Our objectives:
-Explore the mental
health conversation on
Twitter
-Analyze
the literature on mental health and
Twitter
-Get
familiar with how Twitter data can be
used
-Raise
awareness of the related ethical issues
1. Using Twitter as
a community for
researcher mental
health
Stéphanie Gauttier, PhD; Grenoble Ecole de Management; France – ReMO Vice-Chair
Darragh Mc Cashin, PhD; Dublin City University; Ireland – ReMO WG3 Leader
Fatma Guneri, PhD; Lille Catholic University; France
Alexandra Lugova; Grenoble Ecole de Management; France
3. Our objectives today
Explore the mental health conversation on
Twitter
Analyze the literature on mental health and
Twitter
Get familiar with how Twitter data can be
used
Raise awareness of the related ethical issues
5. There are many communities online!
Which ones do you know? Have a look!
6. Why should we care about Twitter?
Amenable to a range of study (text-mining,
qualitative analysis, Big Data) and campaigning
(hashtag promotion, networking, impact metrics)
Over 313 million monthly users – open access,
free, anonymity available
Evidence of therapeutic benefits of Twitter via the
provision of support & information in communities;
the potential for self-management strategies;
combatting stigma; raising mental health awareness
(Berry et al., 2017).
High profile examples of relevance to ReMO –
academic mental health hashtags,
@AcademicChatter etc. and temporal effects
surrounding key events (McClellan et al., 2017)
7. Why should we care about Twitter?
Amenable to a range of study (text-mining,
qualitative analysis, Big Data) and campaigning
(hashtag promotion, networking, impact metrics)
Into the future, evidence suggest an important role
for social media–based peer support to guide
information seekers to helpful content and local
resources, but also to combat key barriers. However,
challenges exist regarding: appropriate
measurement; understanding Twitter functionality
(Saha et al., 2019) and avoiding common pitfalls.
10. #hashtags from other tweets with your keywords
#hashtags from the pages of related to your topic communities
special #hashtags tracking websites and apps
How to find relevant #hashtags ?
12. Official Twitter API – Academic Research Access
Free
Special request using online form
Up to 10 million Tweets per month
Access to the full-archive of tweets
Detailed tutorials on the official website + ready-to-use codes
13. Understanding what happens in context
Focusing our query on mental health and research
Selecting based on location to be able to understand the context
36 867 unique tweets
28 954 unique users
27 329 tweets from USA
7 067 - Great Britain
1 412 - Australia
1 092 - South Africa
14. Sentiment polarity analysis
Define which tweets are positive, negative and neutral
Assign sentiment polarity scores (from -1 to 1) to reflect the
intensity of a sentiment
Understand the « mood » of discussion
Discover negative and positive sub-topics
Find factors influencing people’s opinion about your topic
positive
negative
neutral
16. Single-word analysis
• Define interesting words and evaluate
influence of their presence in the text on its
sentiment polarity
17. N-grams
Constant word combinations that appear repeatedly in texts
• Learn more about sub-topics raised during discussion
• Find the most discussed ones in negative and positive ways
19. Topic modeling
Latent Dirichlet Allocation model (LDA)
• Uncover hidden structure in the collection of texts
• Define main topics figuring throughout the data
• Estimate each topic presence in each tweet
• Compare topics in terms of sentiment polarity
20. My current PhD supervisors are now supporting me through
my diagnostics. With my mental health and workload. Also
make effort to take into account my needs as someone with
ADHD. Especially regarding feedback. Mentor me with
publications. Allow me space to discuss my life.
My PhD advisor literally pulled this on me. He told me on the
phone “I don’t need to see it again before you submit”… so I
submitted. He emailed the editor to put a hold on it because
he didn’t approve it yet (he did). It was goddamn
embarrassing, stressful. and I’m still mad 😬
Interestingly, SA is also a country where mentions of
excellence came back strongly and where Tweets are
more negative than in other places
Topic focus (supervision)
21. Case study 2 – Analysing Academic Mental Health
Twitter Communities
22. Channel Number of Tweets (sum) Number of Words
Academic Chatter 365 969
Happy Researchers 49 154
Open Academics 116 398
PhD_Genie 99 308
PhDVoice 154 418
CACTUS Mental Health Initiative 23 78
Mental Health Europe 39 120
PhDExhausted 6 28
PhD_Mindfulness 50 315
Thephdstory 69 171
DragonflyMH 26 258
NetworkSmarten 20 20
Phdfriendsana 119 347
PhD Tweets 81 309
Vitae_news 10 30
24. Academic Chatter - Interpretation
The tweets come out from Academic Chatter are about the topics such as
mentalh health pf Phd ( connected to tweets from Phd chat, Phd Voice; Phd Life
and Monday motivation); resilience&leaving academia paradoxes, tweets from
Academic Twitter (connected to the tweets from academic jobs and phd chat);
tools and methods (Python, econometrics, javascript, webdev); intersectionality
(Women who code; gender, activists) and finally education&training (course
design, learning management).
28. Implications how to use twitter to drive your community
Study which accounts to tag and hashtag to
follow
Interact with other accounts
Identify topics of interest to the community,
and try out new ones!
Understand which campaigns work or not
29. A call for more qualitative studies
What does it mean to experience mental
health issues during one’s research journey?
What are shared perspectives on
mental health issues in academia?
What are the case-studies we should
investigate? A cohort? Volunteers?
30. The ethics behind such research
Data available publicly online vs private and
sensitive character of this data
Ethics Committee Approval
Protecting the « unvoluntary »
participants in the research
31. The ReMO network is here to help!
Short Term Scientific Missions
ReMO Ambassadors
Provide local knowledge to create
case-studies and allow data
interpretation