=Paper=
{{Paper
|id=Vol-3276/SSS-22_FinalPaper_26
|storemode=property
|title=Well-being Data Origination Using MROCs with
Variable Quest: A Case Analysis of Gloom during COVID-19 Pandemic
|pdfUrl=https://ceur-ws.org/Vol-3276/SSS-22_FinalPaper_26.pdf
|volume=Vol-3276
|authors=Teruaki Hayashi, Yumiko Nagoh, Kai
Ishikawa, Hirohiko Ito, Kenichiro Tsuda, Yukio Ohsawa
|dblpUrl=https://dblp.org/rec/conf/aaaiss/HayashiNIITO22
}}
==Well-being Data Origination Using MROCs with
Variable Quest: A Case Analysis of Gloom during COVID-19 Pandemic==
https://ceur-ws.org/Vol-3276/SSS-22_FinalPaper_26.pdf
Well-being Data Origination Using MROCs with Variable Quest:
A Case Analysis of Gloom during COVID-19 Pandemic
Teruaki Hayashi*1, Yumiko Nagoh1, Kai Ishikawa2,
Hirohiko Ito2, Kenichiro Tsuda2, Yukio Ohsawa1
1
The University of Tokyo, 2NEC Corporation
*hayashi@sys.t.u-tokyo.ac.jp
Abstract requests, imaginary parts, and events yet to be converted
Technologies using artificial intelligence (AI) have been im- into data, or, the potential data of unobserved events. In this
plemented as services to solve various social problems. How- study, we used marketing research online communities
ever, the contributions of AI to people’s mentality and un- (MROCs) and variable quest (VQ) to identify unexplored
known/unobserved events have not been extensively dis- data on the gloom brought on by the COVID-19 pandemic,
cussed. In this study, we focus on people’s mental changes
caused by the COVID-19 pandemic and discuss the origin of while discussing the origin of data sources for well-being.
data sources for well-being using marketing research online
communities (MROCs) and variable quest (VQ). In the ex- Methods
periment, we selected 15 females aged between 20 and 40
who were interested in exploring how daily life has changed MROCs are research methods for extracting consumer in-
since the emergence of COVID-19 using MROCs. The anal- sights by building a closed online community of people who
ysis results by VQ revealed that the variable sets of the events are knowledgeable about or interested in specific themes
differed with the situations, mental states, and attitudes, while
(Baldus, 2015). Unlike an ordinary questionnaire survey,
not being featured in any of the MROC topics as keywords.
The result suggests that abstracting the features of unob- this method obtains more substantial opinions by using the
served events as variable sets, can help us acquire infor- communication between participants to shed light on the re-
mation potentially contributing to unexplored data discovery lationships between their issues and the themes. VQ is a sys-
for human well-being from texts not containing any infor- tem for estimating variables that serve as compositional el-
mation related to the data. ements of data (Hayashi, 2017). Variables are data attribute
sets, and one of the main compositional elements of the data.
Introduction The advantage of using VQ in this study is that the sets of
AI technology can provide effective and automated solu- variables relevant to the event in question can be accessed
tions for issues that manifest in the physical world. However, from data-related collective intelligence by providing a text
AI’s contribution to people’s mentality and unknown/unob- outline of an event that has not yet been noticed, or observed,
served events has not been sufficiently investigated. Various even when keywords related to variables are not included.
industries worldwide have been severely affected by the We selected 15 females aged between 20 and 40, who
coronavirus since the end of 2019, revealing gaps between were interested in exploring changes in daily life since the
social systems and enforcing major transformations in our emergence of COVID-19 and received 314 comments re-
lives. Data consisting of accumulated records of past events garding seven topics with MROCs. To understand how fam-
may not be effective in calculating infection risk or prevent- ilies, friends, and workplaces have changed, we chose the
ing wider infections. Moreover, not much is understood re- comments of the three topics as follows:
garding the emotional issues that accompany changes in ・ Topic 2: Feelings about lifestyle changes due to the
daily life. COVID-19 pandemic (66 comments)
This research considers accomplishing well-being data ・ Topic 5: How cohabiting with relatives, colleagues, and
origination of unexplored data. Data origination is the act of friends has changed (30 comments)
data design/acquisition/utilization that reflects the subjec- ・ Topic 6: How feelings have changed after spending half
tive knowledge and diversity of perspectives of humans and a year living with COVID-19 (25 comments)
aims to elucidate as well as support this process (Hayashi,
2020). Unexplored data signify a source of data containing
___________________________________
In T. Kido, K. Takadama (Eds.), Proceedings of the AAAI 2022 Spring Symposium
“How Fair is Fair? Achieving Wellbeing AI”, Stanford University, Palo Alto, California,
USA, March 21–23, 2022. Copyright © 2022 for this paper by its authors. Use permitted
under Creative Commons License Attribution 4.0 International (CC BY 4.0).
45
Table 1: Example comments and variable sets for each topic
# Example comments Example variable sets
・ While most companies allowed working from home, it was scary to go into office, “Attitudes about health,” “work-related atti-
and I felt different from everyone else. […] I was haunted by this constant anxiety,tude,” “attitudes about money,” “occupa-
Topic
so even on my days off, I couldn’t get much emotional respite. tion,” “number of steps,” “age,” “number of
2
・ I avoided behaviors that seemed to harm my physical condition […] and became cases,” “number of deaths,” “number of sui-
sensitive to changes in my body. cides,” “prefecture,” …
・ Even if you’re asking about something small, putting it into writing in an email or “Annual salary,” “sales,” “labor hours,” “af-
chat takes a surprisingly long time. filiation,” “email address,” “email sender,”
Topic
・ More contact via email and phone, and more pressure to respond from family, “occupation,” “number of steps,” “age,”
5
friends, colleagues—everyone in general. Maybe it’s feelings of unease, going out “number of cases,” “number of deaths,”
less, and spending more time at home. “number of suicides,” “prefecture,” …
・ Even in my daily life, I’m thinking about the future, a career change, studying; it “Occupation type,” “author name,” “know-
feels like everything is looking toward the future. ing about an event in advance,” “purchase
Topic
history,” “birthday,” “attitudes about
6 ・ I used to choose where to go based loosely on hyped-up places and popular events,
but since we’ve begun living with the coronavirus, whenever I need to run an errand, health,” “work-related attitude,” “attitudes
I actively go to places with fewer people. about money,” …
The experiment revealed that the groups of variables ex-
Results and Discussion tracted as unexplored data differed among the three topics.
Topic 2 revealed many negative lifestyle changes during the Participants often made comments in Topics 2 and 6 that
COVID-19 pandemic; in particular, many participants re- centered on themselves, thereby reflecting changes in their
ferred to changes in their physical condition, daily habits, mental states. Therefore, such topics can be characterized by
and the shift to remote working (Table 1). This suggests that having many variables that originate from psychological
it is important to capture these variables when observing changes, including attitudes. Meanwhile, for Topic 5, par-
how work habits have changed, including in remote working. ticipants often mentioned changes in other people; conse-
Moreover, a variable group was considered important for quently, easily quantifiable variables such as “annual salary,”
grasping changes in attitudes toward work and lifestyles. “sales,” and “labor hours” appeared more than variables
For example, “attitudes about health,” “work-related atti- stemming from mental states.
tudes,” and “attitudes about money” were found to be a par- Moreover, the number of variables obtained by entering
ticularly important variable for grasping and understanding the comments for each topic into the VQ were 24, 25, and
the state of mental health. Furthermore, “occupation,” 27, respectively. However, the number of cases for all three
“number of steps,” and “age” can provide an idea of the con- topics in which these variables were included as words in
text of insufficient exercise as a result of remote working. the set of comments collected by the MROCs was zero. In
Next, “number of cases,” “number of deaths,” “number of other words, the keywords representing the variables were
suicides,” and “prefecture” form a group of variables that not included in any of the words spoken by the subjects.
reflect news about the alleged rise in suicide and daily in- These variables are from past data such as the “Sense of Cri-
fection rates during the pandemic. These variables provide sis Database,” “Suicide Statistics Data,” or “Event and
more quantifiable data than the other variables. Shopping Mall Sales Data” and are thus difficult to apply
Topic 5 deals with variables about changes in participants’ directly on ongoing events such as the COVID-19 pandemic.
personal circles and shared the same variables as remote Nonetheless, decomposing the past data in variable units
working and news in Topic 2. Compared with Topic 2, the and reconstructing them using VQ could contribute to unex-
variables for Topic 5 did not concern psychological issues, plored data discovery, as a set of variables from texts not
instead centering on events occurring in the physical world, containing any information related to the data.
such as “annual salary,” “sales,” and “labor hours”.
Topic 6 discussed the changes since the first declaration Acknowledgments
of emergency in Japan in April 2020. There were still many This study was supported by NEC Corporation. We would
comments regarding the discomfort regarding work and like to thank PLUG-Inc. in the survey design of MROCs.
daily life; however, contrary to Topic 2, there were many References
positive statements about accepting societal changes, begin- Baldus, B.J. 2015. Insight Generation with Marketing Research
ning new activities, and gradually adjusting feelings. Ac- Online Communities (MROCs). J. Internet Commer., 14:476–491.
cordingly, such variables appeared as “occupation type,” Hayashi, T., Ohsawa, Y. 2017. Matrix-based Method for Inferring
owing to career changes, “author name,” “knowing about an Variable Labels Using Outlines of Data in Data Jackets. PAKDD.
event in advance,” and “purchase history” from new hobbies Hayashi, T., Ohsawa, Y. 2020. Data Origination: Human-centered
and activities. Approach for Design, Acquisition, and Utilization of Data, Int’l
Conf. Soft Computing and Pattern Recognit.
46