Meyer felt he knew the people personally – those who described smells in terms of tea and fruit, or meat and gasoline, or blue powerade and lollipops. The way they described their senses felt so intimate that he would later say, “You could almost see what kind of person they are.” He believed that people believed they could smell bad describe just because so often in laboratories they are asked to sniff single, isolated molecules (when the more familiar smell of coffee is a mix of many hundreds of them) away from the context of their real life and the smells that actually mattered to them . On the right occasion he said, “People get very, very verbal.”
This was exciting news for Meyer, an IBM researcher who specializes in using algorithms to analyze biological data and who insisted that the GCCR surveys contain open text boxes. For years, scientists studying odors have only worked on a few extremely flawed sets of data relating different chemicals and the way people perceive them. For example, there was a record made by a single perfumer in the late 1960s describing thousands of smells, and study after study was based on a single “Atlas of Odor Character Profiles” published in 1985. It relied on the observations of volunteers who had been asked to smell various single molecules and chemical mixtures, to rate and name them according to a list of descriptors provided, which many scientists believed to be flawed and dated.
More recently, Meyer and many others had used a new data set carefully compiled by scientists at Rockefeller University in New York and published in 2016. (I visited the lab in 2014 while Leslie Vosshall and her colleagues were compiling their data.) And was surprised to see that I could “smell” one of the vials, even though it probably only triggered my trigeminal system. When I told Vosshall that it seemed minty, she replied, “Really? Most people say ‘dirty socks’. Although the new dataset was a significant improvement, 55 people smelled 480 different molecules and rated them for intensity, comfort, familiarity, and how well they matched a list of 20 descriptions, including “garlic”, “spice”, “flower”. “Bakery,” “musk,” “urine” and so on – it was still a sign of how limited the field was.
For this reason, Meyer and his colleague Guillermo Cecchi pushed for these open text fields in the GCCR survey. They were interested in the possibilities of natural language processing, a branch of machine learning that uses algorithms to analyze patterns of human expression. Cecchi was already using the technology to predict the early onset of Alzheimer’s when it was most treatable by analyzing details of the way people speak. Many researchers had written about the possibilities of using artificial intelligence to finally create a predictive odor map and study relationships between changes in odor formation and any diseases that these changes are associated with, but adequate data was never available.
Now Covid had provided the researchers with a large, complicated data set that linked the olfactory experience and the progression of a particular disease. It wasn’t constrained by numerical rankings, monomolecules, or some adjectives on offer, but instead allowed people to speak freely about real smells in the real world in all their complex and subjective glory.
When Meyer and Cecchi’s colleague Raquel Norel had finished analyzing the open-ended responses from the English-speaking respondents, they were surprised and delighted to find that their text analysis predicted a Covid diagnosis as well as the numerical ratings of odor losses. The algorithms worked because people with Covid used very different words to talk about odor than those without Covid. Even those who had not completely lost their smell tended to describe their sensations in the same way and use words like “metallic,” “decayed,” “chemical,” “sour,” “sour,” “burned” and ” Urine ”to repeat. “It was encouraging finding to examine a proof of concept that they couldn’t wait to look further into – first in the GCCR responses in other languages, and then in the future in other datasets related to other diseases. Meyer was excited when he talked about it. “Anything where the smell changes,” he told me. “Depression, schizophrenia, Alzheimer’s, Parkinson’s, neurodegeneration, cognitive and neuropsychiatric diseases. The whole enchilada, as they say. “
I had a hard time Imagine the olfactory “map” that scientists have dreamed of for so long. I asked Mainland, would it look something like a periodic table? He suggested I think instead of the maps that scientists have made out of “color space” and arrange the colors to show their mathematical relationships and mixtures. “We didn’t know how useful color space was until people started inventing things like color TV and Photoshop,” he explained, adding that the map itself isn’t the goal, but the ability to use it to understand why we are what do we smell. What will be really interesting after that are the applications that we cannot yet imagine. “It’s hard to understand how useful the card is,” he said, “until you have the card.”