« Wednesday Open Mike: Dog Days Book Report (OT) | Main | Changing the Subject (Apropos Yesterday) »

Wednesday, 24 August 2016


Possibly more helpful for depression are the reports that some/much of it is due to inflammation: http://www.bbc.co.uk/news/health-37166293

An interesting article regarding depression and the immune system, was on the BBC news today:

More likely you get a pre-packaged reply from an automated email server telling you a robot decided, from looking at an algorithm, that you are depressed. And it's from a do not reply, unattended mailbox.

No that's depressing.


First Instagram, then every photo site. This is a trifle chilling to me.

Well that just makes me sad.

This sort of thing depresses me.

From the report: "Depressed participants were less likely than healthy participants to use any filters at all." (p. 9)

Better start using those filters, folks. And no, B&W doesn't count: "When depressed participants did employ filters, they most disproportionately favored the “Inkwell” filter, which converts color photographs to black­-and­-white images."

From the abstract ...

Photos posted by depressed individuals were more likely to be bluer, grayer, and darker.
[ ... ]
Several different types of information were extracted from the collected Instagram data. We used total posts per user, per day, as a measure of user activity. We gauged community reaction by counting the number of comments and “likes” each posted photograph received. Face detection software was used to determine whether or not a photograph contained a human face, as well as count the total number of faces in each photo, as a proxy measure for participants’ social activity levels. Pixel­level averages were computed for Hue, Saturation, and Value ...


This worrisome note appeared at the end of the Significance Statement ...

 These findings demonstrate how visual
social media may be harnessed to make accurate
inferences about mental health. 

Get a blood test to check for thyroid activity first. Low thyroid activity leads to depression, among other things. Fortunately, a hormone replacement, oroxine, is readily available at a cost of pennies per day. one of the reasons that while I live in paradise, I hope there is not a compete collapse of the worldwide ponzi scheme before I lose my grip on the twig.

Cheers, Geoff

When I was leaving a partnership and serving out the six months leaving requirement, I used to slip out the back door to go to the sandwich place across the road, and I would think of Three Days of the Condor.

Like Microsoft's attempt a while back at determining a person's age from a photograph. Whereas in person most people will give my age up to 10 years younger than I am, when I tried the website for this software out of curiosity it put me 10 years over!

I really don't see the use in these things which just seem like a way for the researchers to keep their jobs.

I clicked the link to MIT Tech Review and got a blank page. Went to Review site and still didn't find the article. Do you know why??
Richard Newman

You make 'sad' and 'depressed' equivalent when you talk about authentic versus poseur -- a very photographer kind of thing to say I think. Authenticity seems to be the prime directive in photography.

Obviously sadness and depression are not equivalent. Maybe there's an overlap.

There are enormous philosophical problems with the premise that (true) states of mind can be inferred (by anyone or anything) from a piece of art. Whether it's post-transition (to digital presumably) or not.

Of course, I haven't read the article so I could be blowing smoke. But on the face of it, it just seems fanciful and deluded to me. Perhaps because treating art as data is just kind of beside the point. Isn't it?

I liked it better when we talked about important stuff, like what kind of jeans Mike wears, and his lack of sponsorship......
Now if you switched to Speedo, there might be an opening there.....

This algorithmic approach actually may be a useful adjunct diagnostic. It's worth recalling that the standard psychological evaluation tests, such as the MMPI, only tend to be about 60-65% accurate and that clinical correlation is always necessary for accurate diagnosis and treatment.

It's also worth noting that much of what the algorithm examines can be considered affirmative self-descriptives, such as the filters that are intentionally applied. That's a fair use of data that's likely more reliable than potentially self-conscious responses to a formal psychological diagnostic test.


Regarding your comment on "I also don't know how the machine could separate authentic sad photos from poseur sad photos."

The basic answer to this is that the researchers probably used a supervised learning machine algorithm. Essentially the photos are tagged by real people, and this is done at a large scale as in machine learning the machine gets better at the task with more experience..

So it's ability to detect 'authentic' vs. 'poseur' sad photos ultimately depended the response of the people who tagged the photos. thus, the machine is a reflection of the biases of the people it learned from....

That's really interesting, and raises even more interesting questions (as some commuters have already noted).

Of course, all of this leaves aside the value of creative expression in mental health.

"How's that supposed to work—you get an email from someone saying that their robot has detected that you're depressed, based on pictures you've posted online?"

It depends on which robot finds you first. They won't actually tell you you're depressed, but Google's robot will send you uplifting links to click on. Amazon's robot will send you ads for hard liquor and razor blades.

More seriously, it might work through your health insurance. You could get a break on your health care payments if you let your health care provider monitor your activity on the Internet.

Insurance companies are starting to do this with car insurance. If you agree, a small device is installed in the car that sends speed, miles driven, and other data to the insurance company. In theory you pay less for your insurance that way.

There are plenty of psychometric tools based upon the premise that emotional states as reflected in the facial expressions are accessible to others instinctively and many psychiatric disorders are characterized by impairments in this ability. However this is a cross-sectional assessment. There is this online tool that will tell you the emotional state of a person in a photograph: https://www.microsoft.com/cognitive-services/en-us/emotion-api
The instagram tool probably uses a similar algorithm.
Depression on the other hand is a pervasive, emotional disorder and as has been pointed out above is not equivalent to sadness.

This is a perfect example of bad science. There is so much assumption and subjective analysis in it that it cannot possibly have any value.

So let's rename all landscape photography to 'sad selfie photography'. (By the way, some Indian gurus wouldn't protest calling every photo a selfie by definition - a lovely idea, but another matter entirely.)

Typical silly season news, but depression is a serious thing. Too serious for this kind of distraction, I'd say. People tend to use the term 'depression' as a blanket word that applies to all kinds of sad/blue moods, but it is actually a clinical condition. And one that can have very serious consequences.
It happened to me once, after a break-up and a rather draconian shortage of work: inactivity and loneliness got me so down I had to be medicated. It took me several months to pull through. I mightn't have made it if I hadn't been properly diagnosed by a doctor. Now some bunch of nerds is trying to tell us an algorythm can replace the knowledge of an MD. Some people get so infatuated with computer technology they lose the sense of reality.
As Carl Gustav Jung would have put it, "Oh well."

Lots of people say they find my photos depressing. Wonder what that says about me?


A lot of interesting and sometimes contradictory stuff out there on inflammation and mental well being.
One study suggests that Statins reduce the risk of dementia. Another suggests that the same drugs interfere with memory. Go figure.
Makes a guy want to go cook up a steak, wash it down with a martini and pile some more dirty laundry on top of the treadmill but that would be bad.

Will the next version of Photoshop have a "sadness" or "depression" correction button similar to red-eye correction? Or maybe they should get to work on a radicalization algorithm, to find all the terrorists via their selfies.

I know it's too easy to make snide comments about something that I don't know much about. I'm weak. But history has shown that these things are usually badly reported on or oversold. Remember the story of the zen master and the little boy from the movie Charlie Wilson's war? We'll see.

I would not be surprised if there is something to this. But whatever it is, it's probably not what we think about it now.

I don't think it's about photos... I think it's about social media. There has been a wide range of research published on the links between depression and daily use of social media accounts.

One that I read (but can't find in 10 seconds of googling) indicated something like >20 minutes per day of 'online social media' = propensity to depression.

It probably stems from where one finds happiness; from our own interests and pursuits, or from trying to see what others think of us. (my $0.02 worth).

I would think that the truly depressed wouldn't post any pictures on Instagram because they felt they weren't worthy.

Just think, that research is probably far behind that which analyzes blog posts...

I'd write more but probably should spend the time brightening up my Instagram!

Cause or effect, I wonder?

I wish they'd make the data openly available. Calling the importance() function on a random forest model will give you a relative-quantified importance of each factor, for starters.

Back on topic - I seem to make some of my best photos when the mood is non-average, either a bit depressed or a bit sunny.

Also, they should try running it on 500px users...

Ok, what about those who don't have any idea of what instagram is?
Idle chatter?

Wouldn't it better to just go out and meet real people? While you can tell a lot from photographs, there's no substitute for the real thing, and a fleeting expression may give completely the wrong impression.

Is depressing a shutter button a form of depression?

[No, it's a form of release! --Mike]

Well this is a novel use of Amazon's Mechanical Turk! I daresay it's not a representative sample though - if you look at the amount that is being paid for the tasks requested you'll see that the pay scale is very very (let me add another one...) very low and thus likely attracting people from a very thin socioeconomic sliver of society. Take a look - https://goo.gl/RBWjlL.

It will be interesting to see this theory tested by other researchers to ascertain it's real value. If it pans out, it will be even more interesting to point the machine at the works of Cartier-Bresson, Kertesz, Eisenstaedt, Lange, and others.

Without some level of the so called depressed mood, where would art be?

I wish I could be one of those "artists" making expressive images when my depression takes hold but I'm not. When it hits, the cameras sit. And that makes me sad.

Reviewing the comments to this article makes me wonder how many commenters actually read the article in MIT Technology Review before posting dismissive or negative comments.

I've battled depression and anxiety throughout my life. My passion for photography has been and still is a lifesaver--I enjoy taking pictures of gas stations at night, abstract images of the sea, and dogs.

I read the item in Technology Review. My knee jerk reaction: Sure, it's interesting to "draw" conclusions about people based on doodles, photos, music, prose, and poetry. I'm sure the MIT algorithm has an "AI" module embedded so that it will become more accurate over time. Will it reach or exceed six sigma?

I am rather skeptical this technology will be used to reach out to help people who suffer from mental illness. My guess is HR departments, health insurance providers, lending institution, and security agencies will be the beneficiaries.

I imagine a PI (principal investigator) at MIT suggested that analyzing billions of pictures from social media will serve a multitude of interests/end users. The PI applied for and received a grant (probably from the DoD). The methodology for codifying visual elements from photos taken by happy and sad people (developing a new pattern recognition algorithm) is labor intensive/costly.

The technology for acquiring and storing photos posted on social media has been around for years. Security agencies are clamoring for cheaper, better, faster ways of cataloging/codifying/profiling visual content and the people who post it.

I am sure more than a few grad students drank gallons of Red Bull while "mapping dots and vectors" onto happy and sad photos. (Were the photos culled from a control groups of normal and depressed people?)

Once the lab built up a sufficient library of "photo maps," it began the coding process. I am sure the task of writing the algorithm went through a long process of testing and debugging. Eventually it worked well enough to quantify qualitative data with a fair degree of predictability/accuracy.

After the lab published its paper and received favorable peer review, Technology Review published an article for the educated masses.

(I earned an MS in visual studies at MIT in 1984. I send annual tax deductible contributions to the McGovern Institute for Brain Research at MIT: http://mcgovern.mit.edu/)

----Mike, please take a look at this. It's way early in the morning and I'm sure there are some typos and other errors.

The comments to this entry are closed.



Blog powered by Typepad
Member since 06/2007