Content uploaded by Eisuke Ono
Author content
All content in this area was uploaded by Eisuke Ono on Nov 18, 2015
Content may be subject to copyright.
978-1-4577-1524-2/11/$26.00 ©2011 IEEE
Relationship between Social Interaction and Mental Health
Eisuke Ono, Takayuki Nozawa, Taiki Ogata, Masanari Motohashi, Naoki Higo, Tetsuro Kobayashi,
Kunihiro Ishikawa, Koji Ara, Kazuo Yano, and Yoshihiro Miyake
Abstract— In order to explore the relationship between hu-
man social interaction and mental health, we investigated the
correlations between the amount of face-to-face contact time
and quantified mental health. Social interaction data were
obtained by wearable sensing system for two organizations
in Japan. In this study, we regarded face-to-face contact
between individuals as social interaction. The mental health
of individuals was measured by psychological questionnaire.
We found that the social behavioral property of individuals
tended to reflect quantified stress in both organizations. The
results suggest the possibility that people who experience large
amounts of social interaction tend to have less stress.
I. INTRODUCTION
Data on human social behavior are now readily avail-
able. Human behavior has been investigated statistically
and dynamically, enabled by developments in sensing and
information technologies [1]. Interevent time distributions of
human activities such as communication, trade transactions,
and library loans have been found to follow power-law [2],
[3]. Although the statistical structure of human behavior has
been partially revealed, the psychological or social meaning
of the structure is not clear yet.
In social psychology, some previous studies have inves-
tigated the relation between social relationship and psy-
chological state. It has been suggested that office layout
arrangement encourages employee satisfaction and improves
work satisfaction [4]. Social capital and mental illness were
also found to relate to each other [5]. Social interaction
and psychological state were indicated to be important for
maintaining quality of life. However, previous studies had
technical limitations to measure observable social interaction
accurately [5].
Manuscript received October 11, 2011.
E. Ono, M. Motohashi, N. Higo and Y. Miyake are with the Department
of Computational Intelligence and Systems Science, Tokyo Institute of
Technology, Yokohama, Japan.
(e-mail: ono@myk.dis.titech.ac.jp, motohashi@myk.dis.titech.ac.jp,
higo@myk.dis.titech.ac.jp, miyake@dis.titech.ac.jp)
T. Nozawa is with the Institute of Development, Aging and Cancer,
Tohoku University, Sendai, Japan. (e-mail: nozawa@idac.tohoku.ac.jp)
T. Ogata is with the Intelligent Modeling Laboratory, the University of
Tokyo, Tokyo, Japan and with the Department of Computational Intelligence
and Systems Science, Interdisciplinary Graduate School of Science and En-
gineering, Tokyo Institute of Technology. (e-mail: ogata@iml.u-tokyo.ac.jp)
T. Kobayashi is with the National Institute of Informatics, Tokyo, Japan.
(e-mail: k-tetsu@nii.ac.jp)
K. Ishikawa is with the Department of Human System Science, Graduate
School of Decision Science and Technology, Tokyo Institute of Technology,
Tokyo, Japan. (e-mail: ishikawa@hum.titech.ac.jp )
K. Ara and K. Yano are with the Central Research Labora-
tory, Hitachi, Ltd., Tokyo, Japan. (e-mail: koji.ara.he@hitachi.com,
kazuo.yano.bb@hitachi.com)
Nakamura et al. found that the resting duration distribution
of human physical activity obeyed power-law. [6], [7]. They
also found that the scaling exponent of resting duration
distribution of depressed patients was statistically lower than
that of healthy people. These findings suggest that depressed
patients tend to remain in their resting state more than healthy
people.
Nozawa et al. investigated the relationship between the
physical activity of individuals and their social interaction
[8]. In their study, social interaction was defined as the
amount of face-to-face contact time. Their results suggest the
possibility that social context of individuals can be evaluated
from their physical activity patterns. In other words, the
effect of social relationship on individual physical activity
state was indicated.
The statistical structure of human social behavior possibly
reflects our qualitative states such as mental health and
sociality. The effect of social interaction on mental health
of individuals is not clear quantitatively yet. What does
interacting with others mean for our psychological states?
In this research, we focus on the relationship between
observable social interaction and mental health. We used
two sets of data measured in Japanese organizations. Our
data are unique in that they are records of face-to-face
interaction, measured by using a wearable sensing system.
Simultaneously, we conducted a questionnaire to measure
individual mental health. We propose a hypothesis that the
amount of social interaction and mental health are related
to each other. The purpose of this research is to verify the
hypothesis by investigating the correlation between social
interaction and mental health.
II. METHODS
A. Data
We analyzed two sets of face-to-face interaction data. The
social interaction data were provided by the Central Research
Laboratory, Hitachi, Ltd., Japan. The sets of data include two
organizations in Japan. The first is a consulting firm (Org.
A). In Org. A, 136 participants were objects of analysis.
The data were measured from January 1, to February 22,
2011. The second is a nursing home (Org. B). In Org.
B, 50 participants were objects of analysis. The data were
measured from March 1, to March 31, 2011. People in Org.
B were separated into two different groups: staff and elderly
users. The data were collected anonymously and analyzed.
The summary of organization is shown in Table I.
- 246 - SI International 2011
Fig. 1. Device used in the measurement of social interaction. The
device contains an infrared sensor and detects face-to-face contact between
participants only if they are within 2 (m) and facing each other.
TABLE I
ORGANIZATION SUMMARY.THE NUMBER IN PARENTHESES IS THE
NUMBER OF ELDERLY USERS IN ORG.B.
Number of Number of Observation Category of
participants valid answers period organization
Org. A 136 57 2011/1/5- Consulting
2011/2/22 firm
Org. B 50(36) 49(35) 2011/3/1- Nursing home
2011/3/31 for the elderly
B. Social interaction
1) Measurement method: In order to measure social inter-
action as conversation events, we used the Business Micro-
scope, which is a wearable sensing system developed by the
Central Research Laboratory, Hitachi, Ltd [9]. The Business
Microscope contains an infrared sensor, and its shape is
similar to a name tag as shown in Fig. 1. Each participant
wore the name tag around the neck and placed at the chest.
The infrared sensor can detect face-to-face contact between
participants if they are within 2 (m) of one another. The
infrared sensor only senses the other within a 120∘circular
area in front of the name tag, and the system records face-
to-face contact events cumulatively when two individuals
are facing each other. We regard the amount of face-to-
face contact as a feature of observable social interaction. It
should be noted that the sensor records the face-to-face state
as one minute even if a momentary contact was detected
between participants. In other words, the time resolution of
the system is equal to 1 (min). Face-to-face interaction logs
were gathered through the whole measurement period. Then,
the amount of face-to-face contact time between every pair of
participants was obtained for the whole measurement period
in each organization.
2) Analysis method: Nozawa et al. investigated social
interaction in an office setting [8]. They analyzed social in-
teraction using the social network analysis method. Inspired
by their previous study, we regard the amount of social
interaction between participants as connection weight in the
Fig. 2. Example of the calculation of social interaction characteristics
social network. For instance, if the total face-to-face contact
time between 𝑖and 𝑗is 100 (min), the connection weight
between 𝑖and 𝑗is 100, denoted as 𝑓𝑖𝑗 in (1). However,
it was problematic to use face-to-face contact time as a
weighted edge because the sensor-wearing time depended
on the participants. To overcome individual differences, a
normalized matrix 𝐺was adopted as an adjacency matrix
for analysis.
𝐺=[𝑔𝑖𝑗]=[(𝑓𝑖𝑗
𝑢𝑖)] (1)
This matrix 𝐺was obtained by dividing each row of
𝐹=[𝑓𝑖𝑗 ]with 𝑢𝑖, where 𝑓𝑖𝑗 represents the total face-to-
face contact time between participants 𝑖and 𝑗through the
whole measurement period and 𝑢𝑖represents the total sensor-
wearing time of participant 𝑖.
We calculated in-degree centrality 𝐶𝑖𝑛 and out-degree
centrality 𝐶𝑜𝑢𝑡 to characterize the quality of each participant
in the social network based on the method of [8]. In-degree
centrality 𝐶𝑖𝑛 was calculated using (2).
𝐶𝑖𝑛(𝑖;𝐺)=∑
𝑗
𝑔𝑗𝑖 =∑
𝑗
𝑓𝑗𝑖
𝑢𝑗
(2)
𝐶𝑖𝑛 quantifies the individual property that a person has been
required to contact from other people. Out-degree centrality
𝐶𝑜𝑢𝑡 was calculated using (3).
𝐶𝑜𝑢𝑡(𝑖;𝐺)=∑
𝑗
𝑔𝑖𝑗 =1
𝑢𝑖∑
𝑗
𝑓𝑖𝑗 (3)
𝐶𝑜𝑢𝑡 quantifies the individual property that the person has
dedicated to social relationship. An example of calculation
of 𝐶𝑖𝑛 and 𝐶𝑜𝑢𝑡 is shown in Fig. 2. These two centralities
were the characteristics for social interaction analysis.
C. Mental health
1) Participants and measurement periods: A question-
naire to measure mental health was conducted in each
organization. In Org. A, 57 responses were obtained from
- 247 - SI International 2011
TABLE II
EXAMPLES AND SUBSCALE FOR QOL AND SCL
Psychological Scale Subscale An Example of Statements
Stress Check List (SCL) Psychological Stress Obsession Obsessed with things
Inattention Worried about carelessness
Social Stress Interpersonal Avoidance Feeling worried about meeting someone
Interpersonal Nervousness Feeling nervous to meet someone
Physical Stress Tired Feeling Feeling tired
Sleep Disorder Trouble falling asleep
Quality of Life (QOL) Life Satisfaction Satisfied with your life
Motivational Life Having dream for your future
0.1 0.2 0.3 0.4 0.5
30
40
50
60
70
80
90
Cin
SCL
R = −0.270
(p < .05)
(a) 𝐶𝑖𝑛 and SCL (N= 57, office workers in Org. A)
0.5 1.0 1.5 2.0
30
40
50
60
70
80
90
Cout
SCL
R = −0.377
(p < .05)
(b) 𝐶𝑜𝑢𝑡 and SCL (N=35, elderly users in Org. B)
Fig. 3. Scatter diagrams showing the relationship between observable social interaction and degree of stress
136 participants. The answers were measured from February
7 to February 14, 2011. In Org. B, 49 responses were
obtained from 50 participants. The responses were measured
from March 4 to March 15, 2011. In Org. B, 35 answers
were from elderly users, 14 were from staff.
2) Measurement method: In this research, the Mental
Health Pattern (MHP) scale developed by Hashimoto et al.
was adopted for the contents of questionnaire. Validity and
reliability of the MHP scale have already been proven in
their research [10], [11]. The MHP scale contains degree of
stress and degree of life satisfaction. The Stress Checklist
(SCL) scale and Quality of Life (QOL) scale were used to
measure. The SCL scale has six subscales. QOL scale has
two subscales. Examples of statements are shown in Table
II. Each subscale had five questions. The total number of
the questions was 40. During the procedure, the participant
selected from four options: strongly agree,agree,disagree
and strongly disagree. The answers were scored from four
points (strongly agree) to one point (strongly disagree).
Reversed question items were calculated by reversing the
points.
III. RESULTS
We found a statistically significant correlation between
amount of social interaction and individual mental health.
Pearson’s correlation coefficients are presented in Table III.
Notably, a significant negative correlation between 𝐶𝑖𝑛 and
SCL was observed in Org. A (p<.05). In addition, a
significant negative correlation between 𝐶𝑜𝑢𝑡 and SCL was
observed in the cluster of elderly users in Org. B (p<.05).
Slight positive correlations between social interaction and
QOL were observed for both organizations.
Two diagrams and each correlation coefficient R are
presented for precise observation of the correlations. Scatter
diagrams describing the relation between social interaction
and degree of stress are shown in Fig. 3. First, the result
of Org. A is presented. The x-axis is 𝐶𝑖𝑛 and the y-axis is
SCL, as shown in Fig. 3(a). A negative correlation tendency
is observed. Second, the result of the cluster of elderly users
in Org. B is presented. The x-axis is 𝐶𝑜𝑢𝑡 and the y-axis is
SCL, as shown in Fig. 3(b). A negative correlation tendency
is observed. Regression lines were drawn in each figure to
aid understanding.
- 248 - SI International 2011
TABLE III
CORRELATION COEFFICIENTS BETWEEN SOCIAL INTERACTION AND
MENTAL HEALTH
Organization Mental Scale Cin Cout
Org A SCL *-0.270 -0.069
QOL 0.060 0.067
Org B(staff) SCL -0.089 0.184
QOL 0.291 0.108
Org B(elderly users) SCL -0.058 *-0.377
QOL 0.124 0.129
∗:p<.05
IV. DISCUSSION
The results shows that the amount of social interaction is
correlated to individual mental health. In particular, we found
that the degree of stress (SCL) possibly reflects participants’
behavioral property. These results support the hypothesis that
there is a correlation between social interaction and mental
health.
However, the degree of the correlation coefficient depends
on the organization or cluster as shown in Table III. The
correlation coefficient of the elderly users cluster in Org. B
was larger than that of the staff cluster. In addition, the calcu-
lated centrality values of Org. B were higher than the those
of Org. A. We assume that centrality value differences were
caused by the organization’s characteristics and shortness of
the measurement period.
The results indicate that social supports were possibly
exchanged through face-to-face social interaction. Social
support means supports as help, advice that is exchanged
through social relationships. Social support has been consid-
ered to be a buffer for stress generated in social life [12].
Allen et al. also reported that an open-space layout in an
office setting encouraged communication and improved ease
of communication [4]. The results in this research support
the previous studies in social psychology.
Findings in this research are new because we investigated
the relationship between social relation and psychological
state quantitatively. The results suggest that not only physical
activity, but also individual mental health possibly reflects the
amount of social interaction. Nozawa et al. found significant
positive correlations between the degree of physical resting
state and network centralities, suggesting the effect of social
relationship on individual physical activity [8]. Our results
added psychological meaning on their previous study.
In this research, we did not consider individual attribute
such as sex, age, and position. The experiment design also
should be modified. There is still a possibility that the results
occurred because of other causes such as home problems.
Deeper investigations must be done in the future work.
V. CONCLUSION
In order to explore the relationship between observable
social interaction and psychological state, we have conducted
a correlation analysis of the amount of face-to-face contact
time and quantified mental health. Significant negative corre-
lations between the amount of social interaction and mental
health were observed. These results suggest that people who
interact with others relatively tended to have less stress.
There has not been a study of how the amount of observ-
able social interaction is mutually related to psychological
states. We have first partially clarified the relationship.
VI. ACKNOWLEDGMENTS
This work was supported by the Homo Contribuence
Research and Development Institute.
REFERENCES
[1] C. Cattuto, W. Van den Broeck, A. Barrat, V. Colizza, J.-F. Pinton,
and A. Vespignani, “Dynamics of Person-to-Person Interactions from
Distributed RFID Sensor Networks,” PLoS ONE, 5, e11596, 2010.
[2] A. L. Barabasi, “The Origin of Bursts and Heavy Tails in Human
Dynamics,” Nature (London), vol. 435, pp. 207-211, 2005.
[3] D. Rybski, S.V. Buldyrev, S. Havlin, F. Liljeros, and H. A. Makse,
“Scaling Laws of Human Activity Patterns on the Dynamics of
Information Diffusion,” Proceedings of National Academy of Science
U.S.A., vol. 106, 12640, 2009.
[4] T. J. Allen, P. G. Gerstberger, “A Field Experiment to Improve
Communications in a Product Engineering Department: the Non-
territorial Office,” The Journal of the Human Factors and Ergonomics
Society, Vol. 15, Number 5, pp. 487-498, 1973.
[5] M. J. D. Silva, K. McKenzie, T. Harpham, S. R. A. Huttly, “Social
capital and mental illness: a systematic review,” Journal of Epidemiol
Community Health , vol. 59, pp. 619-627, 2005.
[6] T. Nakamura, K. Kiyono, K. Yoshiuchi, R. Nakahara, Z. R. Struzik, Y.
Yamamoto, “Universal scaling law in human behavioral organization,”
Physical Review, Letter, vol. 99, 138103-1-4, 2007.
[7] T. Nakamura, K. Kiyono, K. Yoshiuchi, R. Nakahara, Z. R. Struzik,
Y. Yamamoto, “Of mice and men-universality and break down of
behavioral organization,” PLoS One, vol. 3, e2050-1-8, 2008.
[8] T. Nozawa, N. Higo, T. Ogata, K. Ara, K. Yano, Y. Miyake, “Rela-
tionship between characteristics of individual physical activity pattern
and social interaction,” 11th SICE System Integration Division Annual
Conference, 2G2-3, 2010.
[9] R. Otsuka, K. Yano, N. Sato, “An organization topographic map for
visualizing business hierarchical relationships,” IEEE Pacific Visual-
ization Symposium 2009 (PacificVis ’09), April 20-23, Beijing, China,
pp. 25-32, 2009.
[10] K. Hashimoto, M. Tokunaga, “Development of a diagnostic inventory
for Mental Health Pattern (MHP): reliability and validity of the MHP
Scale,” Japan Health Society, vol. 21, pp. 53-62, 1999.
[11] K. Hashimoto, M. Tokunaga, S. Takayanagi, “A study on character-
istics of Mental Health Pattern,” Journal of Health Science, Kyushu
University, vol. 16, pp. 49-56, 1994.
[12] T. C. Antonucci, J. E. Lansford, K. J. Ajrouch, Social support,
Encyclopedia of stress, San Diego, CA: Academic Press, pp. 479-482,
2000.
- 249 - SI International 2011
