Translating Local Concepts of Distress

Karan

Last week, we asked about technologies that could help narrow the global mental health gender data gap (https://community.xprize.org/discussion/754/global-mental-health-gender-data-gap).

One issue that arose in both online and focus group discussions is how the diversity of local mental health experiences could be lost when placing them into Western-based survey and diagnostic categories.

• How can we contribute to global mental health data needs without completely transforming or erasing local concepts of distress in the process?
• Is there a way to preserve capturing local experiences of distress while also building global data needs?

Shashi

Hi @Andrea, @stellunak, @sarahb, @KarenBett, @areff2000 and @trigoana - with your backgrounds in data science, we feel you may be able to advise us on ways of capturing local experiences of distress while also building global mental data needs. We would like hear your thoughts on this discussion. Thanks.

Andrea

@Shashi This is a very interesting topic and maybe beyond the scope of my work.

Shashi

@shihei, @AnnalijnUBC, @erickson, @Tapman, @bwilcher and @Lyndall -- What are your thoughts on ways of capturing local experiences of distress while also building global mental data needs. Please join the discussion to share your thoughts.

farah

HI All! My research in the area of mental health has conceptualized distress from a culturally specific perspective; I did a study of working class English women's mental health issues related to depression, amongst other studies. Most of my work, however, looks at mental well being and how to achieve it. I have published in the area of yoga and do seminars on various self-care methods that cultivate wellness such as gratitude the game changer, deep breathing, yoga nidra and other practices. It is so important that we maintain a focus on resilience, 'upstream' aspects of mental health!

stellunak

Hello everyone. I find this issue of critical importance and I am sorry I didn't have the chance to respond earlier.

Two initial thoughts on this and I would love to know more about your experience and perspectives.

1. I would say we need to tie culturally informed terms/keywords, used by individuals to describe mental health states and issues, with more formalised language structures or vice versa, in order to create pipelines of "foreign" data towards usable data categories.
2. We would also need to identify the hosting contexts where those pipelines could be really activated -those contexts shouldn't just be healthcare-related, but could also operate through educational and community settings, digital spaces etc.

In my experience from working with youth who have experienced or are currently experiencing eating disorder-related health challenges, language plays an integral part in how these challenges are formed, communicated, diagnosed and healed (or not communicated and not healed) within communities. And because we are talking about notoriously unreported challenges affecting notoriously inaccessible and undiagnosed communities (teens), the language of the everyday becomes even more important for gathering data and charting the territory -as varied as this territory maybe among different places and cultures.

Shashi

Thanks @farah and @stellunak for your sharing your thoughts.

@farah - Great to hear about your research on Mental Health. During your research have you come across any technology or model that helps capturing local experiences of distress while also building global mental health data needs. We would love to hear your perspective.

@stellunak - Thanks for your insightful comment. All strong points. Can technology help us in overcoming challenges mentioned by you? Are you aware of any models / methods used to overcome these challenges?

Shashi

@DrLiliaGiugni, @EVSwanson, @aakanksha_k, @mustafanaseem, @Tsion and @staceyo - We feel you might have some thoughts on this discussion and the comments. Thanks.

stellunak

Thanks @Shashi
No, unfortunately I don't. My work so far has centered on individuals and I haven't had the chance to work with systems and tech solutions. I will be really interested in exploring such solutions to more experienced tech professionals, both from the language and the mental health perspective.

farah

I can't think of any technologies that would help. Although right now we need communications tech to keep in touch and this has been critically important

Shashi

Thanks @farah and @stellunak.

@ingmarweber - Can we capture cultural-specific data in addition to global mental health data using the method that you'll use to map digital gender gaps or may be a modified version of it. What do you think?

@YaelNevo, @kbeegle, @Suneetharani, @qlong - You might have some thoughts on Stellunak's comment above. Join the discussion to share your thoughts.

@ssolomon - We are curious to know if you'll have done any research which takes into consideration culture-specific inputs to cure mental health.

ingmarweber

@Shashi What is easy to collect is gender-disaggregated data on where Facebook [and/or Instagram] users are (not) interested in certain topics, such as mental health. That can be done with global coverage and at the sub-national level (even sub-city for big cities). See https://sha.qcri.org/ for an exploration of this kind of data, collected three years ago. (More about the data visualization in https://www.aaai.org/ocs/index.php/ICWSM/ICWSM17/paper/view/15576 and the methodology in https://dl.acm.org/doi/10.1145/3091478.3091513.)

What is harder to do is use this advertising data on audience estimates to reason about prevalence or health outcomes as "being interested in X" is at best a weak proxy for "being affected by X". Also lots of biases, e.g. time spent on a platform. See reference above for mixed results on that.

Targeted ads on social media can also be used for recruiting otherwise hard-to-reach populations for either online surveys or clinical studies. That has been done repeatedly in various settings around the globe. See e.g. https://scholar.google.com/scholar?q=targeted+social+media+advertising+recruitment+clinical+studies.

Aaron_Denham

@stellunak This is spot on. We've been thinking about how we can consider the intersections of language, context, idioms, and diverse meanings. Some of the AI advances in this area are interesting (and they ideally will enable one to consider more than just the direct translation across cultures and linguistic groups). Your attention to language here made me think about narrative-based approaches to therapy and research--in terms of looking at how content (or a lack of it), meaning, structure, and context shape emotional worlds and can be in turn reshaped as part of the healing process. As part of this, we've been looking at some ways to leverage the strengths of natural language processing with cultural context. We'll share any good findings here.

@ingmarweber Thank you so much for this example. It's exactly what we needed to think with this coming week, particularly around how we might look at data, markers, and the potential pitfalls.

Shashi

Thanks @ingmarweber for sharing those insightful resources. We are in the process of identifying big EHR data source to further research mental health conditions.

Hi @lepri - As you have been leading the mobile and social computing lab, we feel you may have thoughts on @ingmarweber comment. Please share your thoughts.

stellunak

Aaron_Denham wrote: »

@stellunak This is spot on. We've been thinking about how we can consider the intersections of language, context, idioms, and diverse meanings. Some of the AI advances in this area are interesting (and they ideally will enable one to consider more than just the direct translation across cultures and linguistic groups). Your attention to language here made me think about narrative-based approaches to therapy and research--in terms of looking at how content (or a lack of it), meaning, structure, and context shape emotional worlds and can be in turn reshaped as part of the healing process. As part of this, we've been looking at some ways to leverage the strengths of natural language processing with cultural context. We'll share any good findings here.

Great thinking, Aaron, would love to har more about this.