Emotional Resiliency of Families Dealing with Autism in

Social Media

Amit Saha

and Nitin Agarwal

Center for Distance Health, University of Arkansas Medical Science, Little Rock, Arkansas, U.S.A.

University of Arkansas at Little Rock, Arkansas, U.S.A.

Keywords: Autism, Stress, Twitter, Blogger, Community, Social Support, Health 2.0, Social Media.

Abstract: Nowadays online social media is used extensively by families dealing with various health issues, such as

autism, diabetes, obesity, etc., to share experiences with other members of the community. The interaction

between members of health community can be systematically analyzed to build a knowledge base for others

who are dealing with the same health conditions. In this study, we analyze one such health community, i.e.,

the autism community and evaluate stress dispersed among the community members using social network

analysis along with sentiment analysis methodology. We found that the autism blogger community provides

nominal stress during the interaction with other community members. Differences across various classified

groups like autistic bloggers, mother bloggers with autistic kids, father bloggers with autistic kids, and autism

support group blogs in different social media platforms (blogs and Twitter) were analyzed in context of stress.

Families dealing with autism have a better quality of life with reduced stress by interacting with fellow autism

community members in social media.

1 INTRODUCTION

Social media has provided users an open platform for

discussions, communication, and information

exchange on various health related topics. Families

dealing with health issues like autism use social

media almost daily to share their experiences with

others. The information exchanged by the interactions

of individuals dealing with same health problems are

archived by default and has become an immense

source of knowledge for others dealing with the same

situation.

Approximately 1 in 68 children in the USA are

diagnosed with Autism Spectrum Disorder (ASD) as

estimated by Centers for Disease Control and

Prevention (CDC, 2014). Governments and non-

profit groups advocating autism awareness

encourages autism community members to share their

experiences freely in social media and get advice

from others. Jordan in her noteworthy study found the

benefits of the Internet technology in spreading the

education and awareness of autism using Internet

technologies (Jordan, 2010).

Shared experiences by an individual dealing with

autism in the social media platform, especially blogs,

Twitter and Facebook shed light on various issues of

autism. As a moto to generate an autism awareness

premier non-profit organizations like Autism Speaks

(www.autismspeaks.org) recognizes top autism

bloggers based on feedback from families dealing

with autism. For families dealing with autism, the

shared know-how about autism helps them to lead a

better life.

Stress can be defined as the non-specific response

of the body to any demand for change (Selye, 1936).

Stress arises when individuals perceive that they

cannot adequately cope with the demands being made

on or with threats to their well-being (Larzelere and

Glenn, 2008).

The purpose of the current study is to provide a

research-based understanding of the conversations in

social media platforms among families dealing with

autism and to shed light on community characteristics

of autism toward its members. This study is aimed to

analyze the topics of discussion among the members

of the autism blogging community and deduce

whether the families, individuals, or caregivers

dealing with autism can utilize this information to

enhance the quality of life. In this study, sentiment is

quantitatively analyzed from the conversations of the

autism community on blogs and Twitter to understand

how autism blogger community engages with stress.

Saha, A. and Agarwal, N.

Emotional Resiliency of Families Dealing with Autism in Social Media.

DOI: 10.5220/0005774703770382

In Proceedings of the 9th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2016) - Volume 5: HEALTHINF, pages 377-382

ISBN: 978-989-758-170-0

377

The autism community members promote

disseminating positive and upbeat messages to

counter stress often experienced by other members of

the community on blogs and Twitter.

This paper is organized as follows: the prior

related works are described in Section 2. Section 3

describes our proposed model to quantify stress in

social media interaction. Section 4 depicts the

methodology and data collection. Section 5 shows the

result. Section 6 discusses about the inferences drawn

from the study. We draw conclusions and possible

future works in Section 7.

2 RELATED WORK

Many clinical studies have been conducted to get an

in-depth knowledge of ASD. These studies provide

understanding of the cause, issues and effectiveness

of different therapy on autism. Clinical trial option

being sluggish and costly, the use of social media

content in research analysis to assess the different

intervention mechanism for autism could be an

economically viable option. Our study does not

intend to provide a substitute for clinical tests of the

intervention strategies. On the contrary, our

methodology would provide a justification to build a

knowledge base of the intervention strategies or the

therapies from a receiver’s perspective, which would

help prioritizing resources on the testing procedures

of intervention strategies for dealing with different

health issues. In this study, however, we address a

tiny part of this bigger research agenda, which is, does

health communities like autism blogger community

provide a feeling of solidarity to other community

members? Are there differences in stress mitigation

across various characteristics of autism bloggers (i.e.,

autistic bloggers, mother bloggers with autistic kids,

father bloggers with autistic kids, and autism support

group blogs) and in different social media platforms

(e.g., blogs and Twitter)?. Answers to these questions

will help conduct a more systematic evaluation of

interactions occurring on various online platforms,

especially the social media, and thereby helping us

evaluate the efficacy of a knowledge base constructed

using social media based interactions.

Sociologists and health scientists have published

profusely on the stress and health/wellbeing concepts.

The link between stress and health is addressed by the

buffering effect hypothesis. In the buffering effect

hypothesis, social support enhances good health by

reducing the impact of stressful life events (Wallace,

2005). Stress is linked to all leading physical causes

of death - cancer, and stroke (Cohen et al., 2007).

3 STRESS: PROPOSED MODEL

Social science literature lacks a formal definition of

stress. Selye defined stress as ‘the non-specific

response of the body to any demand for change’

(Selye, 1936). We leverage various empirical

definitions of stress available in social science along

with computational science literature that overlap

with the healthcare domain. Stress in an interaction

between individuals can be approximately deduced

by collectable statistics when influential factors

groups can be evaluated.

3.1 Model Parameters

Below we examine factors that influence stress

assessment and objectively measure these factors by

collectable statistics from social media interactions.

 Personal Concern: American Psychological

Association listed many causes of stress that

include fear and uncertainty in personal domain

like relationship conflicts and major life changes

(APA, 2000). Measure of personal concern can be

estimated using sentiment analysis methods.

 Anger: Studies indicate anger suppression as a

significant factor in perceived stress within thin

the sample of adults of America and Japan

(Yamaguchi, 2015). Correlation study among

cancer patients found high degree positive

correlation among anger-out and anger-in (Lee et

al., 2005). Anger expression can be measures by

rating in the affective process of anger.

 Anxiety: Study by Vallee found positive

correlation in adults between anxieties and stress

(Vallée et al., 1999). Study shows academic stress

would show a significant positive correlation with

anxiety (Mishra and McKean, 2000). Expression

of anxiety within a text can be measures by rating

in the negative affective process of anxiety.

These three influence factors with the corresponding

statistics collectable using sentimental analysis

methodology are summarized in Table 1.

Table 1: Factors influencing stress assessment and their

corresponding collectable statistics.

Influence

Factors

Example

InfluenceWeight Notation

Personal

Concern

job, cash, owe

 

Anger hate, kill, annoyed

 

Anxiety Worried, fearful

 

HEALTHINF 2016 - 9th International Conference on Health Informatics

378

Our model uses sentiment analysis mechanism to

get an approximation numerical value of stress. Stress

in an interaction can be interpreted in terms of an

influence by the three factors (or parameters):

; 

and . Hence, stress measure denoted by S in an

interaction is given by,

Stress (S) = w

 + w

 + w



(1)

where w

, w

and w

are the weights that can be

determines the influence of the factors on stress. To

determine the degree to which the factors influence

assessment of stress, we consider the Google distance

approach. Google distance uses Google search

association between two pair of concepts (Cilibrasi

and Vitanyi, 2004). Using the google similarity

distance algorithm, corresponding values of w

, w

and w

were evaluated. The result of the Google

similarity measure is shown in Table 1 in the

influence weight column. Therefore, the

mathematical formula used in the study to determine

stress (S) in an interaction is evaluated using the

Equation 2

Stress (S) = 0.40  + 0.17  + 0.43 

(2)

4 METHODOLOGY AND DATA

COLLECTION

This study evaluates stress through the interactions

among members of the autism blogger community

using sentiment analysis and social network analysis

concepts. The social network analysis features are

used to assess the structural networking aspect of the

propagation of stress within the autism community.

Our methodology consists of collecting data from

autism bloggers in different social media platform

(blogs and Twitter). Data is then processed and

filtered for noise. Topic and word analysis were

performed to ensure that the subject of discussion is

autism. For each social media platform, the network

for the autism community is constructed to deduce

stress propagation. Later analysis of sentiment for the

content of interaction was performed. Lastly, the

degree of stress propagated in the interaction was

calculated and analyzed.

Presently there are more than a thousand active

autism bloggers on the Internet. As an initial phase of

the study, top 40 autism bloggers based on the

recommended list of popular bloggers by the Autism

Speaks organization were selected. The content of the

blogs by the selected autism bloggers was extracted

and analyzed. Further, we cross-referenced their

blogger profile and Twitter profile (wherever the

blogger had provided a link to his/her Twitter profile)

and collected their tweets, and other network

information, including friends and followers. We

retrieved the most recent permissible tweets (up to

3,200 each) for the 40 autism bloggers, resulting in

118,531 tweets.

Some of the tweets by autism bloggers are as

follows, “I myself am opaque, for some reason. Their

eyes cannot see me. Yes, that's it: The world is autistic

with...”, “Do not fear people with Autism, embrace

them, Do not spite people with Autism unite them, Do

not deny people.”.

Profile analysis of the bloggers led to the

classification of bloggers based on different

characteristics. Classification of autism bloggers into

different categories is done to deduce different

capacities of social support based on defined blogger

categories. Of the 40 autism bloggers, 13 were female

bloggers with autistic kids who are termed as

mothers. Male bloggers with autistic children termed

as fathers are 10 in our database. Number of bloggers

who blogged as groups to create autism awareness

termed as autism support group are 13 and rest 4

termed as self-autistic bloggers who are diagnosed

with autism and blogs for themselves.

To infer the amount of stress in text, we used

psycholinguistic analysis methodology. Linguistic

Inquiry and Word Count (LIWC) program (www.

LIWC.net) (Pennebaker, Martha and Roger, 2001)

and with part-of-speech (POS) tagging methodology

was utilized to categorized the tweets and blog

content into different psychological groups. The

amount of stress in the text content is deduced

primarily using the rating in the affective process of

anger, anxiety and personal concern, captured using

LIWC. Many researchers used LIWC for sentiment

analysis and found promising results. Study found a

consistent correlation between emotion rating values

of LIWC with self-reported score for interaction

within health community forum (Tov et al., 2013).

Topics of the tweets of the autism blogger

community are analyzed using topic modeling

methodology. We used Stanford Topic Modeling

(nlp.stanford.edu/software/tmt) for topic modelling.

5 RESULTS

Many social network analysis methodologies was

used in the study to get an understanding of autism

blogger community characteristics. For social

network analysis, the activities of the autism bloggers

like blogpost, tweets, friends, followers, and

mentions in Twitter were analyzed.

Emotional Resiliency of Families Dealing with Autism in Social Media

379

The friend and follower Twitter network of autism

blogger community is shown in Figure 1. Different

colors indicate various communities based on

network modularity. Modularity is one of the

effective function in community detection for the

compound network like autism bloggers twitter

network. This essentially means that there is an

intensive communication between the members of the

same community as compared to cross-community.

From Figure 1, we can observe that autism groups

tend to form one community and converse more

closely among themselves. Most mothers, fathers,

and autistic bloggers form another community and

converse more with fellow mothers, fathers, autisitic

bloggers. Overall analyzed metrics of the Twitter

friends and followers network is shown in Table 2.

Table 2: Overall Twitter data characteristics of the autism

blogger community network.

Number of users 874

Total Edges 2060

In-Degree 6(Max), 1.15(Average)

Out-Degree 105(Max), 1.151(Average)

Connected Components 1 with 874 Maximum Vertices

Geodesic Distance

(Diameter)

8(Max) , 3.952(Average)

Top Words in Tweet autism, austic, out

Top Hashtags in Tweet autism , specialneeds, sensory

Table 2 shows distinct characteristics of the

autism blogger network on Twitter where any

member of the community can reach a colleague on

average 4 hops (average geodesic distance), in

compared to the widely known 4.74 degrees of

separation (average geodesic distance) in Facebook

network of active users (Backstrom, 2012).

The value modularity of the network of autism

bloggers Twitter network is 0.623, which indicates

that the community is well connected. The top

hashtags ‘autism', ‘sensory’, and ‘specialneeds'

indicates the autism bloggers network is highly

focused on autism-based discussions.

Based on the author characteristics of the autism

bloggers and choice of social media platform the

sentiment in the text varies. LIWC provides the

baseline values for psychological groups for the

different style of text writing like control writing,

emotional writing and science articles along with

speech conversation communication.

Our study found that in Twitter, autism support

group desimates less stress as compared to other

autism blogger categories. The variation of stress

based on author’s characteristics is shown in Figure

2. The autism bloggers’ community in Twitter and

blogs as a whole indicates minimal stress as

compared to other health support forum like alcohol

support. Further, amount of stress shown by families

dealing with autism in our study is nominal as compa-

Figure 1: Friend and follower network of autism blogger community. The autism bloggers are annotated based on the

classification. Their real identity is anonymized. Colors indicate different communities based on the network modularity.

HEALTHINF 2016 - 9th International Conference on Health Informatics

380

red to emotional writing or novels that could indicate

autism blogger community doesn't treat autism as a

curse but as an opportunity.

Topic analysis of the Twitter data of the autism

bloggers shows education, technology, sports, health,

and science as the leading subjects of discussion.

Table 3 displays the distribution of topics discussed

in the Twitter by the autism bloggers community. The

tweet content of the autism bloggers found to be

involved in many topics related to autism, and the aim

of the autism community bloggers’ Twitter network

seems to be spreading autism awareness.

Table 3: Distribution of Topics for the Tweet content by

autism bloggers.

Topic Fathers Mothers

Autistic

Bloggers

Group

Bloggers

Education

25% 43% 56% 16%

Technology

26% 26% 16% 73%

Sports

21% 10% 4% 1%

Health

12% 5% 4% 3%

Disasters

5% 3% 1% 1%

Science

2% 2% 5% 1%

Others

9% 11% 14% 5%

6 DISCUSSION

The study sheds light on community characteristics of

online autism blogger community on the different

social media platform. The stress mitigated during

interactions with members of autism blogger

community by identifying the bloggers, and the

community members were unfolded in the study.

Our study revealed that the autism blogging

community is tightly knit within community

members. Members of the autism community relay

minimal stress in the interaction between its

community members, by providing emotional

support. Members of the autism bloggers community

in Twitter and blogs spread minimal stress as

compared to members of other health groups like

alcohol support forum. For the tweets of the mothers,

the amount of stress provided is lower than fathers or

autistic bloggers with a given amount of negative

emotion, but the ratio is highest in emotional writing

as compared to any interaction.

7 CONCLUSION AND FUTURE

WORK

In this research, we analyzed online interaction

among members of the autism blogger community in

different social media platforms. The study extracts

blogging activity of popular autism blogger and their

Twitter activity including their friends, followers,

tweets, retweets, mentions, and hashtags information.

The tightly knit interaction within the autism

blogging community was identified in our study. Our

Figure 2: Variation of stress based on different characteristics. The vertical axis shows the amount of stress and horizontal

axis represent different characteristics. Each circle represents a data point.

Emotional Resiliency of Families Dealing with Autism in Social Media

381

study revealed that Autism blogger community

members on social media, especially on Twitter,

indicates minimal stress in the interaction between its

community members. While negative sentiments are

reflected in some tweets, the minimal stress attributed

by the autism blogger community restricts the

propagation of stressful sentiments within

community members. Stress indicated in the text

content of autism bloggers varies based on blogging

characteristics and social media platform.

We envision our study will provide a mechanism

to access social interaction in online health

communities. However, the fact that autism bloggers

also use other social media platforms such as

Facebook presents a limitation in our study. The

findings of this study lay the groundwork to study our

bigger research agenda, i.e., evaluating the efficacy of

therapies for ASD as perceived by the caregivers

through the experiences they have shared in online

forums and social media. The will help build a

knowledge base for interventions and experiences,

which in turn could assist the clinical research in

better understanding of behavioural interventions for

various health disorders.

ACKNOWLEDGEMENTS

This research was funded in part by the Jerry L.

Maulden/Entergy Endowment at UALR. Any

opinions, findings, and conclusions or

recommendations expressed in this material are those

of the authors and do not necessarily reflect the views

of the funding organization. The researchers

gratefully acknowledge the support.

REFERENCES

American Psychological Association (APA), 2000.

Mind/Body Health: Stress. Orth-Gomer, K. The

Journal of the American Medical Association.

Backstrom, L., et al., 2012. Four degrees of separation, In

Proceedings of the WebSci '12, ACM, New York, NY,

USA, 33-42.

Centers for Disease Control and Prevention (CDC), 2015.

Community Report on Autism, http://www.cdc.gov/

ncbddd/autism/states/comm_report_autism_2014.pdf.

[Online; accessed 14-May-2015].

Cilibrasi, R., and Vitanyi, P., 2004. Automatic Meaning

Discovery Using Google, http://www.citebase.org/

cgibin/citations? id=oai:arXiv.org:cs/0412098.

Cohen, S., Janicki-Deverts, D., & Miller, G. E., 2007.

Psychological stress and disease. Journal of the

American Medical Association, 298, 1685–1687.

Jordan C., 2010. Evolution of autism support and

understanding via the World Wide Web, Intellect

Developmental Disabilities. 48(3):220–7.

Larzelere, M., and Glenn, J., 2008. Stress and

health. Primary Care: COP, 35.4 : 839-856.

Lee, P. S., et al., 2005. A correlational study among

perceived stress, anger expression, and depression in

cancer patients, Taehan Kanho Hakhoe Chi, 35,195–

205.

Mishra, R., and McKean, M., 2000. College Students'

Academic Stress and its Relation to their Anxiety,

AJHS, 16.1: 41-51.

Pennebaker, W., Martha F., and Roger B., 2001. Linguistic

inquiry and word count: LIWC 2001. Mahway:

Lawrence Erlbaum Associates.

Selye, H., 1936. A syndrome produced by diverse nocuous

agents, Nature, 138.3479: 32.

Tov, W., Ng, L., Lin, H., and Qiu, L., 2013. Detecting Well-

Being via Computerized Content Analysis of Brief

Diary Entries, JPA, vol 25(4), 1069-1078.

Wallace, E., 2005. Job stress, depression and work-to-

family conflict: A test of the strain and buffer

hypotheses, RI/IR: 510-539.

Yamaguchi, A., et al., 2015. Effects of anger regulation and

social anxiety on perceived stress, Health Psychology

Open July-December 2015, 2: 2005102915601583.

Vallée, M., et al., 1999. Long-term effects of prenatal stress,

European Journal of Neuroscience, 11: 2906–2916.

HEALTHINF 2016 - 9th International Conference on Health Informatics

382