CATE: An Embodied Conversational Agent for the Elderly

Sean Latrelle Bravo, Cedric Jose Herrera, Edward Carlo Valdez, Klint John Poliquit, Jennifer Ureta,

Jocelynn Cu, Judith Azcarraga and Joanna Pauline Rivera

College of Computer Studies, De La Salle University, Manila, Philippines

Keywords:

Embodied Conversational Agent, Elderly Companionship.

Abstract:

Social isolation and loneliness affect one-third to one-half of the elderly population, and these have a very

negative impact on the physical and mental health of the elderly. Cate is an Embodied Conversational Agent

designed to alleviate social isolation and loneliness of the elderly. It takes in user inputs through buttons and

responds through text-to-speech and emotion animations developed with the use of SmartBody. Interviews

were conducted to identify what the elders like to talk about and how elders interact in general. An end-user

evaluation form and an observation checklist were used to see what the elders thought of the application. Cate

achieved an overall satisfaction score of 3.98 on a scale of 1 (lowest) to 5 (highest). Based on the results,

elders generally like interacting with Cate and can see Cate as a possible companion.

1 INTRODUCTION

Over the past two decades, the number of elderly peo-

ple in the Philippines grew to 9.4 million constituting

8.6% of the population (ABS-CBN, 2018). Elderly

people in the Philippines are traditionally cared by the

family members or relatives. However, as many Fil-

ipinos try to ﬁnd work abroad in an attempt to have a

better paying job, their collective effort to care for the

elderly dwindles (Antonio, 2015). Nowadays, tradi-

tional families are separated by distance, time, and/or

lack of understanding (Jones, 2017).

One-third to one-half of the elderly population are

affected by social isolation and loneliness. These have

negative impact on the physical and mental health of

the elderly. Social isolation have been identiﬁed as a

risk factor for morbidity and mortality with outcomes

comparable to smoking, obesity, and hypertension.

It has also been linked to decrease in resistance to

infection, cognitive decline, and mental health con-

ditions such as depression and dementia (Landeiro

et al., 2017).

To reduce social isolation and loneliness, two

main types of interventions were identiﬁed by Lan-

deiro et al (2017): group-based interventions (e.g.,

support groups, reminiscence therapy, and videocon-

ferencing) and one-to-one interventions (e.g., com-

puter use training, animal companionship, and vis-

itor volunteers). These interventions can either be

technology-assisted or not (Landeiro et al., 2017).

There are already numerous assistive technologies

for elderly. Some of these are CareZone for medica-

tion (Carezone, 2018), and Elderly Care for motiva-

tion, inspiration, pharmaceutical store locating, and

socializing (AB, 2017). Among these, only a few

deals with the emotional state of elderly.

Embodied Conversational Agents (ECA) are con-

versational agents which can be in a form of a ma-

chine or a software application, and has a virtual rep-

resentation of a human (Bevacqua et al., 2007). These

can be designed to provide support and counseling to

a person. There are several ECAs that are used mainly

for conversing and interacting. Some of these are El-

lie (DeVault et al., 2014), Rea (Bickmore and Cas-

sell, 2005), Greta (Niewiadomski et al., 2009), and

Justin/Justina (Kenny et al., 2008). However, very

few ECAs cater to elderly people despite its potential

as a virtual companion.

A conversational agent can be of help in alleviat-

ing social isolation and loneliness of elderly in three

ways (Ring et al., 2013). First, to reduce their loneli-

ness, an agent could serve as a companion and social

support by engaging them to small talk and/or to so-

cial activities such as games (Bickmore and Cassell,

2005). Second, to reduce their social isolation, an

agent can connect them to their friends and relatives

through electronic communication, visit and chat co-

ordination, and interventions in their social behavior.

Third, an agent can perform a talk therapy (Colby,

1995).

Bravo, S., Herrera, C., Valdez, E., Poliquit, K., Ureta, J., Cu, J., Azcarraga, J. and Rivera, J.

CATE: An Embodied Conversational Agent for the Elderly.

DOI: 10.5220/0009174009410948

In Proceedings of the 12th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2020) - Volume 2, pages 941-948

ISBN: 978-989-758-395-7; ISSN: 2184-433X

941

This work focuses on developing a virtual com-

panion for elderly people through the use of ECA to

help alleviate their social isolation and loneliness. In

this paper, we discuss the dialogue design of Cate

which includes the dialogue types and the conversa-

tion topics derived from the interviews with the el-

derly. The other activities in the system are also de-

scribed. Lastly, we show our evaluation metrics and

analysis of results.

2 RELATED WORK

There are a few studies that focuses on providing so-

cial support for older adults using agents.

A preliminary study was conducted by Mival et

al (2004). They made use of AIBO, a robotic dog,

as a companion for elder adults during a chess game.

They have concluded that the agent has to be proac-

tive in starting a conversation. Users ﬁnd an agent

that initiates a conversation more desirable. Another

conclusion is that utility is more important to the el-

derly than micro level details of the agent like physi-

cal characteristics, behaviours, and personality. Once

its usefulness is established, users get engaged to the

interaction and will start to notice the micro level de-

tails (Mival et al., 2004).

Another preliminary study was conducted by Var-

doulakis et al (2012). They deployed an Embodied

Conversational Agent (ECA) in a home for the elderly

to be tested by 12 elderly. The ECA can be controlled

remotely by the research assistants. The satisfaction

rate of their system is 6 on a scale of 1 (lowest) to 7

(highest). From the interactions, they have identiﬁed

conversation topics (e.g. storytelling, family, friends,

music, news, fashion, wheather, activity planning,

and attitude towards aging) and general design princi-

ples for an ECA for the elderly. The main limitation

of their system is the in-home video recording that

made several participants very uncomfortable. The

in-home video recording was necessary since the re-

search assistants need to see and hear the participants

for them to respond appropriately. Thus, their rec-

ommendation is to create an autonomous agent that

would not need a support of a human (Vardoulakis

et al., 2012).

Following the aforementioned studies is the work

of Ring et al (2013). They developed an ECA to pro-

vide automated social support for elderly. The user

selects responses from multiple choices provided by

the ECA. Then, the ECA responds by nodding its

head, speaking, and/or gesturing. Its dialogue design

has two components: (a) companionship and social

support which assesses user’s affect state then pro-

vides empathetic feedback, and (b) loneliness and de-

pressive symptom interventions which provides mo-

tivation and inspiration. The ECA was evaluated by

12 elderly. It has a satisfaction rate of 4.4 on a scale

of 1 (lowest) to 7 (highest) and an ease-of-use rate of

1.9 on a scale of 1 (highest) to 7 (lowest). Their re-

sults suggest that their system is effective in assessing

and managing user’s affect state, and that making the

agent proactive in starting a conversation highly in-

creased the effectiveness of the system. However, the

agent being a social support needs to be improved by

facilitating social connectivity (Ring et al., 2013).

3 CATE: CONVERSATIONAL

AGENT FOR THE ELDERLY

Cate is an ECA that can converse with elderly users.

She is embedded in a mobile application that is avail-

able on smartphones running Android 5.0 and above.

The user can communicate with Cate through but-

tons and text ﬁelds that are displayed on the screen.

Cate, in turn, responds through SmartBody’s text-to-

speech module. Depending on the choices of the user,

the kind of dialogue is selected from the dialogue tem-

plates, and the appropriate emotion (i.e. happy, con-

cerned, or worried) is expressed.

Cate’s age range from late twenties to early thir-

ties. Her upper body is shown. Her eyes, eyebrows,

mouth, neck, and hair are used to show Cate em-

pathizing with the user. SmartBody, a character an-

imation platform, is used for the facial animations. A

woman’s voice, who is ﬂuent in the English language,

is used as the voice of Cate. Both the animation and

the speech are synchronized when Cate’s response is

delivered.

3.1 Dialogue Design

An interview was conducted at a house for the elderly.

There were a total of three interviewees. During the

interview with the elderly, they talked about family at

some point. However, it seemed like a sensitive topic

for them as some of them were wondering why they

were left there by their family. This is the reason why

family is not one of the preset topics, which are topics

always suggested by Cate when the user wants to talk

about something in the application. When the elderly

were asked what they usually talk about, all of them

said they liked talking about what they watched on the

television in the house.

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

942

3.1.1 Dialogue Types

Cate’s responses are based on four dialogue types:

immediate feedback (i.e. positive, negative, and neu-

tral), pumping, suggestion, and summarizing. These

dialogue types are used as templates for the genera-

tion of responses. Cate can provide positive, neutral,

and negative feedback. She can also ask the user for

more information (e.g. “What else?” or “What more

can you tell me?”). She can also suggest some activ-

ities found in Section 3.2. Since the users’ inputs are

mostly limited to buttons, dialogue types are usually

pre-determined when a user makes a choice. An ex-

ample of a dialogue with Cate can be seen at Figure

Figure 1: A sample draft of conversation with Cate.

3.1.2 Conversation Topics

Cate has ten conversation topics: getting to know,

feeling, jokes, food, reading, television, pets, cat or

dog pictures, family, and social media. The ﬁrst topic

is getting to know. The preset topics are food and

reading, while the random topics are feeling, jokes,

television, pets, cat or dog pictures, family, and social

media.

a. Getting to Know

Cate will ask for the user’s ﬁrst and last name, birth-

day, gender, and hobby. Then, Cate will start a con-

versation by asking the user how they currently feel.

The conversation topic will then be changed to feel-

ings. Afterwards, Cate will ask if the user wants to

continue talking. If the user agrees, Cate will suggest

a different conversation topic, or an activity.

b. Feeling

Whenever Cate asks how the user is feeling, there

are always 3 choices which will lead the conversa-

tion to three different paths: happy, sad, and angry.

The happy path is where Cate encourages the user to

continue whatever made them happy as shown in Fig-

ure 1. The sad path is where Cate tries to sympathize

with the user, and tries to make them happy. The an-

gry path is where Cate will try to calm down the user

ﬁrst, then will ask them if they want to talk about it.

Cate will not pry as it might add more fuel to the ﬁre,

but will try her best to make them calm down and talk

about the issue at hand.

c. Jokes

A good humor or joke may ease the struggles that they

are going through in life. It may also help them forget

about their sadness (Flores, 2018). Cate can tell them

a joke to hopefully uplift their mood.

d. Food

The topics for food are about two of the famous fast-

food chains in the Philippines. One is a long-time

establishment, while the other is a relatively modern

fast-food restaurant. These two restaurants represent

the past and the present. Talking about the past gives

the elderly people nostalgia making them feel better

(Gergov and Stoyanova, 2013).

e. Reading

Scientiﬁc studies show that reading helps with en-

hancing memory and sharpening decision skills. It

can also help reduce an elderly person’s stress to the

point that they sleep better than usual (HomeInstead,

nd).

Cate will start by asking what the user usually

reads (i.e. books, magazines, or newspapers), then

she will ask about the title and genre. If they choose

books, she will also ask if they like ﬁction or non-

ﬁction. Cate will also share her preferences to the

user. Afterwards, she will recommend the user to

read more and will motivate them by citing beneﬁts

of reading.

f. Television

The volunteers in the participating house for the el-

derly said that elderly love watching television (TV).

Most of the time, the random topics they talk about

with one another include what they recently watched

on TV (Vardoulakis et al., 2012).

Cate will ask about the title of their favorite show,

their purpose of watching TV (i.e. information, or en-

tertainment), and their favorite TV channel. If they

answered that they watch TV for information, they

will be asked if they usually watch news or TV series.

On the other hand, if they answered entertainment,

they will be asked if they usually watch sports or TV

series. Afterwards, Cate will ask how long they usu-

ally watch TV and will cite effects of watching TV.

g. Pets

Studies have shown that having pets have poten-

tial health improvements, both physical and men-

CATE: An Embodied Conversational Agent for the Elderly

943

tal health, for the elderly (Cherniack and Cherniack,

2014).

Cate will ask about the user’s pet, the name of the

pet, and their sentiment about their pet. She then asks

how they got the pet. If the user says it was given a

family or friend, Cate will suggest to the user get in

touch, and thank them through phone.

h. Cat or Dog Pictures

Looking at animal pictures tends to make people

happy due to their cuteness (Nittono et al., 2012).

Thus, Cate will sometimes, show pictures of cute

dogs and cats which are two of the most common do-

mesticated pets.

i. Family

Elderly always talk about how their families take care

of them and love them (Vardoulakis et al., 2012). For

this topic, Cate will ask if the user talked to their fam-

ily recently. If they do, she will ask about what they

talked about and how they talked (i.e. thru phone or

in person). Otherwise, she will encourage the user to

talk to their family, and suggest the user to call them,

or set a reminder when to call them.

j. Social Media

Social media strengthens ties between friends and

family by making them engage with one another

(Cornejo et al., 2013). When the users pick the so-

cial media topic, Cate will ask them if they use social

media. If they do, Cate will ask them what social me-

dia platform they use and for what purpose do they

use it for (i.e. social interaction, or news and events

updates). If they use social media to interact with peo-

ple, Cate will ask who they usually talk to (i.e. family

or friends), how often, and will they meet soon. She

will then suggest to the user to set a reminder to talk

to their family or friends.

3.2 Activities

Cate features three different activities, aside from

conversations, for the elderly users to help alleviate

their social isolation and loneliness, or to assist them.

3.2.1 Calls

Cate could suggest to make a call for the user. For

example, at the end of the pets topic, the user is given

a choice whether to call their family or friend who

gifted them the pet. Another instance of this is in the

getting to know path when the user picks that some-

one made them happy because of a gift. Figure 2

shows that Cate is suggesting to make a call as one

of the other activities. Figure 3 shows that Cate is

suggesting some contacts for the user to call to thank

them through the phone.

Figure 2: Cate suggesting other activities.

Figure 3: Cate listing people to contact.

3.2.2 Reminder (Calendar)

Cate could suggest to the user to set a reminder. The

application will open the calendar application of the

phone, and the user will conﬁrm the contents of the re-

minder. Once set, the Calendar application will close

and will go back to Cate. For example, the user can

set a reminder at the end of the family topic. If the

user says that they are seeing their family soon, Cate

asks if they want to set a reminder to when this meet-

ing will happen. Same as the call, another instance is

when the user is happy because someone gave them

a gift. Figure 2 shows that Cate is suggesting to set a

reminder as one of the activities. Figure 4 shows that

Cate opened the calendar application of the phone for

the user.

3.2.3 Games

Cate will sometimes ask the user if they would want

to do something else. If they do, Cate could suggest to

play a game. There are currently two types of simple

games in the application: trivia and picture game.

a. Trivia

Trivia is a quiz-like game where Cate will ask the user

some questions about history because it is one of the

most exciting trivia game topics for elderly (Super-

Carers, nd). Sample questions are: ”Who waved the

ﬂag during the Philippines independence proclama-

tion in 1898?”, ”Who coined the name of the longest-

running noontime show ”Eat Bulaga!”?”, and ”Is

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

944

Figure 4: Setting a reminder.

Manny Pacquiao’s nickname Pac-man based on a

Japanese arcade game of the same name?”.

b. Picture Game

The picture game shows the user some landmarks,

buildings, and churches of the Philippines. It may

give a sense of nostalgia since the landmarks and pho-

tos used were from the past, helping them reminisce

about the good old days. Some of the sample pho-

tos used in the picture game are the old photos of the

Malaca

nang Palace, Manila City Hall, and Araneta

Coliseum (FilipiKnow, 2018).

3.3 Knowledge Base

In Cate’s knowlege base, since the dialogues are

template-based, there are three major stored data:

user proﬁle, questions, and responses. The user pro-

ﬁle contains the user’s ﬁrst and last name, birthday,

gender, and hobby. The name is asked so that Cate can

keep mentioning their name when having their con-

versation to build a sense of companionship. Ques-

tions are what Cate can ask to the user. Responses are

Cate’s reactions and suggestions regarding the user’s

answer. It contains the identiﬁer of the current deci-

sion, a string of decision identiﬁers that will conse-

quently display the next question/response/dialogue,

the choices that the user will be having next or the

text-ﬁeld that the user will be answering, and the tag

of the emotion (i.e. happy, concerned, or worried) that

Cate will be executing.

4 METHOD

4.1 Participants

Ten participants aged 60-77 years old were chosen to

test the application. Four of them were from a retire-

ment home and the other four were living with their

family. The elderly from the retirement home were

either abandoned by their family or were placed their

by the government’s social welfare

For some of the elderly who had difﬁculties with

understanding English, especially those from the re-

tirement home, some of the conversations and evalua-

tion questions had to be translated by the researchers

to the local language. Some of the participants also

had to be assisted because they were technologically

challenged.

4.2 Privacy and Ethical Issues

Since Cate was evaluated by human participants, an

informed consent form was given to them. For those

who have disabilities, they are assisted by a family

member, or the volunteers in the participating house

for the elderly. The elderly were informed that their

participation is voluntary, and that they may choose

to withdraw anytime without any consequences. The

system, testing procedure, potential risks and discom-

forts, potential beneﬁts to subject or society, and con-

ﬁdentiality were discussed in the informed consent

form. There might be a potential discomfort in deal-

ing with an ECA since they are not really human, but

only has avatar as its human representation. There

are no potential risks or harm that may be posed to

the participants. The information obtained during the

study regarding the user was kept conﬁdential and

their identity was kept anonymous.

4.3 Evaluation

Cate was evaluated by the participants by answering

an evaluation form which is split into 7 categories:

naturalness, embodiment, interaction and affect, joy

of use, ease of use, acceptability, and utility. For each

category, the participants are asked to answer a set of

questions using a Likert scale of 1 (strongly disagree)

to 5 (strongly agree).

The Naturalness category is used to evaluate how

human-like the virtual agent is when interacting with

the elderly. Embodiment deals with evaluating the ex-

ternal appearance and personality of the virtual agent.

It is also used to measure how well the agent is able

to convey itself as likeable through its facial expres-

sion, speech and gestures. Interaction and affect deals

CATE: An Embodied Conversational Agent for the Elderly

945

with how the elderly feel when they interact with Cate

while using the application. It is used to evaluate how

emphatic and comforting the agent is. Joy-of-use is a

subcategory under the interaction and affect caetgory.

It refers to the positive feeling the user has when us-

ing the application. Ease-of-use is used to evaluate

how much mental effort is required from the user to

do a certain task and how user friendly it is. This is

important because we have to consider the cognitive

and physical abilities of the elderly. Acceptability is

used to determine if the user likes using the applica-

tion and if they would recommend it to other people.

Utility is used to evaluate the functionality of the sys-

tem related to the task that the user wants to accom-

plish. The agent must be capable of meeting the end

goals of the user.

The researchers also answered an observation

checklist while the users were talking to Cate, in or-

der to get more insights and data to support the scores

and answers in the end user evaluation.

5 RESULTS AND ANALYSIS

5.1 Overall Results

Table 1: This table shows the average score for each cate-

gory in the end-user evaulation form.

Category Mean

Naturalness 3.8

Embodiment 3.73

Interaction and Affect 4.18

Joy-of-use 3.99

Ease-of-use 4.03

Acceptability 3.68

Utility 4.29

Total Mean 3.98

Table 1 shows the average results for each category.

Utility had the highest score with an average of 4.29

while the lowest score was for acceptability. Un-

der the naturalnes category, the lowest scoring criteria

was the “I can sense emotions through Cate’s voice”,

which is under the naturalness category. It had an av-

erage score of 3 which is neutral. However, this is

caused by a limitation of the system on how Cate pro-

nounces certain words. This limitation made it difﬁ-

cult to determine the emotion in her speech for some

people. Some also found the voice too robotic for

their liking. “The conversation with Cate is natural”

and “Talking to Cate is similar to/feels like talking to

a human being” both had the highest average score of

4.3. The elderly said that the conversations felt natu-

ral and that they loved talking to Cate because of how

she would keep asking them questions. For embodi-

ment, the criteria with the highest average score was

“I like the way Cate looks” with an average score of

4.2. Based on the observation checklist and the eval-

uation scores, some were amazed by how human-like

she looked while some thought she looked like a car-

toon character. One of the participants initially re-

marked that Cate looked snobbish but she eventually

like talking to her after a while, after seeing how she

looked and acted throughout their conversation.

Interaction and affect had the second highest av-

erage score. The highest average scores under this

category were from the criteria “Cate was friendly”

and “Interaction with Cate is not repetitive”. Over-

all, most of the elderly liked talking to Cate. Although

some of them did not like how she looked, they liked

interacting with her. One elderly said that they were

amazed by how she kept talking to them while an-

other even began to verbally talk to Cate because she

was so immersed in the conversation, that she forgot

that Cate does not accept voice input. There was also

a notable instance with one participant wherein she

was asked by Cate about her ﬁrst crush, ﬁrst date and

ﬁrst kiss and she laughed and smiled and said to the

researchers: “I forgot already, but it is nice to know

that she is curious about me”. The lowest scores en-

countered were for the criteria “The random topics

that Cate offered me were interesting” and “The jokes

that Cate shared were funny”. The scores were 3.7

and 3.6 respectively. The reason for the low scores for

the random topics could be because they got a topic

they did not like . However, for some of them they re-

ally enjoyed the topics that were randomly assigned to

them. One of the elderly participants whose favorite

hobby was reading, was amazed when she saw read-

ing as a topic and when Cate talked about it with her.

For the jokes, the low scores could be due to Cate’s

voice. An essential part of telling the joke is delivery,

and Cate’s robot-like voice is not able to handle this

well.

The category ease of use had a high score. The

only problem that the participants had, especially

those from the retirement home, was that they were

not that knowledgeable about smartphones and appli-

cations. One of the participants had to be assisted

by the researcher in using the application. Most of

the participants also could understand what Cate was

saying despite some not being very ﬂuent in English.

However, one participant had a difﬁcult time because

they were hard of hearing while another forgot to

bring his reading glasses so he could not see prop-

erly. These problems however were minimized be-

cause Cate provide both text and audio versions of

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

946

the conversations.

Acceptability had the lowest score among all cat-

egories. Based on the observation checklist, they en-

joyed talking to Cate but they still prefer conversa-

tions with actual humans. One participant even com-

mented that human emotion can never be replace with

robotics. Another participant said that they would

only use Cate again if there was someone who could

assist them, this is because the participant had an ill-

ness that limited her body movement and had bad eye-

sight.

For Utility, the elderly who were able to test the

call and reminder functionalities were amazed that

she could do that and agreed that it was very helpful

to have these features.

5.2 Comparison between the Two

Demographics

Figure 5 shows side by side the average scores for

each category for both demographics. For the partic-

ipants from the retirement home, utility and interac-

tion had the highest scores. Based on the observation

checklist, they found the call function a helpful fea-

ture. They also said that the reminder function would

be help them keep track of events such as taking their

medicine. They mostly gave high scores for their in-

teraction with Cate because they liked talking to her.

Figure 5: Comparison of average scores for each category

for both demographics.

For the elderly living with their family, the highest

score was for joy-of-use which falls under the inter-

action and affect category. Based on the observation

checklist, they found Cate interesting and amusing.

They also found it easier to use the application.

The participants from the retirement home gener-

ally gave higher scores than the participants who lived

with their families. This could be due to the difference

in their living conditions. Those that live with their

families get to talk to them every single day while

those at the retirement home often do not meet their

families at all. Another reason could be due to the

difference in their exposure and familiarity to tech-

nology. The elderly living with their family are more

familiar with smartphones and gadgets compared to

those living in the retirement home. This could also

explain the large difference in scores for the embodi-

ment category.

6 FUTURE WORK

This research is an ongoing work on developing a

conversational agent that can help alleviate the so-

cial isolation and loneliness of elderly people by pro-

viding companionship to them. Based from the re-

sults, the elderly generally liked talking to Cate and

liked the application. However, although they accept

Cate as a companion, it is not to the point that they

would use Cate every day due to certain limitations

in the system and the fact that they would still pre-

fer the companionship of another human over a ma-

chine. Nonetheless, Embodied Conversational Agent

(ECA) like Cate has a potential to alleviate the so-

cial isolation and loneliness of elderly. It is impor-

tant to focus on the activities (e.g. games, setting a

reminder, reminding to call family and friends) ﬁrst

since it catches the attention of the elderly.

There are still a lot of improvements to be made

as the features of Cate right now are still very basic.

The dialogues must be improved since they are too

static due to the limited templates used. The speech

recognition and natural language processing has to be

improved in order to make the conversation more nat-

ural for the elders because right now their answers are

restricted to the options presented in the buttons. We

also plan to add voice input and facial recognition so

that we can also add other processing techniques like

emotion recognition and adjust the conversations and

the facial expressions of Cate accordingly. We also

plan to look at other models aside from those provided

by SmartBody to address the concerns regarding the

robotic voice of Cate and her appearance. Since we

mentioned that some of the participants also had a

difﬁculty in understanding English, it would be also

worth exploring the use of the local language instead.

For the evaluation, since one of the focus is combat-

ing loneliness, use of the UCLA Loneliness Scale, the

most widely used self-report measure of loneliness,

will be explored.

CATE: An Embodied Conversational Agent for the Elderly

947

REFERENCES

AB, S. (2017). Elderly care. Retrieved from

https://play.google.com/store/apps/details?id=

com.sci.elderlycare.

ABS-CBN (2018). Philippines moving toward aging

population: Popcom. Retrieved from http://news.abs-

cbn.com/news/01/03/18/philippines-moving-toward-

aging-population-popcom.

Antonio, A. (2015). Challenges to the ﬁlipino elderly as

traditional caregivers: The changing landscape of long

term care management of the ﬁlipino elderly. Master’s

thesis, De La Salle University-Manila.

Bevacqua, E., Mancini, M., Niewiadomski, R., and

Pelachaud, C. (2007). An expressive eca showing

complex emotions. In The AISB Annual Convention,

pages 208–216, Newcastle, UK.

Bickmore, T. and Cassell, J. (2005). Social Dialogue

with Embodied Conversational Agents, pages 23–54.

Springer Netherlands, Dordrecht.

Carezone (2018). Carezone. Retrieved from

https://play.google.com/store/apps/details?id=

com.carezone.caredroid.careapp.medications&hl=en.

Cherniack, E. and Cherniack, A. (2014). The beneﬁt of pets

and animal-assisted therapy to the health of older indi-

viduals. Current gerontology and geriatrics research,

2014:623203.

Colby, K. M. (1995). A computer program using cognitive

therapy to treat depressed patients. Psychiatric Ser-

vices, 46(12):1223–1225. PMID: 8590102.

Cornejo, R., Tentori, M., and Favela, J. (2013). Enriching

in-person encounters through social media: A study

on family connectedness for the elderly. Int. J. Hum.-

Comput. Stud., 71(9):889–899.

DeVault, D., Artstein, R., Benn, G., Dey, T., Fast, E.,

Gainer, A., Georgila, K., Gratch, J., Hartholt, A.,

Lhommet, M., Lucas, G., Marsella, S., Morbini, F.,

Nazarian, A., Scherer, S., Stratou, G., Suri, A., Traum,

D., Wood, R., Xu, Y., Rizzo, A., and Morency, L.-P.

(2014). Simsensei kiosk: A virtual human interviewer

for healthcare decision support. In Proceedings of

the 2014 International Conference on Autonomous

Agents and Multi-agent Systems, AAMAS ’14, pages

1061–1068, Richland, SC. International Foundation

for Autonomous Agents and Multiagent Systems.

FilipiKnow (2018). Can you answer these tricky

pinoy trivia questions? Retrieved from

https://ﬁlipiknow.net/pinoy-trivia-quiz-questions-

and-answers/.

Flores, M. M. (2018). 12 ways to make a sad

friend or any person happy. Retrieved from

https://inspiringtips.com/ways-to-make-a-sad-friend-

or-any-person-happy/.

Gergov, T. and Stoyanova, S. (2013). Sentimentality and

nostalgia in elderly people: Psychometric proper-

ties of a new questionnaire. Psychological Thought,

6:358–375.

HomeInstead (n.d.). Five proven beneﬁts

of reading for seniors. Retrieved from

https://www.homeinstead.com/205/BlogSite/Pages/

Five Proven Beneﬁts of Reading for Seniors.aspx.

Jones, C. H. (2017). Combining daycare for children and

elderly people beneﬁts all generations. Retrieved from

https://theconversation.com/combining-daycare-for-

children-and-elderly-people-beneﬁts-all-generations-

70724.

Kenny, P., Parsons, T. D., Gratch, J., and Rizzo, A. A.

(2008). Evaluation of justina: A virtual patient with

ptsd. In Prendinger, H., Lester, J., and Ishizuka,

M., editors, Intelligent Virtual Agents, pages 394–408,

Berlin, Heidelberg. Springer Berlin Heidelberg.

Landeiro, F., Barrows, P., Musson, E., Gray, A., and Leal,

J. (2017). Reducing social isolation and loneliness

in older people: A systematic review protocol. BMJ

Open, 7.

Mival, O., Cringean, S., and Benyon, D. (2004). Personiﬁ-

cation technologies: developing artiﬁcial companions

for older people. In CHI 2004 conference proceedings

: Connect, CHI letters, pages 1–8. Vienna, Austria,

Association for Computing Machinery.

Niewiadomski, R., Bevacqua, E., Mancini, M., and

Pelachaud, C. (2009). Greta: An interactive expres-

sive eca system. volume 2, pages 1399–1400.

Nittono, H., Fukushima, M., Yano, A., and Moriya, H.

(2012). The power of kawaii: viewing cute images

promotes a careful behavior and narrows attentional

focus.

Ring, L., Barry, B., Totzke, K., and Bickmore, T. (2013).

Addressing loneliness and isolation in older adults:

Proactive affective agents provide better support. In

Proceedings of the 2013 Human Association Confer-

ence on Affective Computing and Intelligent Interac-

tion, ACII ’13, pages 61–66, Washington, DC, USA.

IEEE Computer Society.

SuperCarers (n.d.). Brain games for the elderly that im-

prove mental health and memory. Retrieved from

https://supercarers.com/blog/games-for-elderly/.

Vardoulakis, L. P., Ring, L., Barry, B., Sidner, C. L., and

Bickmore, T. (2012). Designing relational agents as

long term social companions for older adults. In

Proceedings of the 12th International Conference on

Intelligent Virtual Agents, IVA’12, pages 289–302,

Berlin, Heidelberg. Springer-Verlag.

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

948