Tuesday, October 13, 2015

The Q-Test: What's Cute, and What's Not?


This post is about a survey I created on October 12th, 2015. I encourage you to make sure that you take the survey before you read any further.

It was Thanksgiving Sunday in 2015, and I was at a café with my extended family. There are some babies and toddlers in the family, so I brought up the topic of Sophie the Giraffe, a wildly popular toy designed for babies to chew on while their teeth are growing in. My aunt Andrea looked it up on her phone, and thought that the design looked terrible -- she said it was creepy, and that Sophie's lifeless black eyes reminded her of a shark. ("A cute shark, perhaps?" I suggested to no avail.)

(Above: Sophie the Giraffe. Photo courtesy of Toys R Us. Used for educational purposes)

This surprised me. I thought Sophie's designed seem rather cute and harmless. I asked some other relatives, and many of them agreed: Sophie's eyes looked disturbing, like an alien.

This puzzled me, enough that I was still thinking about it when we had finished Thanksgiving dinner. It interested me partly because I've drawn cartoons in a professional capacity (in that I have been paid at least once in my life to draw a cartoon) and partly because of my interest in psychology. You would think that cuteness would be a universally agreed upon thing. It's an emotion we feel in order to motivate us to take care of our young, isn't it? So, shouldn't we all find the same things cute?

Sophie's face uses what I thought was a universal recipe for cuteness: large pupils, eyes far apart. It's what you see in many popular emojis, cute logos, and in cute baby animals of almost any species.

(Above: A baby sloth, courtesy of your-daily-baby-sloth.tumblr.com. Used for educational purposes.)

I asked my relatives what they would do to make Sophie cuter. They showed me examples of cartoons and designs they considered to be cute, and they all had eyes closer together with smaller pupils. I don't hate that design, but done poorly, it seems kind of tacky to me. Indeed, it describes almost all of the results that come up when I look up "bad cartoon clipart" on Google Images.

(Above: The result of an actual Google Image search. If you look in the upper-right corner, you can see remnants of my research for the previous photo in this blog post.)

My theory, although I didn't dare say it out loud, was that the older family members would prefer small pupils close together, and younger family members would prefer large pupils far apart. I wasn't sure why yet, but it seemed like a reasonable prediction, especially since it seems to be the younger generation that's always using emojis and sharing photos of baby sloths on the internet. So, I performed a little survey of everyone in my family, or at least, everyone who was at Thanksgiving dinner that year. I drew two simple cartoon faces, one with tiny pupils and eyes close together, and one with big black pupils and eyes far apart, as follows:

(Above: A scan of the paper towel I quickly drew the faces on. The words "Left" and "Right" are primarily for my benefit. I've written down the names of each family member, but I trust that my crappy handwriting will help maintain their privacy. You can see my note to myself "Facebook Poll?" at the top.)

The left face got 14 votes. The right face got 12 votes. I was surprised that it was so close to a 50-50 split, especially since some of my family members were very passionate on one side or the other. I mentioned that my aunt Andrea thought the right image was creepy, and some other family members said it looked like an alien. On the other hand, a few of my cousins said that the right face was cuter by far, and that the left one looked creepy.

The participants included my first cousins once removed, Lyla (age 5) and Luca (age 3). Lyla said that the one on the right was cuter. Luca preferred the one on the left, because it reminded him of Cookie Monster.

I wasn't able to immediately figure out whether age was a factor; the factions didn't neatly divide the older generation from the younger generation.

There were some flaws in my methodology. For one thing, my strategy was to go around and ask people what they thought, while they were distracted by other things, and while other people were listening to their answers. This might have affected their answers.

In addition, when I drew the two pictures, I was really changing two variables, which is a no-no in science: I was changing both the pupil size, and the distance apart. If I wanted to get some decent findings, I would need to use four faces, not two.

So, I decided to set up an online survey.

The Survey

On the morning of October 12th, I taught myself to use Google Forms to create a survey and post the link on Facebook so my friends could see it. The survey centred around four smiling faces I drew with MS Paint. I assigned them the arbitrary numbers 1 to 4.

This forms a more complete set of possibilities than the two I drew on the paper towel. You can see that I've included large pupils far apart (1), large pupils close together (4), small pupils far apart (2), and small pupils close together (3).

Strictly speaking, it probably would have been better if I had randomized the numbering for each survey participant so that the order would not affect the overall results. But I had just learned to use Google Forms that morning, so give me a break.

In order to test my prediction, I had each participant give their age. Then, instead of asking the participants to just tell me which face was the cutest, I had them rank the faces from cutest to least cute. I did this because it gives me more information to work with than simply telling me which one they like the best, and I like having information to work with.

Plus, it gave me a chance to interpret the results in terms of voting methods. After all, each participant, in giving a ranking of the faces, was expressing their preference, and that's basically what a vote is. Traditionally, most voting we encounter in everyday life works by simple plurality, but there were two other methods I wanted to try: Instant-Runoff voting (IRV, also known as Single Transferable Vote or STV in Canada), and the Beatpath method, both of which work best when the voters give a ranking of their preferences, rather than just putting an X beside the one they like best.

This is where it gets a little boring and technical, because this section has less to do with cute faces, more to do with mathematics of social choice. I won't be offended if you skip to the next section.

For those of you who chose to read on: Given a set of rankings given by voters (also known as a preference schedule) my job was to aggregate the votes and figure out how they determine a ranking for the entire population. This is easier said than done. In fact, according to Arrow's Impossibility Theorem, it's, well, impossible to do it in a way that totally satifies a small set of reasonable criteria.

Still, even in the face of possibility, I thought it was important to do as well as I could to aggregate the results. And in fact, some vote-counting methods are superior to others. In particular, the Beatpath method is better than IRV and Plurality in almost every way. The Wikipedia article on voting systems has a nice table that explains how various voting systems fare according to different criteria, each criterion more complicated and abstract than the last. But the important facts are this:

- Both IRV and Beatpath prevent the simple Ralph Nader-style vote splitting that happens when the plurality method is used.
- In addition, unlike IRV, Beatpath is monotonic, meaning that there is no way for a "candidate" to change from winner to loser by rising in a voter's ranking. This page contains some examples which should clarify what it means to say that IRV fails to be monotonic.

I don't read many psychology papers; I'm sure I'm not the first person who's had to take a group of participants' rankings, and figure out how to summarize them. I'm not sure what most psychologists do in this situation, so I just aggregated the rankings using Plurality, IRV and Beatpath to see what it looked like. However, I ultimately place more trust in Beatpath than in the other two methods.

The Results

As of writing this blog post, my survey has had just over 100 responses. The median age is 25. I turn 24 this year, so it makes sense that most of my friends, many of whom were classmates in high school and university, would cluster near my age. The youngest participant in the Google survey was 17, and the oldest was 73.

You can see the results for three categories: General, 24 and Under, and 25 and Older. Since 25 was the median, it seemed like a natural cutoff point to separate older and younger participants.

In each of the three categories, I've used three methods of aggregating the participants' rankings, as described in the previous section. Since I trust Beatpath the most, I have put it in boldface.

For convenience, I'll reproduce the image of the four faces again here:

General voters:
Plurality: [1, 3, 2, 4]
IRV: [1, 3, 2, 4]
Beatpath: [1, 2, 3, 4]

Voters 24 and Under:

Plurality: [1, 2, 3, 4]
IRV: [1, 2, 3, 4]
Beatpath: [1, 2, 3, 4]

Voters 25 and Older:
Plurality: [3, 1, 2, 4]
IRV: [3, 1, 2, 4]
Beatpath: [3, 2, 1, 4]

A few observations immediately jump out at me:

Generally, the face with big pupils far apart (1) was the favourite of the group. This makes sense to me, since (1) is the category of puppies, baby sloths, emojis, and so on.

However, if you look at voters 24 and under vs. voters 25 and over, it becomes clear that (1) and (3) are very polarizing -- just like it was among my family at Thanksgiving dinner. Younger people think (1) is the cutest and (3) is the third cutest, whereas older people think (3) is the cutest and (1) is the third cutest. This is in line with my prediction that younger people prefer large pupils far apart while older people prefer small pupils close together.

(2) was universally considered to be second cutest, and (4) was universally considered to be least cute. This seems odd to me. You would think that if big pupils far apart (1) are cute, then big pupils close together (4) would be almost as cute. But putting big pupils close together seems to ruin the cuteness, even for the 24 and under group.

Those are the questions that I was focussed on answering, but perhaps you have others. For example, do the answers give us a way of measuring the intensity of preference for one face over another, at least for the population? What if we were to divide the group into further age subdivisions, or perhaps subdivisions by which face they thought was cutest, or second cutest, and so on? If you want to try to answer any of these questions, or questions of your own, feel free to look at the data yourself.

Crazy Speculations (a.k.a. Discussion)

What theory of cuteness would explain these results? Hell if I know. But I can still make some guesses.

For example, I still think I was essentially correct in my belief that big pupils far apart (1) is a reliable recipe for cuteness. As I've mentioned, it's the face of most baby animals.

It's not clear to me why humans should find other species' babies cute. But the fact remains that we do. The big, black pupils are a trait we're more likely to associate with other animals than with humans.

In addition, having the eyes far apart might make the face look less threatening, because it's the trait of herbivores. Herbivores usually have eyes closer to the sides of their heads, so that they can always have a good panoramic view of their surroundings to look for predators (e.g.: deer, horses, hummingbirds, baby sloths). Carnivores and omnivores, in contrast, are more likely to have their eyes on the front of their head in order to get good depth perception, which is useful for hunting (e.g. humans, spiders, cats, dogs, owls, raccoons).

This would help to explain why big pupils close together (4) was universally considered to be the least cute. Big pupils make it look more animalistic than human, and the fact that the eyes are close together makes the face look more like a predator than like prey.

However, this doesn't explain why the older crowd found small pupils close together (3) so cute. One clue towards figuring this out might be the fact that (3) looks more like a human baby than the other faces. Humans have eyes close together rather than on the sides of our heads, and we have small pupils surrounded by big white sclera. Perhaps the older group contains more parents and grandparents, who are more likely to find babies cute -- perhaps becoming a parent actually causes a person to find babies cuter than they did before. Younger people find babies less cute, as evidenced by the fact that most of the cute pictures that get shared on the internet are of animals, rather than baby humans. However, it's not clear to me why young people like myself are so indifferent towards babies.

I have another guess as to why older participants found (3) to be the cutest: perhaps they somehow learned to find it cute. My 3-year-old cousin Luca might have hit upon a profound observation when he said that the face with small pupils close together looked like Cookie Monster. Sesame Street and The Muppets have been around for over 40 years, and most of them are designed to have small pupils, close together. Looney Tunes and Disney characters also have small pupils close together, and they've been around even longer. Perhaps we're born with the capacity to experience cuteness, but what triggers that feeling is something we have to learn -- kind of like how we're all born with the capacity for language, but the specific language we learn is determined by our environment. Visual styles go in and out; perhaps big pupils far apart are simply "in" these days, and the next generation will defy us by preferring small pupils close together.

It would be a mistake to finish this discussion without mentioning the Uncanny Valley. We like abstract, simple, cute faces (e.g. emojis, baby sloths) and we like them even more when they're slightly more realistic (e.g. Disney cartoons, superhero comics). But there comes a point where a face is almost perfectly realistic, but not quite: this is called the Uncanny Valley, and contains the almost-human faces that creep us out (e.g. zombies, horror movie masks, sex robots, Michael Jackson). But if we push past the uncanny valley, towards realistic human faces, eventually we find them appealing again (e.g. Mona Lisa, Neil Patrick Harris... do I really have to list examples of people with human faces?)

I mention this because it might be useful in understanding the results of the survey. Young people liked (1) -- it's simple, with big black eyes that are far apart. It's not human enough to be even close to the Uncanny Valley, so it's just cute. Everybody, young and old, was okay with (2) -- it's slightly more human than (1), but not enough to be creepy. (4) is where it gets interesting. I think that, at least relative to the other faces in this survey, (4) represents the uncanny valley. The proportions of the face are similar to a human baby, but the eyes are too animalistic. It looks like some kind of alien baby. Finally, (3) was the face that the older group loved. (3) is the face that is most like an actual human. Perhaps for the older crowd, (3) was well past the uncanny valley, but for the younger crowd, (3) was still not quite human enough to be cute.

So, that's my analysis. Feel free to look at the data and find patterns on your own, and maybe try to convince your friends to take the survey before they see this blog post.

No comments:

Post a Comment

Creative Commons License
It Seemed Funny at the Time by Ben Buckley is licensed under a Creative Commons Attribution-Noncommercial 2.5 Canada License.