Game Pitch- Heroes of Niflheim (a mobile game).pdf
Unintended Emotional effects of Online Health communities.pptx
1. Unintended Emotional effects of Online
Health communities: A text Mining –
supported Empirical Study – Jiaqi Zhou
et al. MISQ 2023
- Summary by Manish Kumar, EFPM10011
2. Need of the study – 3-4 points
• Online Health Communities (OHCs) facilitate cross-region healthcare knowledge
transfer[1] and enable patients to help each other[2]. Involvement in OHCs
improves patients’ health[3]
• The paper studies impact of patients’ involvement in OHCs on their emotional
status which is an important part of their health, in particular, mental diseases
• Existing studies[4] have covered the emotional impact on the targeted audience
alone
• The paper studies the impact on non-targeted audience & the emotional impact
of auxiliary content (general informational support, ads etc.)
[1] Mein Goh et al., 2016 [2] Lou et al., 2018 [3] Yan & Tan, 2014, 2017
[4] Yan & Tan, 2014, Yoo et al. 2014, Chen et al. 2019, Huang et al. 2019
3. Context & Relevance
Source: Unintended Emotional effects of Online Health communities: A text Mining
– supported Empirical Study, Jiaqi Zhou et al. MISQ 2023
• Basis Appraisal theory[5] & Emotional contagion
theory[6]
• The same content can derive different emotional
consequence in targeted and unintended recipients
leading unexpected outcomes such as social
comparison
• The emotional state can be influenced by stimuli in
content (emotional support +auxiliary content)
• The paper is an empirical study to identify
unintended emotional effects thru a deep
learning model for emotional support
differentiation
[5] Lazarus, 1991 [6] Elfenbein, 2014
4. Context & Relevance
Source: Unintended Emotional effects of Online Health communities: A text Mining
– supported Empirical Study, Jiaqi Zhou et al. MISQ 2023
• Data used in the study from ‘Douban’ – China’s
popular social media platform for liberal topics.
User base of 62 million
• Studied the Interest group (Public) - Major
depressive Disorders (MDD), for depression
diagnosed
• 3565 threads, 47,247 posts generated by 5013
users
[5] Lazarus, 1991 [6] Elfenbein, 2014
5. Research objective
An empirical study to identify unintended motional support at Daily-
level model & Post-level model
Discussion 1 : The need for emotional support differentiation
Discussion 2 : The proper intervention strategy in OHCs
6. Hypothesis
H1: The sentiment of the emotional support content targeting other support
seekers will negatively influence the emotion of the support seeker (reflected by
expression sentiment).
H2: The sentiment of the auxiliary content will positively influence the emotion of
the targeted support seeker (reflected by expression sentiment).
H3-0: The sentiment of the auxiliary content targeting other support seekers will
NOT influence the emotion of a support seeker (reflected by expression sentiment).
H3-1: The sentiment of the auxiliary content targeting other support seekers will
negatively influence the emotion of a support seeker (reflected by expression
sentiment).
7. Methodology
A. Preprocessing: Emotional Support
Differentiation
a. To support the study, a deep learning model was
built to differentiate between:
• Posts’ targets & content type (emotional support or
auxiliary content)
b. Three models used for capturing three information
vectors –
1. Bidirectional long short-term memory(BiLSTM) model to
process initial post of the thread to capture semantics
2. Hierarchical Attention Networks (HAN) model for reply
relationship of a post by combining focal post, the post
replied to by the focal post, and the first post reply to
the focal post into a tuple
3. HAN model to process sequential relationship of a post
by combining focal post with the post before & after it
into a tuple
Source: Unintended Emotional effects of Online Health communities: A text Mining
– supported Empirical Study, Jiaqi Zhou et al. MISQ 2023
8. Methodology
• ML model training – 901 posts from gold standard
members were coded by 3 MBA coders. Fliess’s kappa
of .750 indicated substantial agreement.
• The model was compared with other baseline
methods such as SVM, BiLTSM, GCN etc. thru 10-fold
cross validation for AUC, Accuracy, Precision & recall
B. Emotional impact assessment –
Econometrical analysis
Source: Unintended Emotional effects of Online Health communities: A text Mining
– supported Empirical Study, Jiaqi Zhou et al. MISQ 2023
where SENTIMENTi,t-1 is the support seeker’s posting sentiment on day t-1, which accounts for the serial correlation of the
variable.
Xi,t-1 represents the vector of independent variables on t-1 and
Zi,t-1 denotes the vector of control variables, which will be elaborated later.
Φt accounts for the time-variant effect, such as economic, social, and environmental factors,
ηi accounts for the heterogeneity of support seekers. εi,t denotes the random noise that cannot be explained by this model.
9. Methodology
C. Testing Hypothesis through regression results
i. Activity Level of support seekers vs. Other Users
ii. Sentiment Level of Emotional Support vs. Auxiliary Content
10. Findings
• Negative impact of off-target emotional support on the emotional
state of focal users (H1 supported)
• Impact of auxiliary content is positive & significant (H2 supported)
• Off-target auxiliary content does not have a significant effect (H3-0
supported)
11. Contribution – Theoretical
• The paper enriches the understanding of OHCs in terms of the within-
community emotional influence
• Moreover, if emotional support and auxiliary content are not
differentiated, it will only be possible to observe the effect of on-
target content (as shown in previous research), with the negative
effect of off-target content remaining unobserved.
• Missing such information can lead to a misunderstanding of the
mechanisms of OHCs and could even have fatal consequences for
users.
12. Contribution – Managerial
• The findings shed light on how to best use OHCs to enhance the
mental health conditions of patients and suggest potential OHC-
based intervention strategies.
• Combining advanced machine learning models with a deeper
understanding of OHCs offers invaluable insights that could
significantly improve patient welfare.