How Artificial Intelligence Is Bringing Objectivity To Aesthetics
Since the time of the ancient Greeks, mathematicians, philosophers, and scientists have been searching for an elusive secret: What makes someone beautiful? Whether its Fibonacci’s Golden Ratio, to Leonardo da Vinci’s Vitruvian Man, or Charles Darwin’s competing theories of selection, we’ve long been obsessed with pinpointing the precise formula that makes one person’s face more aesthetically pleasing than another.
Enter artificial intelligence (AI). The science-fiction staple is now our modern reality, and researchers have been employing the technology — essentially a genius-level, one-track mind — in everything from rideshare apps to Alexa and, now, aesthetics.
The Study: Machine Learning, Rhinoplasty, & Aging
Led by associate professor Jason Roostaeian, MD, researchers in the Department of Plastic and Reconstructive Surgery at UCLA recently completed a study that brings us one step closer to objectively quantifying what defines beauty.
Why is that so significant? The answer lies in matching art with science. “As plastic surgeons, we try to be as science-based as possible to be certain everything is working for our patients,” Dr. Jason explains. “We can think we’re creating beauty. But without a way to measure that objectively, we’re not really using the scientific method.”
The UCLA researchers, including plastic surgery residents Sean Saadat, MD, and Robert Dorfman, MD, set out to test whether or not existing machine learning technology (which forms complex calculations based on a set of test images) could accurately recognize what humans consider beautiful. While the initial results showed promise, they quickly realized the high degree of subjectivity would require a greater quantity of images to get objective results.
So, they shifted focus — instead honing in on just one of the many facets of beauty: youth. The surgeons used established AI technology capable of detecting a person’s age more precisely than the human eye to an accuracy within four months. “If you gave me a hundred photos could I guess virtually their exact age? No,” Dr. Jason says. “This proves that the computer is figuring out something that we simply cannot. It’s incredible.”
For the test subjects, researchers passed over the obvious options like facelift and brow lift patients (we already know those procedures reduce apparent age), instead choosing a sampling of Dr. Jason’s most common clients: 100 female rhinoplasty patients, ranging in age from 16 to 72 years old. The results were eye-opening. In patient after patient, the machine returned a postoperative age that was, on average, three years younger than the preoperative age.
The Results: Objectivity in a Subjective Field
Surgeons have long suspected that certain changes to the nose (like refining and rotating the tip, for example) make patients seem younger. “Conceptually, we’ve known for awhile that a droopy tip is associated with facial aging,” Dr. Jason shares, adding that the study, recently published in the Aesthetic Surgery Journal, represents “the first time rhinoplasty has been proven to show a decrease in age.”
While textbooks and journal articles rarely (if ever) correlate facial rejuvenation and rhinoplasty, Dr. Saadat says it should come as no surprise that the popular cosmetic surgery procedure has such an impact.
When the nose is out of balance with the rest of the face, it draws undo attention to itself, Dr. Saadat explains. A rhinoplasty aims to manipulate the angles of the nose — the highlights and shadows — to bring it into better harmony with the surrounding facial features. Refining the nose opens up the face, highlighting the lips and eyes. “Bright, open eyes are a key indicator of youth,” he says. “In youth, facial harmony is optimized. You look more awake, more alive.”
Long term, Dr. Saadat foresees the study’s findings being a critical component of evaluating facial rejuvenation patients. “Of course, not every aging patient needs a rhinoplasty, but it’s an important piece of the puzzle, particularly when we’re fat grafting the mid-face,” he explains. “In some patients, maybe we balance it with a rhinoplasty to keep everything in proper proportion for the most natural results.”
The Future of AI in Aesthetics
So, what do these findings mean for the future of aesthetic medicine more broadly? According to Dr. Saadat, figuring out how beautiful a person is on a scale of one to 10 misses the point. The focus instead is on using AI data — symmetry, proportion, and shape — as a tool to complement our more subjective sense of beauty. “It’s about asking: ‘What would this individual look like were they to be improved in their objective aesthetic harmony,’” he says.
The technology may prove particularly useful in facial reconstruction cases, allowing surgeons to “reverse engineer” what changes would yield the most pleasing outcome for the patient. “Artificial intelligence could provide a roadmap for reconstruction, rather than having to rely solely on what’s in the mind of the plastic surgeon,” Dr. Saadat says.
Down the road, plastic surgeons may even be able to pinpoint the subtle differences in technique that create the most ‘beautiful’ result. “Over time, if we run enough numbers and enough people use this technology, we may be able to potentially figure out those differences” Dr. Jason says.
On the flip side, machine learning may also help patients in their search for the perfect plastic surgeon. “We may be able to get an objective score for a surgeon’s quality of work,” he adds, while conceding there will always be an element of subjectivity. “Whether you agree with it or not, that’s a question everyone will have to answer for themselves,” he shares.
No matter how close we come to objectively cracking the beauty code, subjectivity will always remain. After all, it’s humans who have to tell the technology what is beautiful in the first place. “You feed the machine a bunch of photos, and the computer gets smarter and smarter,” says Dr. Jason. “But, absolutely, the input affects the output.” Given human nature, it’s unlikely that we’ll abandon our own individual sense of beauty, even when faced with quantifiable AI data. Beauty — for now — remains in the eye of the beholder.