They used the color of the photographs, number of people in the photos, level of engagement, and happiness rated by Mechanical Turk to determine depression.
"Data collection was crowdsourced using Amazon’s Mechanical Turk (MTurk) crowdwork platform. Separate surveys were created for depressed and healthy individuals. In the depressed survey, participants were invited to complete a survey that involved passing a series of inclusion criteria, responding to a standardized clinical depression survey, answering questions related to demographics and history of depression, and sharing social media history. We used the CES-D (Center for Epidemiologic Studies Depression Scale) questionnaire to screen participant depression levels [26]. CES-D assessment quality has been demonstrated as on-par with other depression inventories, including the Beck Depression Inventory and the Kellner Symptom Questionnaire [27, 28]. Healthy participants were screened to ensure no history of depression and active Instagram use. See Additional file 1 for actual survey text."
So it sounds like they found depressed and healthy subjects via MTurk and verified them using surveys. I guess that's slightly more rigorous than just "people on MTurk thought the person in this photo looks depressed"
I'm curious how ML will do that, not how but accuracy would be cool to make (train) the model. Also with thumbnails that capture the viewer's attention.
The other jobs were brute searching sites in hopes to improve SEO.
Transcribing photos of receipts (funny those ads, some poor schmuck being paid 12 cents per 10 min job).
Oh I remember an awkward one it was some social site's user photo uploads, I had to tag images with babies. So weird to see so many 'public' yet seemingly intimate photos. Not nude, just random people's photos.
I did this for a whole week as much as I could do sanely, made $55 not worth it. Washing plates makes more money.
I'm curious on how this works. Do you basically search for specific keywords and click on specific URLs in order to signal Google that the URL is more relevant?
Yeah you had to search some term and find the link. It seemed to work as they would be beyond the first page result usually or at least not #1 but say #4 on page 1.
But it could be biased you know based on your own search history I often check externally with say Tor to see how sites rank from other viewpoints than my own.
Anyway hope that is helpful. You could also just sign up and look around. I think I still have a 99.99% satisfaction rate a lot of jobs you need something close to that to apply but it is shit work though. Some go up to $1.00+ woweee but those take like half an hour to do it's ridiculous.
How many comments vs shares does a content post have?
Comments - disagree with it, want to correct something, controversy
Shares - agree with it, want to spread it more widely, echo chamber
Also, the fact that you wrote n:n with the colon in between would make it by definition a ratio.