizes intrusiveness in the user’s daily living using a combination of a

deterministic algorithm and incremental machine learning model. The home becomes an augmented
Citation: Beristain Iraola, A.;
reality environment, using a combination of projectors, cameras, microphones and support sensors,
Álvarez Sánchez, R. User Centered where common objects are used for projection and sensed. Older adults interact with this virtual
Virtual Coaching for Older Adults at coach in their home in a natural way using speech and body gestures on projected user interfaces
Home Using SMART Goal Plans and with common objects at home. This paper presents the concept from the older adult and the caregiver
I-Change Model. Int. J. Environ. Res. perspectives. Then, it focuses on the older adult view, describing the tools and processes available
Public Health 2021, 18, 6868. https:// to foster a positive behavior change process, including a discussion about the limitations of the
doi.org/10.3390/ijerph18136868 current implementation.
Academic Editor: Paul B. Tchounwou Keywords: virtual coaching; data visualization; incremental machine learning; feature engineering;
self-management; healthy and active aging
Received: 12 May 2021
Accepted: 22 June 2021
Published: 26 June 2021
1. Introduction
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
According to data from World Population Prospects: the 2019 Revision presented by
published maps and institutional affil- the United Nations [1], by 2050, one in six people in the world will be over age 65 (16%),
iations. up from one in 11 in 2019 (9%). By 2050, one in four persons living in Europe and Northern
America could be aged 65 or over. In 2018, for the first time in history, persons aged 65 or
above outnumbered children under five years of age globally. The number of persons aged
80 years or over is projected to triple, from 143 million in 2019 to 426 million in 2050. Despite
Copyright: © 2021 by the authors.
living longer, natural age-related decline combined with the frequent comorbidity [2] may
Licensee MDPI, Basel, Switzerland.
require that the older adults have some degree of support and supervision. Furthermore,
This article is an open access article
most older adults prefer living on their own instead of staying in a healthcare institution,
distributed under the terms and which does not guarantee better outcomes than at-home care [3]. This has fostered policies
conditions of the Creative Commons to promote active and healthy aging, trying to delay or even avoid dependency, and being
Attribution (CC BY) license (https:// able to enjoy this stage in life as a healthy individual.
creativecommons.org/licenses/by/ Coaching is a form of counseling to maximize personal potential, according to the
4.0/). ECVision funded by the European Commission [4], and the coaching strategy includes a
Int. J. Environ. Res. Public Health 2021, 18, 6868. https://doi.org/10.3390/ijerph18136868 https://www.mdpi.com/journal/ijerph
Int. J. Environ. Res. Public Health 2021, 18, 6868 2 of 24
plan of actions or guidelines. Optimal guidance is achievable when the coach is aware and
adaptable to the coached person’s context, goals and preferences. Preventive and online
care must deal with long-term user monitoring and support, maintaining user adherence.
The role of a coach is very important in this sense to motivate the user and set, track
and adapt training plans with specific goals in mind. This coach needs to have access to
detailed data about the user status, interests, behavior and goals to properly guide the user.
Unfortunately, in practice it is not feasible to have this kind of support for most people due
to limitations in the healthcare services.
As a solution to this limitation, virtual coaching tries to attain similar support to that of
a human coach. Siewiorek [5] defined a virtual coach system (VCS) as an always-attentive
personalized system that continuously monitors the user’s activity and surroundings, and
delivers interventions—e.g., intentional messages—when these are desired. Coaching in-
terventions for older people to adhere to healthier habits could prolong their independence
and relieve costs from healthcare systems.
Traditional coaching has been seen as a human-to-human relationship. Yet, this
concept is shifting as technology becomes part of our lives. The literature review conducted
by Wolever et al., 2013 [6] found that only 5% percent of the analyzed studies about
health coaching were completely technology based. Further, those traditional technology-
based behavior change systems were implemented as simple if-then-else algorithms like
decision trees, with limited potential for adaptation and self-learning from participants’
feedback. Early works such as Bacciu et al. (2015) [7] focus on user interface and Internet
of Things aspects for a VCS at home. This is also the case for the study by Gordienko et al.
(2017) [8], which also focused on the same aspects, but they are far from Siewiorek’s vision
about what a VCS constitutes. In a recent survey, Lete et al. (2020) [9] also found limited
technological development on virtual coaching for older adults. More recently, thanks to
several research projects funded in 2017 by the European Research and Innovation Action
(RIA) on virtual coaching for older adults, more technologically sophisticated systems
have started to emerge. This is the case of the EMPATHIC project [10], focused on natural
language speech interaction enhanced with emotional cues, where the coach continuously
adapts to previous interactions and provides support on several aspects of older adults’
well-being. Council of coaches [11] also focused on natural language interaction but
included the concept of embodied conversational social characters. Supporting Active
Ageing through Multimodal coaching (SAAM) [12] stresses the participation of the older
adult social circles in the coaching process and social activity to combat isolation. Virtual
Coaching Activities for Rehabilitation in Elderly (vCare) [13] considers tele-rehabilitation
and frailty aspects supported by gamification and advanced vital sign monitoring. The
Novel Empowering Solutions and Technologies for Older people to Retain Everyday life
activities (NESTORE) [14] project certainly underlines the importance of behavior change
models and carries a systematic review of virtual coaching for well-being (2020). This
review tried to analyze these areas: (i) the different coach definitions, (ii) the behavior
change models used, (iii) which domains are tackled and (iv) the technical implementation
of the system. It concludes saying that there is a trend for the convergence about what a
virtual coach for older adult well-being should be and which domains it should tackle.
Coaching implies fostering behavior change processes in the person, therefore it is
important to consider a behavior change model to guide this process. The Integrated
Change Model (I-Change Model) is a behavior change model derived from the attitude–
social influence–self-efficacy model [15] which states that covert and overt behaviors
are determined by a person’s motivation or intention to carry out a particular type of
action, originated by awareness factors (e.g., knowledge, risk perceptions and cues to
action). This sequence from awareness to motivation, action and behavior depends on the
person’s preceding factors (e.g., biological, psychological, previous behavior, social and
environmental) and information factors (e.g., message, channel, source). The I-Change
model has been successfully blended with technology to promote healthy behaviors in
several domains, such as smoking cessation [16], hypertension reduction [17] and healthy
in general [18], among others. Therefore, it seems well suited for the purpose
these behaviors under the umbrella of a VCS. This is in line with a pragmatic
to design digital health programs which was recently published, in which
tions are strongly guided by goals and objectives [19].
According to Turner-Stokes [20], every individual sets lifelong objec
considered goals that take a long time to achieve (i.e., years). These long-te
be splitininto
lifestyle smaller
general medium-term
[18], among achievements
others. Therefore, or intermediate
it seems well suited for the purposemileston
of
term
promoting objectives, whichunder
these behaviors permit measuring
the umbrella of a VCS. self-evolution.
This is in line withFinally, those sho
a pragmatic
tives can betotranslated
methodology into
design digital a plan
health withwhich
programs individual action
was recently items.inFollowing
published, which th
the interventions are strongly guided by goals and objectives [19].
termAccording
objectives become specific
to Turner-Stokes andindividual
[20], every the progress can beobjectives
sets lifelong measured. Goals m
that are
able for the
considered person,
goals action-oriented,
that take a long time to achieve realistic, relevant,
(i.e., years). and finally,
These long-term goals cantime-limit
be
tures inspire the acronym of a SMART (specific, measurable, achievable,
split into smaller medium-term achievements or intermediate milestones and short-term
objectives, which permit measuring self-evolution. Finally, those short-term objectives
time-limited)
can be translated goal. For with
into a plan example,
individuala SMART
action items. goal mightthis
Following be idea,
[21]:short-term
“Being able to
ter of mile
objectives to reach
become myand
specific local supermarket
the progress within six
can be measured. Goalsweeks.”
must beTolchin
achievable et al. [22
SMART goals for treatment planning in psychotherapy were more benefic
for the person, action-oriented, realistic, relevant, and finally, time-limited. These features
inspire the acronym of a SMART (specific, measurable, achievable, relevant and time-
alternatives.
limited) goal. For example, a SMART goal might be [21]: “Being able to walk a quarter of
mile to Inreach
thismycontext, H2020 within
local supermarket CAPTAIN is Tolchin
six weeks.” a research and
et al. [22] innovation
proved that SMARTaction
pursues the creation
goals for treatment planning inofpsychotherapy
a virtual coachwere more system
beneficialto support
than older adults
other alternatives.
In this context, H2020 CAPTAIN is a research and innovation action project which
pendently at home [23,24]. This VCS proposes a digitalization of the I-Chan
pursues the creation of a virtual coach system to support older adults living independently
will
at home guide older
[23,24]. adults
This VCS towards
proposes the successful
a digitalization achievement
of the I-Change of SMART
model that will guide go
their lifelong objectives (Figure 1), circumscribing them into four main dim
older adults towards the successful achievement of SMART goals as part of their lifelong
objectives (Figure 1), circumscribing them into four main dimensions: cognitive, social,
nitive, social, physical and nutritional.
physical and nutritional.
Figure
Figure 1.1. I-Change
I-Change and SMART
and SMART coaching.coaching.
This paper presents the coaching aspects in CAPTAIN from different perspectives
Thisa top-down
following paper presents
approach.the coaching
Section aspects
2 describes in virtual
the overall CAPTAINcoachingfrom different
concept
following a top-down approach. Section 2 describes the overall virtual coac
and ecosystem proposed and gives an overview of the features provided from the older
adult perspective and from the caregiver perspective. Then, in Section 3 the whole coaching
and ecosystem proposed and gives an overview of the features provided fr
process is described from the older adult side, presenting the available tools and processes
adult perspective
to foster and from
a positive behavior change theprocess.
caregiver perspective.
Section 4 presents a Then,
detailedinreview
Sectionof 3 the
ing processintervention
the coaching is described from
tools. the5older
Section adult
evaluates side,and
the VCS, presenting
ultimately the available
Section 6 t
summarizes the final conclusions.
cesses to foster a positive behavior change process. Section 4 presents a deta
Some of the preliminary ideas and user interface were presented at the 2020 Interna-
the coaching
tional Conferenceintervention
on Multimodal tools. Section
Interaction 5 evaluates
(ICMI 2020), in thethe VCS, and
Multimodal ultimately S
e-Coaches
marizes the final conclusions.
workshop [24]. The current work presents a refined version and describes in detail several
Some of the preliminary ideas and user interface were presented at the
tional Conference on Multimodal Interaction (ICMI 2020), in the Multimod
workshop [24]. The current work presents a refined version and describes in
technical aspects not presented before. The main contributions of this paper over earlier
works are:
• An updated review of state-of-the-art virtual coaching for older adults’ well-being.
• An in-depth and consolidated description of the complete CAPTAIN ecosystem, also
considering the caregiver perspective.
• A detailed description about how the I-Change and SMART goal concept have been
integrated and implemented as the cornerstone pieces of the VCS.
• A precise specification about how the day plan is built in three stages: weekly, daily
and momentary.
• An enhanced incremental machine learning-supported algorithm for momentary
intervention planning.
• The addition of the analytics aspects of the coach, highlighting its relevance to (1) prop-
erly estimate motivational insights and (2) create goal performance statistics to enhance
the adherence to the system.
• A discussion on the limitations and future work.
2. CAPTAIN Virtual Coaching Ecosystem

2.1. Overall Coaching Approach
The CAPTAIN virtual coaching ecosystem is focused on supporting behavior change
processes of older adult people towards active and healthy aging (AHA), to extend inde-
pendent living not only quantitively but also qualitatively. Effective coaching is a complex
process that additionally requires close cooperation between at least healthcare and social
services, as well as the community, family and friends, including formal and informal
caregivers. Furthermore, it must foster a user’s empowerment by placing them at the
center of the decision-making process.
Our proposal to properly support this complex scenario is founded by digitally
implementing two main concepts:
(1) The I-Change model, previously introduced in Section 1, as the reference framework to
achieve optimal behavior change. This contrasted model will act as a main conductor
which guides our system workflow at every moment. More specifically, we have
mainly focused on providing users with tools that will guide them towards the
successfully consecution of the three pillar phases described by the I-Change model:
awareness, motivation and action (Figure 2).
(2) As a vehicular instrument that guides the process, we based the digitalization of care
plans on SMART goals, as previously described in Section 1. This concept represents
a proper choice since its foundations correlate with the selected methodology:
• Specific: the specific nature of SMART goals helps the system to better design
the action planning suggested by I-Change.
• Measurable: this feature allows to achieve the awareness and motivation phase
while evaluating the impact of the actions and behavior change.
• Achievable: this characteristic also is in line with the motivation phase, but
correlates with the selection of proper actions and their intensity.
• Relevant: the awareness phase relies on highlighting the relevancy of the care
plans for the older adult, who is in the center of the coaching process.
• Time-limited: this feature is also key for the action planning as well as for the
motivation phase.
As presented earlier, the virtual coaching ecosystem (VCE) puts the older adult in the
center of the coaching process. Older adults might be living independently in their home
or supported by caregivers. Each older adult decides which SMART goals to have and
their personalization, either on their own or after a discussion with their caregivers.
plementing two main concepts:
(1) The I-Change model, previously introduced in Section 1, as the reference framework
to achieve optimal behavior change. This contrasted model will act as a main con-
ductor which guides our system workflow at every moment. More specifically, we
have mainly focused on providing users with tools that will guide them towards
Int. J. Environ. Res. Public Health 2021, 18, 6868 5 ofthe
24
successfully consecution of the three pillar phases described by the I-Change model:
awareness,
Int. J. Environ. Res. Public Health 2021, 18, x FOR motivation and action (Figure 2).
PEER REVIEW 5 of 24
(2) As a vehicular instrument that guides the process, we based the digitalization of care
plans on SMART goals, as previously described in Section 1. This concept represents
a proper choice since its foundations correlate with the selected methodology:
• Specific: the specific nature of SMART goals helps the system to better design
the action planning suggested by I-Change.
• Measurable: this feature allows to achieve the awareness and motivation phase
while evaluating the impact of the actions and behavior change.
• Achievable: this characteristic also is in line with the motivation phase, but cor-
relates with the selection of proper actions and their intensity.
• Relevant: the awareness phase relies on highlighting the relevancy of the care
plans for the older adult, who is in the center of the coaching process.
• Time-limited: this feature is also key for the action planning as well as for the
motivation phase.
As presented earlier, the virtual coaching ecosystem (VCE) puts the older adult in the
center of the coaching process. Older adults might be living independently in their home
or supported by caregivers. Each older adult decides which SMART goals to have and
Figure 2.
Figure 2. I-Change
I-Change behavior
behavior change
change model
model [25].
[25].
their personalization, either on their own or after a discussion with their caregivers.
Finally,the
Finally, theVCE
VCEalso
alsoallows
allowscaregivers
caregiversto towork
worknot
notonly
only individually
individuallywith
withusers,
users, but
but
also at the population level by designing SMART goal plans tailored to
also at the population level by designing SMART goal plans tailored to the populationthe population in
a dynamic way. Caregivers can see the evolution of the population characteristics
in a dynamic way. Caregivers can see the evolution of the population characteristics and and
group them,
group them,thenthenevaluate
evaluatethethe impact
impact of of
thethe designed
designed plans
plans andand improve
improve them.them.
Under Under
this
this paradigm,
paradigm, the virtual
the virtual coach coach
systemsystem
can alsocan also suggest
suggest SMART SMART goals adults
goals to older to older adults
based on
based on the effectiveness on other older adults with similar characteristics. Thanks
the effectiveness on other older adults with similar characteristics. Thanks to this approach, to this
approach, population
population grouping,
grouping, plan designplan design andas
and validation, validation, as well plan
well as coaching as coaching
delivery,plan
can de-
be
livery, can be improved iteratively based on
improved iteratively based on the collected feedback. the collected feedback.
2.2. Physical
2.2. Physical Platform
Platform at
at Home
Home
As presented
As presented inin the
the introduction,
introduction, the
the virtual
virtual coach
coach needs
needs to
to be
be able
able to
to track
track the
the user
user
and the
and the environment
environment and and to
to interact
interact with
with them
them inin the
the most
most natural
natural manner
manner to to deliver
deliver the
the
interventions. In
interventions. Inthis
thissection
section
wewe present
present howhowthesethese aspects
aspects were implemented
were implemented in the
in the H2020
H2020 CAPTAIN
CAPTAIN project byproject by describing
describing thearchitecture
the overall overall architecture
under whichunder which the
the virtual virtual
coach was
coach was implemented.
implemented.
The CAPTAIN platform is
The is composed
composedof oflocal
localhardware
hardwareand andsoftware
softwaredeployment
deployment at
each older adult’s home and a central server-side platform (Figure 3). The local and server-
at each older adult’s home and a central server-side platform (Figure 3). The local and
side parts ofparts
server-side the platform interactinteract
of the platform using ausing
RESTa API
REST [26].
API [26].
Figure 3.
Figure 3. Physical
Physical view
view of
of CAPTAIN
CAPTAIN coaching
coachingplatform.
platform.
The server side provides long-term and high-level aggregated and anonymized data
storage, analytics and visualization on the stored data and system management tools, in-
cluding software updates.
The server side provides long-term and high-level aggregated and anonymized data
storage, analytics and visualization on the stored data and system management tools,
including software updates.
The local side at each home follows a distributed and extendable design approach.
Nevertheless, there is always a master device node, taking care of home orchestration and
taking the role of hub with the server side. This device is called the CaptainBox, and it is
implemented using a Raspberry Pi 4 device with 4GB memory and a 64GB SD card. It also
includes a stereo microphone and speakers, as well as a camera and pico-projector. All these
devices are integrated in a 3D-printed case so that the device is self-contained. The rest of
the devices at home contain a subset of these features and extend the sensing and interface
range to all rooms at home. There is an additional hardware which is the MentorAge®
device, produced by the Nively company (partner in the CAPTAIN project), which provides
3D detection of the user location as well as the detection of user actions, protecting user
privacy. Depending on the hardware components that each additional device contains,
devices are categorized as either CaptainSatelliteCamera or CaptainSatelliteProjector. The
first contains a MentorAge® while the second contains a pico-projector, but both include
stereo microphones and speakers and limited computing capabilities.
Regarding the software components, they are distributed among the devices and
interact with them using a publish-subscribe messaging paradigm.
The older adult interacts with the system using speech and video projections in
different locations at each room at home, with multi-language support. This user interface
is supported by the following components:
• Automatic speech recognition (ASR) of several commands for interacting with the
system: numbers, yes/no and certain key words.
• Speech synthesis for interacting with the user (text to speech module, TTS).
• Narrow field of view with 3D-tangible video projection where the user can touch the
surface where the GUI is shown to interact with the controls (e.g., buttons).
• 2D video projection on objects and walls at home, supported by body gesture recogni-
tion (linked to the sensing capabilities).
The CAPTAIN platform provides the following sensing capabilities:
• User location in 3D space at each room, including gait features, posture recognition
and identification of dangerous postures
• User-tangible interaction with physical objects in the room
• Face detection
• Facial recognition and authentication at home
• Facial expression recognition for emotion estimation
• Speech analysis for emotion recognition
• Environment weather sensing
The VCE continuously monitors the user and environment status to provide an
aggregated and updated view of any user recognized by the system as the coached person,
including its location at home and in the room, the object they are interacting with, the
speech command they said and their facial expression. In addition, the environment status
provides information such as the weather in the home surroundings. All this sensing
permits the preceding factors and information factors in the I-Change model, gaining
knowledge about the user and permitting more personalized interactions.
The coach uses the app dispatcher component to identify the most suitable device
to provide the user interface to the user based on his location at home and requests it to
launch an intervention tool (e.g., Agenda), through the instance of the tool projected GUI
component running in that device. This component manages the interaction with the user
and the business logic of the intervention tool while providing a unified graphical user
interface (GUI) for any intervention tool. After finishing the intervention, the tool-projected
GUI reports the to coach that the intervention is finished, along with its outcomes.
Int. J. Environ. Res. Public Health 2021, 18, x FOR PEER REVIEW 7 of 24
2.3. Older Adult Perspective

The overall idea of this virtual coaching system from the participant (i.e., older adult)
platform for them
perspective is towith the purpose
become a lifelong oflearning
fostering(i.e.,
healthy habits through
coaching), motivationala behavior change
and supportive
process executed themselves.
platform for them with the purpose of fostering healthy habits through a behavior change
Long-term
process executedengagement
themselves. and adherence to the plans are two of the key issues within
this kind Long-term engagementcompletely
of approach that are and adherence related to the
to the plansmotivation
are two ofphasethe keysuggested by
issues within
thethis
I-Change
kind ofmodel.
approach Thethatwayare
in which our system
completely relateddeals
to thewith these challenges
motivation is threefold:
phase suggested by the
first, by splitting broad, long-term objectives into time-bound,
I-Change model. The way in which our system deals with these challenges is threefold: specific milestones that we
callfirst,
SMART goals. Second, each of the SMART goal plans dynamically
by splitting broad, long-term objectives into time-bound, specific milestones that we adapts to the user’s
availability
call SMART and capabilities
goals. Second, each (given certain
of the SMART plangoalconstrains to guarantee
plans dynamically adapts a to
behavior
the user’s
change). Finally, by providing the user with periodic feedback about
availability and capabilities (given certain plan constrains to guarantee a behavior change). his progress and
presenting
Finally, by motivational
providing messages
the user withhighlighting
periodic his achievements
feedback about his soprogress
that the user
and can have
presenting
a sense of succeeding
motivational messages and highlighting
can keep track hisofachievements
the evolutionso (i.e.,
thataccomplish
the user can thehave
motivation
a sense of
phase).
succeeding and can keep track of the evolution (i.e., accomplish the motivation phase).
TheThecoaching
coaching process
processitself is presented
itself is presented in in
detail in in
detail Section
Section 3: virtual coach
3: virtual coachsystem.
system.
Nevertheless, it is important to highlight some aspects of this
Nevertheless, it is important to highlight some aspects of this approach. Users can approach. Users can joinjoin
anyany SMART
SMART goalgoalfromfromthethepre-populated
pre-populated catalog
catalog(only
(onlyif the older
if the olderadult
adultfulfills allall
fulfills itsits
requirements).
requirements). Every
Every SMART
SMARTgoal goalcontains
containsaapredefined
predefined initial interventionplan
initial intervention planscheduled
sched-
uled
byby weeks.
weeks. The The system
system stores
stores andand analyzes
analyzes sensing
sensing datadata to choose
to choose the best
the best moment moment to
to deliver
deliver the intervention
the intervention plan.plan. All along
All along this this process,
process, the system,
the system, in addition
in addition to the
to the sensing
sensing data,
data,
also also stores
stores goal-related
goal-related performance
performance andand engagement.
engagement. Using
Using statisticaland
statistical andAIAItools,
tools,the
thecoach
coachcancanevaluate
evaluatethe theuser
userstatus
statusandandsurroundings
surroundingstotocompare compareit it with
with hishis achieve-
achievement
ment in the
in the goals
goals to to consequently
consequently adapt
adapt thethe coachingprocess.
coaching process.GoalGoalplans
planscan
canalsoalso evolve
evolve as
as described
described in innext
nextsection,
section,based
basedon onpopulation-level
population-levelanalytics
analytics comparing
comparing thetheuser
userstatus
status
before
beforeand andafter performing
after performing thethe
goal plans.
goal plans.This process
This processis illustrated
is illustrated in in
Figure
Figure 4. 4.
Figure 4. Older
Figure adult
4. Older perspective
adult of the
perspective virtual
of the coaching
virtual ecosystem.
coaching ecosystem.
2.4.2.4.
Caregiver Perspective
Caregiver Perspective
FromFromthethe
caregiver
caregiverperspective,
perspective,the
theVCE
VCEprovides
providessupport
supportin inthe
theprocess
process ofof provid-
providing
ing care in a more personalized, effective and efficient way iteratively. Different types
care in a more personalized, effective and efficient way iteratively. Different of of
types
caregivers and institutions can take part in this process.
caregivers and institutions can take part in this process.
TheThemain concept
main conceptis that thanks
is that thanksto to
thetheVCE,
VCE, caregivers
caregiverscancananalyze
analyzethethe
participant
participant
(i.e., older adult) population in a structured manner by grouping similar individuals and
(i.e., older adult) population in a structured manner by grouping similar individuals and
analyzing
analyzing their
their evolution
evolution to designed
to designed SMARTSMART goalgoal plans.
plans. Then,Then,
they cantheyevaluate
can evaluate
the ef-the
effectiveness of the plans they design. This process becomes
fectiveness of the plans they design. This process becomes a loop in which existing a loop in which existing
SMART goal plans are refined and replaced by better plans and
SMART goal plans are refined and replaced by better plans and new plans are generatednew plans are generated
as the
as the population
population evolves,
evolves, newnew users
users areare added
added and and
newnew coaching
coaching needsneeds
areare identified.
identified. In In
addition, certain goals can be suggested to the participants based on their effectiveness onon
addition, certain goals can be suggested to the participants based on their effectiveness
other similar participants with similar interests. These ideas are presented in Figure 5. The
technical details of this stratification tool are out of the scope of this paper.
other similar participants with similar interests. These ideas are presented in Figure 5. The
technical details of this stratification tool are out of the scope of this paper.
Figure 5. Caregiver
Figure perspective
5. Caregiver of of
perspective the virtual
the virtualcoaching ecosystem.
coaching ecosystem.
3. Virtual Coach
3. Virtual System
Coach (VCS)
System (VCS)
ThisThis section
section describes
describes thethe virtualcoach
virtual coach system
system from
fromthe
theconceptual
conceptual to to
thethe
technical
technical
aspects. The visual analytics clustering and plan definition tool is out of the scope of
aspects. The visual analytics clustering and plan definition tool is out of the scope of this
this paper.
paper. The VCS focuses only on SMART goals and the I-Change behavior change model, as
The VCS later
described focuses only
in this on SMART goals and the I-Change behavior change model, as
section.
describedEvery
later SMART
in this section.
goal is composed of a weekly plan with coaching actions provided
Every
by SMART
digital goal istools
intervention composed
such as of a weekly plan
recommender with quizzes,
systems, coachingtrainers
actionsandprovided
games. by
digital
Allintervention tools such
of them are focused as recommender
on supporting a behavior systems,
change quizzes, trainers
process. The scopeand games. All
of coaching
of them arethese
covers focused on areas
specific supporting a behavior
or dimensions: change
cognitive, process.
physical, Theand
social scope of coaching co-
nutritional.
vers these specific areas or dimensions: cognitive, physical, social and nutritional.
One key aspect that the VCS focuses on is to simplify the self-empowerment process.
In CAPTAIN, the participant (i.e., older adult) has an active role in two levels. First, it is he
One key aspect that the VCS focuses on is to simplify the self-empowerment process.
(on his own or supported by a caregiver) who decides which SMART goal to join from a
In CAPTAIN, the participant (i.e., older adult) has an active role in two levels. First, it is
catalogue (created by experts in the area and described later in this section), including the
he (on his own
selection of or
thesupported
most suitableby level
a caregiver) whofor
of difficulty decides which
his needs. SMART
Then, goalthe
whenever to join
coachfrom
a catalogue (created by experts in the area and described later in this section),
suggests an intervention to the participant, he can postpone it and the coach will try to including
the selection
adapt theof theto
plan most
meetsuitable
the SMART levelgoal
of difficulty
planning. for his needs. Then, whenever the coach
suggests Participants
an intervention to with
interact the participant, he can
this virtual coach postpone
at home using it and the coach
a combination willand
of 2D- try to
adapt the plan to meet the SMART goal planning.
3D-tangible projections, supported by natural language processing. This system is also
Participants
able to sense theinteract with this
participant’s virtual
location andcoach
bodyatmotion,
home using a combination
recognize of 2D- and
his facial expression
and even identify him thanks to the combination of several image and audio analysis
3D-tangible projections, supported by natural language processing. This system is also
techniques. The devices forming the CAPTAIN system are blended into the participant’s
able to sense the participant’s location and body motion, recognize his facial expression
home, mimicking conventional furniture and appliances.
and even identify him thanks to the combination of several image and audio analysis tech-
niques.
3.1. The devices
I-Change Modelforming the CAPTAIN system are blended into the participant’s
Application
home, mimicking conventional
As introduced in Section furniture and appliances.
2.1, the I-Change model has inspired several aspects of our
VCS, focusing on digital support as its cornerstone: awareness, motivation and action
3.1. I-Change Model 2).
phases (Figure Application
To tackle awareness, each SMART goal plan includes tips to learn
about the aspects in which the goal focuses, such as physical activity. Then, there are
As introduced in Section 2.1, the I-Change model has inspired several aspects of our
several intervention tools (presented in Section 4) which lie among one of the three phases:
VCS, focusing on digital support as its cornerstone: awareness, motivation and action
phases (Figure 2). To tackle awareness, each SMART goal plan includes tips to learn about
the aspects in which the goal focuses, such as physical activity. Then, there are several
intervention tools (presented in Section 4) which lie among one of the three phases: aware-
awareness, motivation and action. The coffee session intervention tool promotes periodic
self-reflection, where the user can check his progress throughout the SMART goal plans
and receive suggestions for next goals. The motivational day agenda shows motivational
messages when the user achieves certain milestones when completing the target day plan.
The motivational day insights provide daily motivational messages when the user has
either a personal or SMART goal-related positive evolution. Most of the intervention tools
try to provide positive feedback of completion when succeeding. Finally, the main purpose
of the coach is to trigger actions, which are interventions contained in the SMART coaching
plans in the most suitable circumstances.
Information factors, also depicted in Figure 2, are covered by using natural user
interfaces based on speech technologies and tangible projections on top of common objects
at home. Simple graphical user interfaces, appropriate speech volume and content density,
provides the possibility of personalizing many setup parameters in the system.
Regarding preceding factors, they are considered requirements for joining each
SMART goal by describing the effort required and the minimum health requirements
to meet the plan, as well as by adapting certain coaching interventions to the environment
of the user. For example, considering the weather in the neighborhood for suggesting
outdoor activities.
3.2. SMART Goals and SMART Goal Plugins

SMART goal plugins are care plans aimed to reach a very specific goal regarding any
of the VCS coaching dimensions (physical, cognitive, social or nutritional) in a specific
period (i.e., range of 4–8 weeks), following the SMART goal concepts. They must rely on
digital intervention tools which are pieces of software that will be detailed in Section 4. A
SMART goal plugin is a self-contained set of data including weekly intervention plans,
trigger rules for the interventions, and configuration parameters so that the participant can
choose the difficulty of the goal. The idea is that third-party institutions and companies can
create their own specific SMART goal plugins and make them available for the participants
to join. This opens the possibility of a market niche in which participants could offer their
own of plans to other participants in exchange for microtransactions.
The coach relies completely on these plugins to schedule and dispatch interventions
to the participant. Consequently, the responsibility derived from the amount and intensity
of the coaching interventions provided to a given participant by the coach following a
SMART goal plugin falls to both the author of the SMART goal plugin as well as in the
participant who accepted to follow the recommendations enumerated within the plugin,
instead of the VCS itself.
When a participant selects and configures a SMART goal plugin, a SMART goal plugin
instance is created with the specific parameters for that goal selected by the participant.
From the participant perspective, he has access to a broad catalogue of SMART goals,
describing their focus (e.g., increase fiber intake habits in eight weeks), requirements (e.g.,
not having Crohn’s disease) and an overview of the expected effort from the participant to
fulfil it. The participant can join a goal in a certain difficulty level (which affects the length
and activities in the plan) and once he finished the plan, join another or even the same (the
coach suggests a suitable new difficulty for the same goal based on his performance). With
this approach, we combine short-term commitment and follow up to very specific goals
with long-term progression.
Each SMART goal plugin plan follows the structure shown in Figure 6. It contains:
• A description of the goal and the requirements to join it.
• The difficulty personalization.
• The weekly plans for each of the difficulty levels including the type and minimum and
maximum number of interventions to carry out, as well as some trigger conditions.
• All data required by the intervention tools in the weekly plan to produce the
proper interventions.
week Target #
intervention
Description and
range
requirements
day Target #
Intervention tool
intervention
personalization X
range
(difficulty)
week 1
Trigger
...
Per week plan conditions
(per difficulty)
SMART goal
plugin week 2 ...
Messages
Questions
Intervention tool
data
Recipes
...
Figure structure
Figure 6. SMART goal plugin 6. SMARTschema.
goal plugin structure schema.
3.3. Dynamic Coach Planning of SMART Goal Plans

3.3. Dynamic Coach Planning of SMART Goal Plans
The coaching framework proposal is depicted in Figure 7. The system is installed in
The coaching framework proposal is depicted in Figure 7. The system is installed
the participant’s home where the hardware is tuned, the system personalized, and the
in the participant’s homeiswhere
participant theregarding
assessed hardwarethe is different
tuned, the systemdimensions.
coaching personalized,Thisand the
assessment is
participant is assessed regarding the different coaching dimensions. This assessment
useful for the clustering process and to get a baseline to measure the evolution. Once is the
useful for the clustering processthe
system is installed, and to get
coach a baseline launches
automatically to measure the session
a coffee evolution. Once
for the the
participant to
join one
system is installed, theorcoach
several SMART goals.
automatically After thea participant
launches has for
coffee session joined
theany SMART goal,
participant to the
Int. J. Environ. Res. Public Health 2021,join onePEER
18, x FOR coachSMART
or REVIEW
several starts delivering interventions
goals. After based on those
the participant goals.11any
has joined of 24SMART goal, the
coach starts delivering interventions based on those goals.
Figure 7. Proposed older adult coaching framework.

Figure 7. Proposed older adult coaching framework.
The VCS works in three timescales—week, day and momentary—using intervention
tools linked to each of these timescales (Figure 7).
Once a week, the coach suggests the participant to have a coffee session. The partici-
pant can decide to postpone this intervention, just like any other intervention. This is an
intervention tool for the participant to reflect on the past week and commit to the next
week’s plan. First, the participant can see several simple charts showing his performance
The VCS works in three timescales—week, day and momentary—using intervention

tools linked to each of these timescales (Figure 7).
Once a week, the coach suggests the participant to have a coffee session. The partici-
pant can decide to postpone this intervention, just like any other intervention. This is an
intervention tool for the participant to reflect on the past week and commit to the next
week’s plan. First, the participant can see several simple charts showing his performance
regarding the rate of activities carried out from the active goal plans as well as the overall
scores. This is meant to be a motivational step. After this step, the participant can join
additional goals or drop active goals if the participant decides to do so. Finally, the partici-
pant can see a summary of the activities planned for the next week. The coffee can also
be requested by the participant, and it is automatically launched if the participant has no
active goals.
Every day when the participant is detected at home, the coach shows the agenda for
the day containing the set of intervention tools for that day as well as the minimum and
maximum amount with the purpose of achieving the SMART goal plugin demands. Then,
the coach shows daily insights if they exist. These interventions can be of three kinds: (a) a
piece of information to increase the awareness about any active goal, (b) a motivational
message about the participant’s performance yesterday compared to historical data, (c) a
motivational message regarding his mood or activity at home, not linked to any goal.
After the agenda and insights have been delivered to the participant, the rest of
interventions can take place. Since the coach does not have access to the participant’s
calendar, it does not know in advance when the participant will be available at home.
Therefore, whenever the participant is detected by the system, it tries to balance between
delivering the maximum number of interventions to maximize daily activities while not
being overwhelming for the participant. To do that, the coach performs a dynamic ranking
of the pending activities based on several aspects and waits between interventions based
on the priority value of the highest ranked intervention combined with an incremental
learning (i.e., type of machine learning) algorithm personalized for the participant, which
can decide to skip delivering an intervention in a certain moment since it anticipates that
the participant would reject it.
When the coach decides to deliver an intervention, it relies on external intervention
tools, which are described in Section 4.
3.3.1. Weekly Plan

Coaching is completely based on the SMART goal plugin plans already presented
in this section which contain per week target activities (i.e., milestones). The coach is
responsible for supporting the participant in achieving those weekly milestones by creating
adaptive per day activity plans and distributing them appropriately during the day using
a momentary planning algorithm.
By combining the static per week plans with the dynamic per day and momentary
planning, we can have the best of both approaches. On the one hand, we can guarantee
that the participant receives a validated coaching plan including the fixed maximum and
minimum intensity ranges so that they are safe and effective for the participant. On the
other hand, the coach can dynamically distribute the interventions through the week and
day to adapt the dispatching to his habits and balance the target of achieving the maximum
intensity with not being too overwhelming for the participant.
3.3.2. Day Plan Generation

Every day the coach reviews all the participant’s active goals to create a combined
day plan. The day plan consists of a list of intervention tools to be launched during the
day, as well as the minimum and maximum number of times to be launched. It can also
include certain requirements to launch, such as specific time ranges or trigger conditions
(e.g., sitting on the sofa for more than 30 min). The adaptive algorithm that builds this
daily plan for every user is described in Section 3.4.
3.3.3. Momentary Planning

Once the day plan has been generated, the coach will try to deliver all the interventions
in the plan to the participant throughout the day. This momentary planning must also
strike a balance between achieving the maximum number of interventions in the agenda
and keeping the participant engaged without being overwhelming. In addition, the coach
cannot know in advance when the participant will be available at home since it cannot
have access to the participant’s calendar.
Therefore, the coach has been designed to solve the next problem description.
Goals:
• Primary: To achieve the minimum target intervention set for the day.
• Secondary: To achieve the maximum target intervention set for the day.
User-friendly restrictions:
• Avoid launching all interventions in a row.
• Avoid launching the same type of intervention tool twice in a row.
• Adapt to the participant’s behavior.
Intervention tool restrictions:
• Time range-based restrictions.
• Trigger condition required to launch an intervention.
• Intervention tool priorities.
Limitations for the momentary coach:
• Cannot know in advance when the participant will be at home.
• Cannot know the participant’s preference for doing certain interventions in certain
situations.
• If a participant rejects a certain intervention, the coach does not know why. It could
be because it is not the proper situation, moment of the day or intervention tool.
Therefore, it does not know if it should insist or not and when.
Based on the problem description, the proposed algorithm performs as follows. Each
time the participant is detected by the sensing system or his status changes over time, the
coach evaluates if there is any intervention that could be launched and decides whether it is
a good moment to launch it or not. If the participant is already doing another intervention,
the coach sticks to the previous intervention unless the participant leaves the room. In that
case, the intervention is considered rejected by the participant, so it can be launched again
later. The algorithm follows two steps: (a) selecting the most suitable intervention from
those pending in the day plan and (b) deciding if the moment is appropriate to launch it.
To select the most suitable intervention, only those interventions which can be
launched under current circumstances are ranked. This filtering removes those inter-
vention tools which have already achieved the maximum number or those which cannot
be triggered, either because it is not the proper time range (according to the plan) or it
contains certain trigger conditions that are not met in this moment (e.g., person touching
the fridge). The priority ranking formula uses a linear combination of several parameters,
as presented:
priority ranking
= 1.1 ∗ participant requested + 0.6(not minimum target interventions reached)
+0.2 ∗ (it was not the previous intervention) + 0.1
∗(intervention requires condition) + 0.06 ∗ (1 − (remaining time/(time range)))
+0.04 ∗ (1 − accomplishment rate)
(
1 i f (carried out interventions >= minimum target interventions)
acomplishment rate = carried out interventions
minimum target interventions , otherwise
where participant requested is a Boolean value which is one if this intervention has been
requested by speech command and zero if not. The rest of the Boolean parameters are
coded with the same representation.
This formula gives top priority to interventions directly requested by the participant
(participant requested). Then, it gives more priority to intervention tools which have not
reached the minimum target (not minimum target interventions reach). It tries to avoid
launching the same intervention tool twice in a row (if it was not the previous intervention)
and gives more priority to those which require a trigger condition (intervention requires
condition). Finally, it gives more priority to those intervention tools which have less of a time
range available to be launched. The set of parameters and their importance (i.e., weight
in the linear combination) were obtained empirically after discussion with non-technical
partners in the CAPTAIN project.
To decide if this is an adequate moment to deliver an intervention to the participant,
two tests are carried out. The first one is checking if “enough time” has passed from the last
intervention to avoid overwhelming the participant. This inter-intervention time is based
on the priority of the selected intervention. Therefore, in the component configuration, we
can set different waiting times for different intervention priority value ranges. Thanks to
this mechanism, the coach can be more insistent for interventions with higher priority and
less for less important ones. For the second check, we use an incremental machine learning
Int. J. Environ. Res. Public Health 2021,model
18, x FOR PEERtries
which REVIEW
to figure out if the participant will accept a certain intervention in the14 of 24
status before actually delivering it. In any case, since this machine learning algorithm could
end up biased by the participant’s behavior, for example, by rejecting all interventions, it
cannot cancel more than a configurable number of interventions in a row.
The incremental machine learning model learns each participant’s patterns for ac-
The incremental machine learning model learns each participant’s patterns for accept-
cepting and rejecting
ing and rejecting interventions.
interventions. Theisinput
The input is the participant
the participant status
status when thewhen the interven-
intervention
tion is about to be delivered, timing information, historic information
is about to be delivered, timing information, historic information about previous rejected about previous re-
jected interventions
interventions and environment
and environment information,
information, as shown as in shown in With
Figure 8. Figure 8. With
these these fea-
features,
tures,
the model the predicts
model predicts if it is
if it is a good a goodtomoment
moment to launch
launch that that intervention
intervention or not with
or not with certain
certain probability.
probability. There is anThere is an intermediate
intermediate feature
feature selection selection
step step the
to improve to improve
accuracy the accuracy
of the
of the model, which is also dynamically adapted to each specific participant.
model, which is also dynamically adapted to each specific participant.
Figure
Figure 8.8.Incremental
Incrementallearning
learning
MLML model
model training
training and and inference.
inference.
One of the characteristics of this model is that it uses incremental learning, which
means that it is continuously improving and adapting as the participant carries out new
interventions or rejects them.
Finally, the feature selection itself adapts to the participant behavior. Feature selec-
tion is a common technique in machine learning to improve the model’s accuracy for sim-
ple models and a low amount of data points. Each time a participant finishes an interven-
One of the characteristics of this model is that it uses incremental learning, which
means that it is continuously improving and adapting as the participant carries out new
interventions or rejects them.
Finally, the feature selection itself adapts to the participant behavior. Feature selection
is a common technique in machine learning to improve the model’s accuracy for simple
models and a low amount of data points. Each time a participant finishes an intervention, all
features described in this section are stored in a local database, including if the participant
accepted or rejected that intervention. Periodically, when enough new samples are available,
the feature selection algorithm is trained to use the most meaningful features from all
collected for that participant.
3.4. Coach Analytics

The virtual coach performs these batch computation data analysis processes periodically:
• Day plan generation: this module generates the day plan introduced in Section 3.3.2
daily. To do so, it combines information from three different sources: (1) the corre-
sponding goals a user is subscribed to, (2) the weekly configuration of those goals
and (3) the set of interventions the user has already carried out. The planification
process can be described as follows: between the first half of the week, the coach
evenly distributes the weekly interventions to try to achieve the maximum number of
interventions set by the goal configuration. From Thursday onwards, if the person
has been able to easily reach the minimum set of interventions, the algorithm keeps
pushing the participant towards the maximum but if the participant has not been
able to keep up with the plan, the target is lowered towards the minimum. This
algorithm has been designed to balance between achieving the maximum and keeping
the user motivated.
• Goal performance statistics and chart generation: to keep the user updated as well as
motivated as much as possible, goal performance statistics are frequently generated.
These statistics are translated into charts that can reflect progress in two different levels:
(1) high-level charts that depict the performance related to the whole goal duration
(long memory) and (2) low-level charts that show the most recent performance (short
memory). Both types of charts unveil information in a quantitative (how many
interventions have they completed compared to the schedule?) as well as qualitative
(how well do they complete them?) manner. All these charts are presented trying
to boost the user adherence and motivation, so they attempt to always deliver the
information in the kindest/most optimistic way possible. They automatically appear
during the coffee sessions explained in Section 4.1.1.
• Motivational message generation: from the daily usage of the system, a lot of informa-
tion related to the user’s general status as well as to the interventions they periodically
perform as part of their goals is collected. These two sources of information are ana-
lyzed to generate insights that could provide some insight or even used as a source
of motivation for the user. These insights are translated into text messages that are
delivered to the users, always according to the goal configuration. As a backup plan
in those cases where the extraction of positive insights cannot be achieved, some pre-
defined plain general messages that must always accompany every goal configuration
are the pieces of information that will be delivered instead. This implementation will
be more specifically detailed in Section 4.1.3.
4. Coaching Intervention Tool Details

The intervention tools are the software applications which provide the actual coaching
orchestrated by the virtual coach. In this work several example software applications were
implemented to provide coaching on the physical, nutritional, cognitive and social areas.
On top of them, the coffee session application corresponds to the goal management and
progress review intervention tool, which is central to the overall coaching process. The
coffee session, motivational day agenda and day insights focus more on the awareness and
motivation aspects from the I-Change model, while the rest of intervention tools triggered
Int. J. Environ. Res. Public Health 2021, by
18, the VCS
x FOR are REVIEW
PEER more focused on the specific actions to take towards the behavior change 16
process, even though they may include some motivational and awareness aspects. Figure 9
shows the intervention tool distribution through the coaching timespan.
Figure
Figure 9. Coaching
9. Coaching intervention
intervention tools in tools in different
different timescales.timescales.
4.1. Core Coaching Intervention Tools

4.1. Core Coaching Intervention Tools
The set of tools described next are core to the system since they organize the coaching
Thefrom
process set of
thetools
user described
perspectivenext aredifferent
in the core to timescales,
the systemassince they organize
previously presentedthe coach
process
in Figurefrom
7. the user perspective in the different timescales, as previously presente
Figure 7.
4.1.1. Coffee Session
As presented above, the coffee session is an intervention tool to perform reflection
4.1.1. Coffee Session
about self-performance of the goals during the past week and the goals that the participant
As presented
has joined. above,athe
This tool follows coffee
fixed session
schema is an
with three intervention tool to perform reflec
steps:
about
1. self-performance
Review one by one the ofperformance
the goals during the past
of the active week
goals withand the goals
a sequence that the particip
of charts
has
2. joined.
Give theThis tool follows
possibility a fixed
to join new schema with three steps:
goals
3. Show a preview of the target activities for the next week
1. Review one by one the performance of the active goals with a sequence of charts
2. Review
Give the performance
possibility to join new goals
This step is carried out per each active goal the participant has joined and it shows two
3. Show a preview of the target activities for the next week
charts, one to depict the overall goal performance and one to show last week’s performance.
Review
An performance
example of an overall goal completion chart, previously introduced in Section 3.4,
This step is carried
is shown in Figure 10. Thereout
is aper
bar each
chart active goal
per week. the
It is participant
filled hasthejoined
in based on numberand it sh
two charts, one to depict the overall goal performance and one to show lastItweek’s
of interventions carried out regarding the maximum target interventions for that week.
is an average for all interventions. The bar color highlights the score of the interventions
formance.
in that period. Emojis are used to show the score of the interventions (not shown in the
figureAn example of an overall goal completion chart, previously introduced in Section
below).
is shown in Figure 10. There is a bar chart per week. It is filled in based on the numbe
interventions carried out regarding the maximum target interventions for that week.
an average for all interventions. The bar color highlights the score of the intervention
that period. Emojis are used to show the score of the interventions (not shown in the fig
below).
J. Environ. Res. Public Health 2021, 18, x FOR PEER REVIEW
Figure 10. Coffee review progress, overall performance.
Then, if the participant wants to see more details, he can rev

formance
Figure
Figure
following
10. Coffee
10. Coffee review
review progress,
theprogress,
overall
same idea for the chart structure where
overall performance.
performance.
day instead of a whole week. After reviewing the goal, the part
Then, if the participant wants to see more details, he can review the most recent
cancel
performance the
Then, goal
if the
following and
the samemove
participant towants
idea for the review
chart tothe
see
structure next
more
where eachgoal. The he
details,
bar representsparticipa
can rev
one
any
day goal.
instead
formance
of a whole week.
following
After reviewing
thethe same
the
idea
goal, the participant has the option to
cancel the goal and move to review next goal. Thefor the chart
participant structure
can skip the review ofwhere
day
any When the
instead
goal. of aparticipant
whole week. finishes
Afterthe whole plan
reviewing for a the
the goal, goal, h
part
message
message theas
as well
welland
asgoal
as amove
a suggestion
suggestion
to review
to try a harder
to trythe a harder
next
or easier difficulty
or easier
When the participant finishes the whole plan for a goal, he receives a motivational
cancel level goal.
based onThe
difficulty
participa
the overall
all performance,
any goal.as presented as
performance, presented
in Figure 11. in Figure 11.
When the participant finishes the whole plan for a goal, he
message as well as a suggestion to try a harder or easier difficulty
all performance, as presented in Figure 11.
Figure
Figure 11. Coffee
11. Coffee review
review progress, goal progress, goal
finish suggestion. finish suggestion.
After reviewing the performance of the active goals, the next optional step is to join
After
a new goal. reviewing
The participant uses athe performance
catalogue of predefined of the
ne
plans active
with differentgoals,
difficultythe
levels grouped by the coaching dimension they are working on, as shown in Figure 12.
a new goal. The participant uses a catalogue of predefined plans
Figure 11. Coffee review progress, goal finish suggestion.
levels grouped by the coaching dimension they are working on,
After reviewing the performance of the active goals, the ne
a new goal. The participant uses a catalogue of predefined plans
Figure 11. Coffee review progress, goal finish suggestion.
After
Int. J. Environ. Res. Public Health 2021, 18, 6868reviewing the performance of the active goals, the next optional step is 17
toofjoin
24
a new goal. The participant uses a catalogue of predefined plans with different difficulty
levels grouped by the coaching dimension they are working on, as shown in Figure 12.
Int. J.J. Environ.

Int. Environ. Res.
Res. Public
Public Health 2021,(a)
Health 2021, 18, xx FOR
18, FOR PEER
PEER REVIEW
REVIEW (b) 18 of
18 of 24
24
Figure 12.
Figure 12. Join
Join goal
goal in
in coffee
coffee session.
session. (a)
(a) Select
Select goal
goal category;
category; (b)
(b) select
select goal
goal in
incategory.
category.
Once
Once the
Once the participant
theparticipant selects
participantselects one
selectsone goal,
onegoal, hehe
goal,he can
cancancheck itsits
check
check its description as well
description
description as well
as wellas the
as the re-
as the
re-
quirements
requirements for the
for participant
the participantto join
to it.
join After
it. Afterthat, the
that, theparticipant
participantcan
canselect
select the
the
quirements for the participant to join it. After that, the participant can select the difficulty difficulty
difficulty
level for
level for
level that
for that goal
that goal (Figure
goal (Figure 13).
(Figure 13).
13).
(a)
(a) (b)
(b)
Figure
Figure 13.
Figure13. Join
13. Join goal
Joingoal in
goalin coffee
incoffee session.
coffeesession. (a)
session. (a) Goal
(a)Goal description
Goaldescription and
descriptionand requirements;
andrequirements; (b)
requirements; (b) set
(b) set goal
set goal difficulty.
goal difficulty.
difficulty.
Finally,
Finally, the
Finally, the user
theuser can
usercan get
cangetgetaa a
preview
preview
preview of of
of thethe
the target
target activities
activities
target for for
for
activities thatthat
that goalgoal
goal per week
per week and
and
per week
confirm that he wants to join the goal. Once confirmed, the participant receives
confirm that he wants to join the goal. Once confirmed, the participant receives a motiva-a
and confirm that he wants to join the goal. Once confirmed, the participant a motiva-
receives
tional
tional message
message
motivational (Figure(Figure
(Figure
message 14).
14). 14).
(a)
(a) (b)
(b)
Figure 14.
Figure 14. Join goal
goal in coffee
coffee session. (a)
(a) Plan summary
summary and confirmation;
confirmation; (b) confirmation
confirmation and motivational
motivational message.
Figure 14. Join
Join goal in
in coffee session.
session. (a) Plan
Plan summary and
and confirmation; (b)
(b) confirmation and
and motivational message.
message.
After finishing
After finishing the
the goal
goal review
review and
and join
join steps,
steps, the
the participant
participant can
can have
have aa preview
preview of
of
the overall target activities that will be scheduled the next week for all goals he has joined.
He will
He will then
then receive
receive aa reminder
reminder toto have
have another
another coffee
coffee session
session next
next week
week (Figure
(Figure 15).
15).
After finishing the goal review and join steps, the participant can have a preview of
He will then receive a reminder to have another coffee session next week (Figure 15).
(a) (b)
Figure 15. Preview target activities next week and reminder for next week. (a) See the target activ
reminder for next week’s coffee.
(a)4.1.2. Motivational Day Agenda (b)

Figure 15.
15. Preview
Previewtarget
targetactivities
activitiesnext
nextweek
weekand
andreminder
reminderforfor
next week. (a)(a)
SeeSee
thethe
target activities for for
nextnext
week; (b)
Figure
reminder for next week’s coffee. The day agenda is an intervention tool which shows the set
next week. target activities week;
(b) reminder for next week’s coffee.
be carried out during the day. It does not set a specific time to do t
4.1.2.
4.1.2. Motivational
Motivational Day Day Agenda
in The
Section 3.3.3: Momentary
Agenda
planning). This tool is launched the
The day agenda is an intervention
day agenda is an intervention tool
tool which
which shows
shows the
the set
set of
of target
target interventions to
interventions to
beiscarried
be detected
carried out every
out during
during the day.
the day.
day. It
It does
does not
not set
set aa specific
specific time
time to
to do
do them
them (rationale
(rationale explained
explained
in
in Section 3.3.3: Momentary planning). This tool is launched the first time the participant
participant
is
SectionIn addition
3.3.3: Momentary to showing the target interventions for the da
planning). This tool is launched the first time the
is detected
detected every
every day.
day.
agenda
In is automatically
In addition
addition to
to showing
showing the launched
the target
target interventionswhenever
interventions for
for the
the day, the
the
day, the participant
motivational
motivational day achi
day
agenda is automatically launched whenever the participant achieves a milestone of com-
pleting
pleting
pleting the
the interventions
the target
target interventions
target interventions for
forthe
thedaydaywith
formotivational
the day
witha amotivational
withasaas
agenda is automatically launched whenever the participant achieves a milestone of com-
message,
message,
motivational
depicted in
depicted
inFigure
Figure 16.16.16.
Figure
Figure 16. Mock-up for motivational day agenda.
The motivational day agenda can detect the next milestones and produce a motiva-
tional message each time any of the following is achieved:
• Minimum interventions for a goal reached
• Maximum interventions for a goal reached
• Minimum interventions for all goals reached
• Figure 16. Mock-up forallmotivational day agenda.
Figure 16. Mock-up for motivational day agenda.
Maximum intervention for goals reached
The motivational day agenda can detect the next milestones and produce a motiva-
4.1.3.
tionalDay Insight
The motivational
message each time any of theday agenda
following can detect the next milestone
is achieved:
The
Minimum interventions for a goal reached piece of knowledge which is delivered to
day insight is a message presenting a
•tional message eachfortime any 17).of thearefollowing is achieved:
•
the participant
Maximum after the day
interventionsagenda (Figure
a goal reached There three kind of day insights based
on the data used to obtain them:
•
• Minimum interventions for a goal reached
Goal-based insight
• Maximum interventions for a goal reached
• Minimum interventions for all goals reached
example, for the “Increase fiber intake” nutritional SMART goa
“Replacing fruit juices with whole fruits is an easy way to increa
Goal performance analytics insights compare yesterday’s p
terventions with the participant’s historic performance to produc
It also compares the last week against the whole month. For exa
•Yesterday you got the
Minimum interventions for alloverall best score in your interventions for
goals reached
• Maximum intervention for all goals reached
goal.”
User model analytics day insights perform statistical analy
4.1.3. Day Insight
The day insight is a message presenting a piece of knowledge which is delivered to
pant-sensing
the data
participant after the to produce
day agenda (Figure 17).aThere
motivational message.
are three kind of day insights For
based exam
ing
on the more
data usedpositive. Good for you.” This message would be genera
to obtain them:
•
•
recognition modules. They aggregate data to estimate the part
Goal-based insight
Goal performance analytics day insight
•use statistical
User tools
model analytics dayto identify trends and outliers.
insight
Figure
Figure 17.insight
17. Day Daygoal
insight goal
performance. performance.
Goal-based insights are predefined messages contained in the SMART goal plugin to
4.2. Goal
increase Plan-Dependent
the awareness Intervention
of the participant regarding theTools
objectives of a SMART goal. For
example, for the “Increase fiber intake” nutritional SMART goal, one message could be:
“ReplacingThese action-oriented
fruit juices intervention
with whole fruits is an easy way to increasetools
your completely
fiber intake.” depen
SMARTwith goal and are not, therefore,
performance key parts of this VCS like the
Goal performance analytics insights compare yesterday’s performance doing the in-
terventions the participant’s historic to produce motivational messages.
Itday agenda the
also compares and lastday
weekinsights.
against the They are exchangeable
whole month. pieces in this s
For example: “Congratulations!!
Yesterday you got the overall best score in your interventions for the increase fiber in-
inggoal.”
take components and user interface. Nevertheless, several action-o
examples
User modelimplemented
analytics day insightsduring the CAPTAIN
perform statistical project
analysis of aggregated are presen
participant-
sensing data to produce a motivational message. For example: “Lately I see you being
more positive. Good for you.” This message would be generated based on the emotion
4.2.1. Gamified
recognition modules. They Trainer: Physical
aggregate data and
to estimate Cognitive
the participant’s mood and then use
statistical tools to identify trends and outliers.
The gamified training approach focuses on two pillars: cog
4.2. Goal Plan-Dependent Intervention Tools
ants. The physical trainer consists of a large set of gamified ph
These action-oriented intervention tools completely depend on the scope of each
controlled
SMART goal and by body
are not, movement,
therefore, recognized
key parts of this by asession,
VCS like the coffee smart video came
motivational
a bigcomponents
pool of and gamified exercises and configure theintervention
amount per s
day agenda and day insights. They are exchangeable pieces in this system, just like the
sensing user interface. Nevertheless, several action-oriented
tensity,
tool exampleswith different
implemented duringtargets ranging
the CAPTAIN from
project are muscle
presented below.strengthenin
During the gameplay, the participant follows the instructions pr

of each exercise with text accompanied by a gif image.
Cognitive games show a similar interface, but they target th
mension.
4.2.1. Gamified Trainer: Physical and Cognitive

The gamified training approach focuses on two pillars: cognitive and physical variants.
The physical trainer consists of a large set of gamified physical exercises that are controlled
by body movement, recognized by a smart video camera. One can choose from a big
pool of gamified exercises and configure the amount per session, as well as the intensity,
with different targets ranging from muscle strengthening, stability or endurance. During
the gameplay, the participant follows the instructions presented at the beginning of each
exercise with text accompanied by a gif image.
Cognitive games show a similar interface, but they target the cognitive coaching dimension.
4.2.2. Recommenders
The system includes several types or recommender systems which basically provide
short motivational interventions in the form of messages calling to specific actions offered
to the participants in a personalized way. The personalization is based on the participant
profile and on the previous interactions with the coach, so it is expected that the suitability
of the recommendations and engagement of the participant with the coach increases
over time.
The recipe recommender provides the participant with suggestions of healthy recipes
aligned with the nutritional SMART goal. The user can see the ingredient list as well as a
video and if he does not like the suggestion, he can get additional recipe options.
The social contact recommender is used to deliver motivational messages prompting
the participant to contact specific relatives and/or friends to encourage him to communicate
with them and improve/maintain his social skills.
The social event recommender aims at encouraging the participants to attend events
and activities that are being organized in their surroundings, thus promoting social interac-
tion outside their home.
The physical activity recommender is intended to promote an active lifestyle in the
participant through motivational messages encouraging the participant to perform some
exercises adapted to the participant’s profile.
The cognitive recommender is designed to deliver motivational messages and recom-
mendations towards fostering the emotional and mental well-being of the participant, thus
preserving their mental capacity and quality of life.
4.2.3. Quiz
The quiz is a tool designed to increase the awareness of the user regarding the goal
topics. It asks questions to the participant about the goal topics. Besides, it provides an
explanation on the rationale behind the correct answer, independent from the answer
submitted by the participant.
Contents and difficulty levels are tailored to the active SMART goal(s) and participant
profile, respectively, and feedback from the participant about the usefulness of the question
proposed is taken into consideration by the coach to further personalize the subsequent
quizzes. This tool is based on the spaced learning methodology [27], aiming at maximizing
participant’s retention of concepts while minimizing the burden to acquire new knowledge.
5. Discussion
As presented in the Introduction, virtual coach systems are still in the early stage of
development, far from Siewiorek’s [5] definition. Nevertheless, the H2020 RIA projects
funded on virtual coaching for older adults are pushing towards the sophistication of
coaching systems. Indeed, an integration of the outcomes of several of these projects
combining advanced sensing; natural language interaction; short-, mid- and long-term
objective management; and dynamic adaptation and personalization of the coaching
process could mostly cover the definition.
Our coach combines SMART goal plans designed and validated by experts on the
coaching domain with an adaptive personalization of the intervention delivery provided
by the virtual coach. This process employs a combination of data analysis techniques with
deterministic algorithms and incremental machine learning algorithms. To our knowledge,
this is the first time that this kind of hybrid approach has been presented.
We see several paths to improve our work, from the sensing and user interface aspects
to the coach itself and intervention tools. Next, we present and open discussion on the
most important improvements.
First, the system coverage is limited currently to older adults’ homes, which constrains
sensing capabilities as well as coaching support. It would be desirable to extend this work
with at least a companion smartphone application to extend the coverage, and even to the
use of wearable devices.
On the sensing aspects, our system currently cannot automatically evaluate the stage
in the I-Change behavior change model in which the VCS should put more emphasis. The
VCS currently does not have enough sensing capabilities to properly deal with information
factors and preceding factors to properly manage the personalization and adaptability
of the system to the user preferences. To overcome this limitation, advances on sensing
devices and software, as well as the extensive use of data intensive machine learning
algorithms, seems to be the path to follow.
On the user interface area, natural-language processing technologies employed by
our coach are too basic. It is mandatory to use the latest language processing technologies
to keep user engagement. This implies using advanced automatic speech recognition
algorithms and speech synthesis technologies. Even more importantly is the use of natural-
language understanding for user intent recognition and natural-language generation for
conveying information in the most effective and real way. All of them are adapted to the
older adult characteristics. Video projection devices also need improvements to deal with
the occlusions, strong light sources and mild visual impairment of older adults. Finally, the
hardware devices providing these services currently need to be integrated manually in the
user’s home and this imposes strong constrains for certain types of houses. Even if they
could be embedded into furniture, this process is manual and requires expert technicians
for installation and maintenance. In addition, this increases the cost of the overall system
and limits its global application. A simplification of the system and evolution of hardware
and low-level sensing are required for commercial exploitation.
On the coaching aspects we see that we need a clearer way to map the SMART goal
concepts into the coaching plans and everyday coaching. In addition, we should explore
alternative approaches to the SMART criteria for goal definition, such as PURE (the right
goal, Positively Stated, Understood, Relevant, Ethical) [28]. It is also important to connect
and group SMART goals under a long-term goal for the user to have a big picture to foster
motivation. We also see that we need to improve the balance between static SMART goal
plans and dynamic scheduling by the coach. This is a complex area because making the
coach decide the interventions for the user on its own (or modifying pre-defined plans)
would make this system a medical device with implications on safety and regulatory
aspects. On the other hand, it could permit finer-grain coaching adaptation to the user
evolution. Finally, we are not yet taking advantage of the data collected in the server
side about population analytics to refine and validate coaching plans nor to improve user
experience and personalization.
Regarding the coaching intervention tools, we see a potential market here for software
tools and support hardware. Our concept is extendable to cover new tools; nevertheless, for
the time being the process of adding new intervention tools is not completely transparent.
It requires extending the unified GUI application, designing the interaction flow with it, as
well as some modifications of the coach itself.
Finally, a proper long-term validation of this concept implementation is required. Due
to COVID-19 restrictions, the extension and coverage of the validation of this implementa-
tion was dramatically reduced, therefore it was not included in this paper.
We consider that the preventive virtual coaching concept and ecosystem presented
here can be a key piece in the future active and healthy aging systems for independent
living, combined with other services such as integration with healthcare, as a telemedicine
tool, specifically to manage chronic diseases and pre-frailty stages, monitoring for dan-
gerous situations preventively and in real time, and as an assistive tool for physical and
cognitive decline. Such a complete system for ambient assisted living at home would
extend healthy years, reduce costs for healthcare and social services and provide better
quality of life to older adults.
6. Conclusions
This paper presented a virtual coaching concept and the implementation of the concept
from the CAPTAIN H2020 (2017–2021) research and innovation action project.
The virtual coach system is grounded on the concepts of SMART goals and the I-
Change behavior change mode which aims to empower the older adult in reaching his
goals and monitoring his progress, even if he is supported by a caregiver. Coaching covers
the cognitive, physical, social and nutritional areas on a preventive well-being level.
The virtual coach is complemented by user sensing devices and hardware, developed
during the CAPTAIN project, including speech-based bidirectional interaction with the
user and video pico-projectors at home, following edge computing approaches with the
support of a server-side control panel.
The design and implementation of the approach permits its extendibility to more
use cases and coaching intervention tools, as well as its adaptability to other coaching
domains and types of population. It also fosters the creation of a commercial market where
third-party institutions can contribute by defining new coaching plans.
Finally, the authors consider that the virtual coaching concept presented here will be
part of the future active and healthy aging systems for independent living, combined with
other services. It will become a milestone in ambient assisted living at home, which will
foster extended healthy years, reduced costs for healthcare and social services and provide
better quality of life to older adults.
Author Contributions: Conceptualization, A.B.I. and R.Á.S.; methodology, A.B.I. and R.Á.S.; soft-
ware, A.B.I. and R.Á.S.; investigation, A.B.I. and R.Á.S.; resources, A.B.I. and R.Á.S.; writing—original
draft preparation, A.B.I. and R.Á.S.; writing—review and editing, A.B.I. and R.Á; supervision, A.B.I.;
project administration, A.B.I.; funding acquisition, A.B.I. All authors have read and agreed to the
published version of the manuscript.
Funding: This project has received funding from the European Union’s Horizon 2020 research and
innovation program under grant agreement No 769830.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Acknowledgments: We thank Despoina Petsani (Aristotle University of Thessaloniki), Pjotr Mjakosin
(DigiFLAK) and Francisco J. Núñez-Benjumea (Salumedia) for providing the pictures of the hardware
and some applications, and useful discussions. We thank Evdokimos Konstantinidis (Aristotle
University of Thessaloniki) for his guidelines as the CAPTAIN project technical coordinator.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Department of Economic and Social Affairs (United Nations). World Population Prospects 2019; United Nations: New York,
NY, USA, 2019.
2. World Health Statistics 2019: Monitoring Health for the SDGs, Sustainable Development Goals. Available online:
https://www.who.int/publications-detail-redirect/world-health-statistics-2019-monitoring-health-for-the-sdgs-sustainable-
development-goals (accessed on 3 February 2021).
3. Boland, L.; Légaré, F.; Perez, M.M.B.; Menear, M.; Garvelink, M.M.; McIsaac, D.I.; Painchaud Guérard, G.; Emond, J.; Brière, N.;
Stacey, D. Impact of home care versus alternative locations of care on elder health outcomes: An overview of systematic reviews.
BMC Geriatr. 2017, 17, 20. [CrossRef] [PubMed]
4. Ajduković, M.; Cajvert, L.; Judy, M.; Knopf, W.; Kuhn, H.; Madai, K.; Voogd, M. ECVision. A European Glossary of Supervision
and Coaching. In In the Mirror of Compentences, Supervision and Coaching in Europe; Facultas: Vienna, Austria, 2006; Volume 1,
pp. 110–183, ISBN 978-3-7089-1480-0.
5. Siewiorek, D.P. Invited talk: Virtual coaches in health care. In Proceedings of the IEEE/IFIP International Conference on
Dependable Systems and Networks Workshops (DSN 2012), Boston, MA, USA, 25–28 June 2012; p. 1.
6. Wolever, R.Q.; Simmons, L.A.; Sforzo, G.A.; Dill, D.; Kaye, M.; Bechard, E.M.; Southard, M.E.; Kennedy, M.; Vosloo, J.; Yang, N. A
systematic review of the literature on health and wellness coaching: Defining a key behavioral intervention in healthcare. Glob.
Adv. Health Med. 2013, 2, 38–57. [CrossRef] [PubMed]
7. Bacciu, D.; Chessa, S.; Gallicchio, C.; Micheli, A.; Ferro, E.; Fortunati, L.; Palumbo, F.; Parodi, O.; Vozzi, F.; Hanke, S.; et al. Smart
environments and context-awareness for lifestyle management in a healthy active ageing framework. In Progress in Artificial
Intelligence; Pereira, F., Machado, P., Costa, E., Cardoso, A., Eds.; Springer International Publishing: Cham, Switzerland, 2015;
pp. 54–66.
8. Gordienko, Y.; Stirenko, S.; Alienin, O.; Skala, K.; Sojat, Z.; Rojbi, A.; López Benito, J.R.; Artetxe González, E.; Lushchyk, U.; Sajn,
L.; et al. Augmented coaching ecosystem for non-obtrusive adaptive personalized elderly care on the basis of cloud-fog-dew
computing paradigm. In Proceedings of the 2017 40th International Convention on Information and Communication Technology,
Electronics and Microelectronics (MIPRO), Opatija, Croatia, 22–26 May 2017; pp. 359–364.
9. Lete, N.; Beristain, A.; García-Alonso, A. Survey on virtual coaching for older adults. Health Inform. J. 2020, 26, 3231–3249.
[CrossRef] [PubMed]
10. Gonzalez-Fraile, E.; Gonzalez-Pinto, A.; Tenorio-Laranga, J.; Fernandez-Ruanova, B.; Olaso, J.M.; Montenegro, C.; Santana, R.;
Vazquez, A.; Justo, R.; Lozano, J. Empathic, expressive, advanced virtual coach to improve independent healthy-life-years of the
elderly (the empathic project: Mid-term achievements). In European Psychiatry; Cambridge University Press: Cambridge, UK,
2020; Volume 63, p. 384.
11. Op den Akker, R.; Beinema, T.; Banos, O.; Heylen, D.; Bedsted, B.; Pease, A.; Pelachaud, C.; Salcedo, V.T.; Kyriazakos, S.;
Hermens, H. Council of coaches a novel holistic behavior change coaching approach. In Proceedings of the ICT4AWE 2018—4th
International Conference on Information and Communication Technologies for Ageing Well and e-Health, Madeira, Portugal,
22–23 March 2018; SCITEPRESS: Madeira, Portugal, 2018; pp. 219–226.
12. Dimitrov, Y.; Gospodinova, Z.; Wheeler, R.; Žnidaršič, M.; Ženko, B.; Veleva, V.; Miteva, N. Social activity modelling and
multimodal coaching for active aging. In Proceedings of the 12th ACM International Conference on PErvasive Technologies
Related to Assistive Environments, Rhodes, Greece, 5–7 June 2019; Association for Computing Machinery: New York, NY, USA,
2019; pp. 608–615.
13. Kyriazakos, S.; Schlieter, H.; Gand, K.; Caprino, M.; Corbo, M.; Tropea, P.; Judica, E.; Sterpi, I.; Busnatu, S.; Philipp, P.; et al. A
Novel Virtual Coaching System Based on Personalized Clinical Pathways for Rehabilitation of Older Adults—Requirements and
Implementation Plan of the vCare Project. Front. Digit. Health 2020, 2. [CrossRef]
14. Virtual Coaches for Older Adults’ Wellbeing: A Systematic Review. Available online: https://ieeexplore.ieee.org/document/90
97837 (accessed on 7 June 2021).
15. Devizio, S.M. The Attitude-Social Influence–Efficacy Theoretical Model: Application to cancer Screening Programs among
Medically Underserved Men. Master’s Thesis, The Pennsylvania State University, University Park, PA, USA, 2013.
16. Hors-Fraile, S.; Malwade, S.; Luna-Perejon, F.; Amaya, C.; Civit, A.; Schneider, F.; Bamidis, P.; Syed-Abdul, S.; Li, Y.-C.; de Vries,
H. Opening the Black Box: Explaining the Process of Basing a Health Recommender System on the I-Change Behavioral Change
Model. IEEE Access 2019, 7, 176525–176540. [CrossRef]
17. Russell, A.; Abidi, S.R.; Abidi, S.S.R. Shared decision making: Using theories and technology to engage the patient in their health
journey. Stud. Health Technol. Inform. 2014, 205, 303–307. [PubMed]
18. Schulz, D.N.; Kremers, S.P.J.; Vandelanotte, C.; van Adrichem, M.J.G.; Schneider, F.; Candel, M.J.J.M.; de Vries, H. Effects of a
web-based tailored multiple-lifestyle intervention for adults: A two-year randomized controlled trial comparing sequential and
simultaneous delivery modes. J. Med. Internet Res. 2014, 16, e26. [CrossRef] [PubMed]
19. Cheung, K.L.; Hors-Fraile, S.; de Vries, H. Chapter 8—How to use the Integrated-Change Model to design digital health programs.
In Digital Health; Syed-Abdul, S., Zhu, X., Fernandez-Luque, L., Eds.; Elsevier: Amsterdam, The Netherlands, 2021; pp. 143–157,
ISBN 978-0-12-820077-3.
20. Turner-Stokes, L.; Ashford, S.; Esquenazi, A.; Wissel, J.; Ward, A.B.; Francisco, G.; Lains, J.; Suputtitada, A.; Serrano, S.; Baguley,
I.J.; et al. A comprehensive person-centered approach to adult spastic paresis: A consensus-based framework. Eur. J. Phys. Rehabil.
Med. 2018, 54, 605–617. [PubMed]
21. Bowman, J.; Mogensen, L.; Marsland, E.; Lannin, N. The development, content validity and inter-rater reliability of the
SMART-Goal Evaluation Method: A standardised method for evaluating clinical goals. Aust. Occup. Ther. J. 2015, 62, 420–427.
[CrossRef] [PubMed]
22. Tolchin, B.; Baslet, G.; Martino, S.; Suzuki, J.; Blumenfeld, H.; Hirsch, L.J.; Altalib, H.; Dworetzky, B.A. Motivational Interviewing
Techniques to Improve Psychotherapy Adherence and Outcomes for Patients With Psychogenic Nonepileptic Seizures. J.
Neuropsychiatry Clin. Neurosci. 2020, 32, 125–131. [CrossRef] [PubMed]
23. Konstantinidis, E.I.; Petsani, D.; Conti, G.; Billis, A.; Conotter, V.; Chican, G.; Lorenzo, T.L.; Rodríguez, A.R.; Hors-Fraile, S.;
Beristain, A. A new approach for ageing at home: The CAPTAIN system. In MedInfo; IOS Press: Amsterdam, The Netherlands,
2019; pp. 1704–1705.
24. Beristain Iraola, A.; Álvarez Sánchez, R.; Petsani, D.; Hors-Fraile, S.; Bamidis, P.; Konstantinidis, E. Virtual coaching for older
adults at home using SMART goal supported behavior change. In Companion Publication of the 2020 International Conference on
Multimodal Interaction; Association for Computing Machinery: New York, NY, USA, 2020; pp. 251–255.
25. De Vries, H. An Integrated Approach for Understanding Health Behavior; The I-Change Model as an Example. Psychol. Behav.
Sci. Int. J. 2017, 2. [CrossRef]
26. Fielding, R.T. Architectural Styles and the Design of Network-Based Software Architectures. Ph.D. Thesis, University of California,
Irvine, CA, USA, 2000.
27. Fields, R.D. Making memories stick. Sci. Am. 2005, 292, 75–81. [CrossRef] [PubMed]
28. Krogerus, M.; Tschäppeler, R. The Decision Book: Fifty Models for Strategic Thinking; Profile Books: London, UK, 2011;
ISBN 978-1-84668-395-4.

Ijerph 18 06868

Uploaded by

Copyright:

Available Formats

Ijerph 18 06868

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ijerph 18 06868

Uploaded by

Copyright:

Available Formats

International Journal of

2. CAPTAIN Virtual Coaching Ecosystem

2.3. Older Adult Perspective

3.2. SMART Goals and SMART Goal Plugins

3.3. Dynamic Coach Planning of SMART Goal Plans

Figure 7. Proposed older adult coaching framework.

The VCS works in three timescales—week, day and momentary—using intervention

3.3.1. Weekly Plan

3.3.2. Day Plan Generation

3.3.3. Momentary Planning

3.4. Coach Analytics

4. Coaching Intervention Tool Details

4.1. Core Coaching Intervention Tools

Figure 10. Coffee review progress, overall performance.

Then, if the participant wants to see more details, he can rev

Int. J.J. Environ.

(a)4.1.2. Motivational Day Agenda (b)

Figure 16. Mock-up for motivational day agenda.

During the gameplay, the participant follows the instructions pr

4.2.1. Gamified Trainer: Physical and Cognitive

You might also like