What Algorithms and Humans Can Tell Us About How We Are Perceived by Others
In my heart-of-hearts, I am a troubleshooter.
As a digital designer, A/B testing is at the crux of the work that I do.
Well, for other people anyways.
It’s kind of like any profession where you are an expert at implementing certain systems for clients and businesses, but it doesn’t really occur to you to do these things for yourself, or at least not to the degree that you are doing it for everyone else. There are a myriad of examples I could list, but I’m sure you get the gist. Whatever it is you do for a living, you are probably better at it in the service of others than you are in your own life for your own needs.
So, in my search for new work opportunities (and even though it is still early in the process) I wasn’t getting the responses I was hoping for, that I was qualified for, that I felt I deserved based on my experience and other controllable variables.
I re-tooled all of my online descriptions, and adjusted several other elements about my communication style, availability, and overall branding. But things were still not clicking.
I even had my resumé professionally re-written. Twice.
So, what was it? What could it be?
My logo is on-point, my website is solid, testimonials from happy clients are glowing, the body of work is large and diverse, something isn’t adding up here.
At this point in my investigation, the only thing I had not updated in some time (really, only a few months) was the profile photo I was using across the board (social media, job search websites, etc).
I liked the photo that I had taken and chosen. It was professional, direct, clean.
I was smiling, well groomed; and in my opinion, it was an accurate representation of my overall likeness and ‘attitude’.
Out of curiosity, I Googled for some articles about how to improve your LinkedIn profile, including your photo. After poring through blog post after another, I found that my picture was hitting all the necessary marks. Eyeline on the rule of thirds, proper lighting, show your teeth (but not too many!), correct exposure, etc etc.
Still puzzled, I decided to run it through a couple of “analyzers” to see if my opinion of my photo matched how other people perceived it.
Of course, there will always be a disconnect here, it’s just how our brains are wired.
But I wasn’t expecting such a large gap.
The first website was a Snappr appraisal, which uses an automated algorithm to rate your LinkedIn photo based on a variety of metrics (crop, smile, “squinch”, jawline shadow [seriously!], composition, rule of thirds, background, brightness, contrast, sharpness, saturation, and color temperature).
Photo #1 ranked a 76 out of 100 possible points, most of the weight coming from how technically correct it was (I am a photographer, after all!), but it wasn’t super stoked ‘about my face’ in general.
I have a long, angular face with large, indelicate, and sometimes cartoonish features.
There is no way around this.
Sure, you can do things like adjust the focal length of the lens, but really….you just look how you look.
Or so I thought.
Back to the Snappr point…so that grade was a solid “C”. If this were a test I was taking, that would not be an acceptable score for me. So, I took the advice on the areas that ‘needed work’ and decided to replace this photo with a newer one for comparison, incorporating the things that the original photo was weak on.
But there was one more thing I needed to do before I could even decide on what kind of photo I should take instead, and that was to get a more granular second opinion…from actual humans. And not from people that actually know and like me on private social networks, but rather from the uncaring, cold, objective strangers on a “rate my photo” website.
The second website I used to analyze my photo was photofeeler.com, which has a democratic voting process by real people who judge you based on how well they think your photo reflects the following three traits: “competent”, “likable”, and “influential”.
The criteria range for each is “No”, “Somewhat”, “Yes”, and “Very”.
These results were even more disheartening, and people can leave comments as well.
Photo #1 received the following metrics (after about 40-45 votes):
Competent: 4.9 / 10 (below average)
Likable: 5.2 / 10 (average)
Influential: 5.4 / 10 (average)
People also left comments like “scary” and “forced smile”
Photo #1 received the following metrics (after about 40-45 votes):
Competent: 9.1/10 (top 10%)
Likable 9.8/10 (top 3%)
Influential 8.6/10 (top 20%)
Comments included “too flirty”, “great smile”, and “smile less” (you really just cannot win!)
So, what was the difference?
Let it be known that both photos were edited using Portrait Pro for “face-tuning”, and then in Lightroom/Photoshop to soften/blur the background, and for the tweaking of individual settings (exposure, contrast, color temp, etc).
The major difference between photo #1 and photo #2 is that in photo #2 I basically “shaved” off about 25% of the length of my face.
Yep, that’s right, I doctored it like it was going on the cover of a magazine (the kind that promotes “healthy body image” no doubt!).
I drastically shortened my forehead, chin, and nose, narrowed my jaw and chin, elongated my neck, etc.
I obviously did not feel good about doing this, but if that’s the game, then I’m gonna play it.
A lot of my work is marketing, and marketers don’t do what’s “nice”, they do what works.
On photo #1, I had done some light “boiler-plate” retouching, but did not alter the overall size and shape of my face, or of any of my features. And come on now, I am fairly attractive, so I’ll be honest in admitting that I didn’t think very much work was necessary.
But the data shows that I was clearly wrong.
My natural face is not a face people respond positively to.
A little too Joan Crawford-esque, perhaps?
I needed to be more angelic, more Bambi-esque.
I needed to stop looking like Eva Green (despite her massive success), and start looking more like Emma Watson.
It’s right there in the numbers, and the numbers don’t lie.
In addition to drastically altering my entire face (while still looking “like me”), here is a list of other items that I changed:
Photo #1 / Photo #2
Plain, black shirt / Dainty white blouse with polka dots and collar bow
No glasses / Glasses
3/4 face & body angle / “Full-frontal” face & body angle
Tight Crop / Wider crop (most of the common advice instructs you to cut it off at the armpits)
High contrast & medium saturation / Practically zero contrast (-86!), low saturation
Backgrounds identical (shot in the same place in the same room)
Hair colored and styled exactly the same
So, after getting those stellar results from human judging, I ran photo #2 through the automated algorithm from Snappr, and it yielded and even lower score than the original (but barely though, a 75 out of 100). Such disparate results between the two methods!
Now what does this experiment teach us (besides the painfully obvious)?
It tells us that, with regards to how professional a female is perceived, people of all genders and ages respond more kindly to shorter, rounder faces with softer, pinker, more feminine features. The longer and thinner my face is, the less competent, likable, and influential I am considered. This isn’t really breaking news, as women have been subjected to insane beauty standards since the dawn of time, but it is quite unfortunate, especially in 2018.
Some of us are just sort of…“mannish”, and despite my attempts at “owning it” as many advise, that ownership is clearly not doing us any favors, statistically speaking.
Not only that, but this is an exercise on how literally easy it is to visually deceive without the audience being none the wiser. When I show up to the interview, I’ll definitely still look like ‘the girl in the picture’, and while the differences are pretty drastic, it’s unlikely that anyone will be able to actually pinpoint why that is beyond some basic strategic subtlety.
I uploaded the new shot to my personal social networks and the likes and comments came flowing in. Not one person questioned whether I had “done something to my face”.
Not even my husband when I had showed him the final 4 photos I was deciding between (all having been pre-doctored unbeknownst to him, before the final edits). And as someone who is as intimately familiar with my face as anyone can be, if he thought the photo was an accurate representation, then certainly so will people who have only cursorily glanced at it once or twice.
My own mother, one of my best girl-friends, and my cousin even called it “adorable”.
So if they were fooled by it, then everybody else will be too.
More reporting on this experiment as the job search continues. Stay tuned.