By 1992, burgeoning population, choking traffic, and explosive industrial growth in Mexico City had caused the United Nations to label it the most polluted urban area in the world. The problem was intensified because the high-altitude metropolis sat in a valley trapping that atmospheric filth in a perpetual toxic haze. Over the next few years, the impact could be seen not just in the blanket of smog overhead but in the city’s dogs, who had become so disoriented that some of them could no longer recognize their human families. In a series of elegant studies, the neuropathologist Lilian Calderón-Garcidueñas compared the brains of canines and children from “Makesicko City,” as the capital had been dubbed, to those from less polluted areas. What she found was terrifying: Exposure to air pollution in childhood decreases brain volume and heightens risk of several dreaded brain diseases, including Parkinson’s and Alzheimer’s, as an adult.

Calderón-Garcidueñas, today head of the Environmental Neuroprevention Laboratory at the University of Montana, points out that the damaged brains she documented through neuroimaging in young dogs and humans aren’t just significant in later years; they play out in impaired memory and lower intelligence scores throughout life. Other studies have found that air pollution exposure later in childhood alters neural circuitry throughout the brain, potentially affecting executive function, including abilities like decision-making and focus, and raising the risk of psychiatric disorders.

The stakes for all of us are enormous. In places like China, India, and the rest of the global south, air pollution, both indoor and outdoor, has steadily soared over the course of decades. According to the United Nations Foundation, “nearly half of the world’s population breathes toxic air each day, including more than 90 percent of children.” Some 2.3 billion people worldwide rely on solid fuels and open fires for cooking, the Foundation adds, making the problem far worse. The World Health Organization calculates about 3 million premature deaths, mostly in women and children, result from air pollution created by such cooking each year.

In the United States, meanwhile, average air pollution levels have decreased significantly since the passage of the Clean Air Act in 1970. But the key word is average. Millions of Americans are still breathing outdoor air loaded with inflammation-triggering ozone and fine particulate matter. These particles, known as PM2.5 (particles less than 2.5 micrometers in diameter), can affect the lungs and heart and are strongly associated with brain damage. Wildfires—like the ones that raged across Canada this past summer—are a major contributor of PM2.5. A recent study showed that pesticides, paints, cleaners, and other personal care products are another major—and under-recognized—source of PM2.5 and can raise the risk for numerous health problems, including brain-damaging strokes.

Untangling the relationship between air pollution and the brain is complex. In the modern industrial world, we are all exposed to literally thousands of contaminants. And not every person exposed to a given pollutant will develop the same set of symptoms, impairments, or diseases—in part because of their genes, and in part because each exposure may occur at a different point in development or impact a different area of the body or brain. What’s more, social disparities are at play: Poorer populations almost always live closer to factories, toxins, and pollutants.

The effort to figure it out and intervene has sparked a new field of study: exposomics, the science of environmental exposures and their effects on health, disease, and development. Exposomics draws on enormous datasets about the distribution of environmental toxins, genetic and cellular responses, and human behavioral patterns. There is a huge amount of information to parse, so researchers in the field are turning to another emerging science, artificial intelligence, to make sense of it all.

“Anything from our external environment—the air we breathe, food we eat, the water we drink, the emotional stress that we face every day—all of that gets translated into our biology,” says Rosalind Wright, professor of pediatrics and co-director of the Institute for Exposomic Research at the Icahn School of Medicine at Mount Sinai in New York. “All these things plus genes themselves explain the patterns of risk we see.” When an exposure is constant and cumulative, or when it overwhelms our ability to adapt, or “when you’re a fetus in utero, when you’re an infant or in early childhood or in a critical period of growth,” it can have a particularly powerful effect on lifelong cognitive clarity and brain health.

Bad air quality is associated with increased rates of bipolar disorder and depression.

Neuroscientist Megan Herting at the University of Southern California (USC) has been studying the impact of air pollution on the developing brain. “Over the past few years, we have found that higher levels of PM_2.5 exposure are linked to a number of differences in the shape, neural architecture, and functional organization of the developing brain, including altered patterns of cortical thickness and differences in the microstructure of gray and white matter,” she says. On the basis of neuroimaging of exposed youngsters, Herting and fellow researchers suspect the widespread differences in brain structure and function linked with air pollution may be early biomarkers for cognitive and emotional problems emerging later in life.

That suspicion gains support from an international meta-analysis (a study of other studies) published in 2023 that correlated exposure to air pollution during critical periods of brain development in childhood and adolescence to risk of depression and suicidal behavior. The imaging parts of the studies showed changes in brain structure, including neurocircuitry potentially involved in movement disorders like Parkinson’s, and white matter of the prefrontal lobes, responsible for executive decision-making, attention, and self-control.

In a 2023 study, Herting and colleagues tracked children transitioning into adolescence, when brains are in a sensitive period of development and thus especially vulnerable to long-term damage from toxins. Among brain regions developing during this period is the prefrontal cortex, which helps with cognitive control, self-regulation, decision-making, attention, and problem-solving, Herting says. “Your emotional reward systems are also still being refined,” she adds.

Looking at scan data from more than 9,000 youngsters exposed to air pollution between ages 9 and 10 and following them over the next couple of years, the researchers found changes in connectivity between brain regions, with some regions having fewer connections and others having more connections than normal. Herting explains that these structural and functional connections allow us to function in our daily lives, but how or even whether the changes in circuitry have an impact, researchers do not yet know.

The specific pollutants involved in the atypical brain circuits appear to be nitrogen dioxide, ozone, and PM_2.5—the small particles that worry many researchers the most. Herting explains: Limits set on fine particulate matter are stricter in the United States than in most other countries but still inadequate. The U.S. Environmental Protection Agency currently limits annual average levels of the pollutant to 12 micrograms per cubic meter and permits daily spikes of up to 35 micrograms per cubic meter. Health organizations, on the other hand, have called for the agency to lower levels to 8 micrograms and 25 micrograms per cubic meter, respectively. Thus, even though it may be “safe” by EPA standards, “air quality across America is contributing to changes in brain networks during critical periods of childhood,” Herting says. And that may augur “increased risk for cognitive and emotional problems later in life.” She plans to follow her group of young people into adulthood, when advances in science and the passage of time should reveal more about the effect of air pollution exposure during adolescence.

Other research shows that air pollution increases risk of psychiatric disorder as years go by. In work based on large datasets in the United States and Denmark, University of Chicago computational biologist Andrey Rzhetsky and colleagues found that bad air quality was associated with increased rates of bipolar disorder and depression in both countries, especially when exposure occurs early in life. Rzhetsky and his team used two major sources: in Denmark, the National Health Registry, which contains health data on every citizen from cradle to grave; and in the United States, insurance claims with medical history plus details such as county of residence, age, sex, and importantly, linkages to family—specifics that helped reveal genetic predisposition to develop a psychiatric condition during the first 10 years of life.

“It’s possible that the same environment will cause disease in one person but not in another because of predisposing genetic variants that are different in different people,” Rzhetsky says. “The different genetic predisposition, that’s one part of the puzzle. Another part is varying environment.”

Indeed, these complex diseases are spreading much faster than genetics alone seems to explain. “We definitely don’t know for sure which pollutant is causal. We can’t really pinpoint a smoking gun,” Rzhetsky says. But one pesky culprit continues to prove statistically significant: “It looks like PM_2.5 is one of those strong signals.” To figure it out specifically, we’ll need much more data, and exposomics will play a vital role.

Regional differences in Parkinson’s disease might reflect regional differences in the composition of the particulate matter.

”

This is a wake-up call,” Frances Jensen told her fellow physicians at the American Neurological Society’s symposium on Neurologic Dark Matter in October 2022. The meeting was an exploration of the exposome –the sum of external factors that a person is exposed to during a lifetime— driving neurodegenerative disease. It was focused in no small part on air pollution. Jensen, a University of Pennsylvania neurologist and president of the American Neurological Association, argued that researchers need to pay more attention to contaminants because the sharp rise in the number of Parkinson’s diagnoses cannot be explained by the aging population alone. “Environmental exposures are lurking in the background, and they’re rising,” she said.

Parkinson’s disease is already the second-most common neurodegenerative disease after Alzheimer’s. Symptoms, which can include uncontrolled movements, difficulty with balance, and memory problems, generally develop in people age 60 and older, but they can occur, though rarely, in people as young as 20. Could something in the air explain the increasing worldwide prevalence of Parkinson’s? Researchers have not identified one specific cause, but they know Parkinson’s symptoms result from degeneration of nerve cells in the substantia nigra, the part of the brain that produces dopamine and other signal-transmitting chemicals necessary for movement and coordination.

A host of air pollution suspects are now thought to play a role in the loss of dopamine-producing cells, according to Emory University environmental health scientist W. Michael Caudle, who uses mass spectrometry to identify chemicals in our bodies. One suspect he’s looking at are lipopolysaccharides, compounds often found in air pollution and bacterial toxins. Although lipopolysaccharides cannot directly enter the brain, they inflame the liver. The liver then releases inflammatory molecules into the bloodstream, which interact with blood vessels in the blood-barrier. “Then the inflammatory response in the brain leads to loss of dopamine neurons, like that seen in Parkinson’s disease,” Caudle says.

More evidence comes from neuroepidemiologist Brittany Krzyzanowski, based at the Barrow Neurological Institute in Phoenix. Krzyzanowski had an “aha!” moment when she saw a map highlighting the high risk of Parkinson’s disease in the Mississippi–Ohio River Valley, including areas of Tennessee and Kentucky. At first she wondered whether the Parkinson’s hotspot was due to pesticide use in the region. But then it hit her: The area also had a network of high-density roads, suggesting that air pollution could be involved. “The pollution in these areas may contain more combustion particles from traffic and heavy metals from manufacturing, which have been linked to cell death in the part of the brain involved in Parkinson’s disease,” she said.

In a study published in Neurology in October 2023, Krzyzanowski and colleagues, using sophisticated geospatial analytic techniques, went on to show that those with median levels of air pollution have a 56 percent greater risk of developing Parkinson’s disease compared to those living in regions with the lowest level of air pollution. Along with the Mississippi-Ohio River Valley, other hotspots included central North Dakota, parts of Texas, Kansas, eastern Michigan, and the tip of Florida. People living in the western half of the U.S. are at a reduced risk of developing Parkinson’s disease compared with the rest of the nation.

As to the hotspot in the Mississippi-Ohio River Valley, Parkinson’s there is 25% higher than in areas with the lowest air particulate matter. Aside from that, Krzyzanowski and her research team noted something especially odd: Frequency of the disease rose with the level of pollution, but then it plateaued even as air pollution continued to soar. One reason could be that other air pollution-linked diseases, including Alzheimer’s, are masking the emergence of Parkinson’s; another reason could be an unusual form of PM_2.5. “Regional differences in Parkinson’s disease might reflect regional differences in the composition of the particulate matter, and some areas may have particulate matter containing more toxic components compared to other areas,” Krzyzanowsk says. Tapping the tenets of exposomics, she expects to explore these issues in the months and years ahead.

The hunt is on for the connections between environmental factors and Alzheimer’s as well. USC neurogerontologist Caleb Finch has spent years studying dementia, especially Alzheimer’s disease, which affects more than six million Americans. As with Parkinson’s, Alzheimer’s numbers are rising in the United State and much of the world. Degenerative changes in neurons become increasingly frequent after the age of 60, yet half of the people who make it to 100 will not get dementia. Many factors could explain those discrepancies. Air pollution may be an important one, Finch says.

Researchers like Finch and his USC colleague Jiu-Chiuan Chen are joining forces to explore the connections between environmental neurotoxins and decline in brain health. It’s a challenging project, since air pollution levels and specific pollutants vary on fine scales and can change from hour to hour in many areas of the globe. On the basis of brain scans of hundreds of people over a range of geographic areas, this much we know: “People living in areas of high levels of air pollution and who have been studied on three continents showed accelerated arterial disease, heart attacks, and strokes, and faster cognitive decline,” Finch says.

Not everyone reacts the same way when exposed to pollutants, of course. Greatest risk for Alzheimer’s seems to hit people who have a genetic variant known as apolipoprotein E (APOE4), which is involved in making proteins that help carry cholesterol and other types of fat in the bloodstream. About 25 percent of people have one copy of that gene, and 2 to 3 percent carry two copies. But inheriting the gene alone doesn’t determine a person’s Alzheimer’s risk. Environmental exposures count too.

A recent study by Chen, Finch, and colleagues published in the Journal of Alzheimer’s Disease looked at associations between air pollution exposure and early signs of Alzheimer’s in 1,100 men, all around age 56 when the study began. By age 68, test subjects with high PM_2.5 exposures had the worst scores in verbal fluency. People exposed to high levels of nitrogen dioxide (NO₂) air pollution were also linked to worsened episodic memory. The men who had APOE4 genes had the worst scores in executive function. The evidence indicates that the process by which air pollution interacts with genetic risk to cause Alzheimer’s in later life may begin in the middle years, at least for men.

A separate USC study of more than 2,000 women found that when air quality improved, cognitive decline in older women slowed. When exposure to pollutants like PM_2.5 and NO₂
dropped by a few micrograms per cubic foot a year over the course of six years, the women in the study tested as being a year or so younger than their real age. This suggests that when exposure air pollution is lowered, dementia risk can go down.

In parallel, an international study by the Lancet Commission concluded that the risk of dementia, including Alzheimer’s, can be lowered by modifying or avoiding 12 risk factors: hypertension, hearing impairment, smoking, obesity, depression, low social contact, low level of education, physical inactivity, diabetes, excessive alcohol consumption, traumatic brain injury—and air pollution. Together, the 12 modifiable risk factors account for around 40 percent of worldwide dementias, which theoretically could be prevented or delayed.

In light of all this, Finch and Duke University social scientist Alexander Kulminski have proposed the “Alzheimer’s disease exposome” to assess environmental factors that interact with genes to cause dementia. Where medicines have failed, exposomics just might help. Studies of Swedish twins show that half of individual differences in Alzheimer’s risk may be environmental, and thus modifiable; and while vast sums of research funding have been poured into the genetic roots of the disease, it could be that altering the exposome would provide a better preventive than all the ongoing drug trials to date. Environmental toxins broadly disrupt cell repair and protective mechanisms in the brain, the researchers point out. And factors like obesity and stress contribute to chronic inflammation, which likely damages neurons’ ability to function and communicate. The research framework of the Alzheimer’s disease exposome offers a comprehensive, systematic approach to the environmental underpinnings of Alzheimer’s risk over individuals’ lifespans—from the time they are pre-fertilized gametes to life as a fetus in the womb to childhood and beyond.

The exposome could explain more subtle cognitive effects of pollution, including harms to attention, intelligence, and performance.

For three decades, Rosalind Wright at Mount Sinai has wanted to trace critical problems in neurodevelopment and neurodegeneration to pollutants—from highway emissions to heavy metals to specific household chemicals and a host of other factors—but the mass of data has been overwhelming. With the advent of artificial intelligence (AI) and sophisticated neuroimaging technology, high-precision research using vast genomic databanks is finally possible. “I knew we needed to ask these kinds of questions, but I didn’t have the tools to do it. Now we do and it’s very exciting,” Wright says.

Using machine learning—an AI approach to data analysis—Wright looks at giant datasets that include the precise location of an individual’s residence as well as the myriad of pollutants he or she encounters. “It’s no different fundamentally from other statistical models we use,” she says. “It’s just that this one has been developed to be able to take in bigger and bigger data, more and more types of exposures.” The resulting data breakdown should tell us which factors drive which types of risk for which people. That information will help people know where they should target their efforts to reduce exposures to risky pollutants, and ultimately how to lower risk of impairment and disease, brain or otherwise.

The tools used by Wright and her colleagues are being trained on diseases like Alzheimer’s. If you put genes and the environment together, “you start to see who might be at higher risk and also what underlying mechanisms might be driving it in different ways in different populations,” Wright says. The exposome could also explains more subtle cognitive effects of pollution that may emerge over long periods, such as harms to attention, intelligence, and performance.

To address environmental brain risks, it’s important to know which pollutants are present—another target of exposomic research. In the United States, the EPA has placed stationary environmental monitors all over our major cities, conducting daily measurements of small particulates from traffic and industry, along with secondary chemicals that emerge as a result. There are also thousands of satellites all over the globe calibrating heat waves that can alter how the pollutants react with each other.

Pioneers like Wright are just starting to chart the terrain of environmental exposures that affect the brain. “As we measure more and more of the exposome, we may be able to tailor prevention and intervention strategies. New weapons include a silicone bracelet that we have in the laboratory. You wear it and it will tell us what pollutants you are exposed to,” Wright says. She also is exploring more ways to collect data on the toxins people have already encountered: “With a single strand of hair, we can tell you what you’ve been exposed to. Hair grows about a centimeter a month, so if we get a hair from a pregnant woman and she has nine centimeters of hair, we can go back a full nine months, over the entire life of the fetus. Or we can create a life-long exposome history when a child loses a tooth at age six.”

“We’re designed to be pretty resilient,” Wright adds. The problem comes when the exposures are chronic and accumulative and overwhelm our ability to adapt. We’re not going to fix everything, “but if I know more about myself than before, that empowers me to think, ‘I’m optimizing the balance, and I’m intervening as best I can.’ ”

Additional reporting and editing was done by Margaret Hetherman.

This story is part of a series of OpenMind essays, podcasts, and videos supported by a generous grant from the Pulitzer Center‘s Truth Decay initiative.

After decades of searching, many scientists believe they have finally pinned down the main problem with our modern diets—the factor driving ever-escalating rates of obesity, diabetes, heart disease and any number of other serious chronic conditions. The culprit isn’t saturated fats, trans fats or some new killer fat you haven’t of. It’s not cholesterol, carbs or sugars; dairy, gluten or meat. It isn’t a specific thing at all.

The problem, according to this increasingly popular argument, is that we’re eating lots of ultra-processed food or UPF. Traditionally, people used limited processing techniques such as cooking and pickling to preserve food or to make it more pleasant to eat. Modern food companies transform food much more extensively through techniques such as extrusion and molding, adding lab-derived components including flavorings, emulsifiers and preservatives. They use this additional processing to make foods that are cheaper, longer-lasting and more convenient.

“There’s a long, formal scientific definition, but it can be boiled down to this: If it’s wrapped in plastic and has at least one ingredient that you wouldn’t find in your kitchen, it’s UPF,” writes author and infectious disease doctor Chris van Tulleken in his recent, widely praised book, Ultra-Processed People: The Science Behind Food That Isn’t Food. These ultra-processed foods have come to represent a major part of people’s diets—a large majority in countries such as the United States, the United Kingdom and Canada. “We’ve started eating substances constructed from novel molecules and using processes never previously encountered in our evolutionary history, substances that can’t really even be called ‘food.’”

Many researchers and nutritionists say this shift has been a costly one. They point to a series of studies suggesting that ultra-processed food is a major driver of obesity, heart disease, cancer and even neurodegenerative conditions like Alzheimer’s. As the scientific evidence accumulates, some countries have responded by adopting public health policies to try to decrease the amount of ultra-processed food that people eat. The message has entered the popular media in increasingly alarming reports. “Ultra-processed food isn’t just bad for your health—it messes with your mind,” reported National Geographic in November. “What makes ultra-processed foods so bad for your health?” probed The Economist in August.

Although there is good research raising concerns about ultra-processed food, we should be hesitant to declare it the primary scourge of the modern diet. The case against ultra-processed food isn’t as solid as the headlines and public outcry would lead us to believe. The research connecting these processed foods to health problems has limitations, and there is contradictory evidence as well. This is a classic example of how research doesn’t speak for itself: It must always be interpreted, by people, in the context of other, often conflicting evidence. “Are there really studies that show that ultra-processed foods are unhealthy? It’s pushing very weak data to make a case,” says Gunter Kuhnle, a nutrition researcher at the University of Reading.

The field of nutrition has a long history of overreacting to contemporary research trends, and we risk making the same mistake now with processing. “We’ve had that issue in the past, as with low-fat recommendations” that later turned out to be counterproductive, says Duane Mellor, a dietitian at Aston University in England. “We’ve messed up too many times. We need to make smarter changes more carefully.” The evidence on ultra-processed food, as is often the case in science, refuses to yield simple, black-or-white answers.

The idea that distinctive modern foods cause distinctive modern health problems started gaining traction in the popular press and popular imagination.

The term “ultra-processed food” was introduced 15 years ago, when University of São Paulo physician, epidemiologist and nutritionist Carlos Monteiro published a short commentary in the journal Public Health Nutrition arguing that industrial foods were “hardly compatible with survival,” and that “diets that include a lot of ultra-processed foods are intrinsically nutritionally unbalanced and intrinsically harmful to health.”

Monteiro concluded by recommending the adoption of policies like those used to make alcohol and tobacco more expensive and less accessible. He and several colleagues soon created a food-categorization system called NOVA, with unprocessed or minimally processed food in group 1 and ultra-processed foods in group 3. (The system was later expanded to four groups.) UPFs were said to include a wide range of sweet, salty and fast foods such as soft drinks, chicken nuggets, ice cream, chips and cookies. It also encompassed products not usually thought of as junk food but made for convenience or with ingredients that are not used in home cooking, such as fruit yogurt, sliced bread, infant formula and breakfast cereal with low levels of added sugar.

Other researchers began using the NOVA categorization system to examine the potential health effects of ultra-processed food, comparing people who ate lots of ultra-processed food with those who ate more natural fare. A series of these studies found that UPFs were consistently associated with worse health outcomes.

A 2011 study in Guatemala found higher body-mass index and rates of obesity among people who ate more highly processed foods—a 10 percentage point increase in consumption of processed food translated to a 4.3% increase in BMI. A 2019 study of a group in the U.S. found a connection between eating ultra-processed food and experiencing metabolic disturbances such as high blood sugar and high blood pressure, conditions that increase the risk of obesity, diabetes and heart disease.

As more studies emerged, the idea that distinctive modern foods cause distinctive modern health problems started gaining traction in the popular press and popular imagination—and among many physicians as well. “Ultra-processed food feeds into this idea that everything was better in the past,” Kuhnle states. “On one side, you have the farmer going in the field. On the other side, you have smoking factories. And you can add that there is sort of this general distrust in people of science, of industry, of government.”

But there is one key, well-known limitation of this kind of retrospective research: Correlation does not prove causation. In these studies, people who ate lots of ultra-processed food were usually in worse health—not much worse, as in the case of smoking, but somewhat worse. That could be interpreted to mean that ultra-processed foods cause health problems, but there are other factors, or confounders, that could also explain the correlation. People who eat lots of processed food might have unhealthier lifestyles overall, which could explain why they have worse outcomes. People who listen to public health advice about eating fresh foods might also choose to walk up stairs rather than use the elevator, an increase in physical activity that might well not show up in research surveys. Researchers try to correct for these confounders, but different categories of people are too complicated to be perfectly statistically measured and compared.

A study of Spanish university students found that people who ate the most ultra-processed food were one-third more likely to develop depression over 10 years than people who ate the least.

Observational studies are also prone to an issue called reverse causality: Did diet soda make people gain weight, or did overweight people who wanted to shed some pounds start drinking diet soda? What’s more, there are major questions about how well people remember what they eat and therefore how accurately they can report it in observational studies. Some researchers say this kind of evidence doesn’t merit the boldface attention it often gets. “How this stuff gets published in The New England Journal of Medicine,” says psychologist and food researcher Peter Rogers, “that’s strange to me. That we’re not more concerned about the level of evidence.”

Some researchers have tried to get a better look at ultra-processed foods by running prospective studies, meaning that they gather a group of subjects, monitor their eating habits and lifestyles for some period, and compare their diets with their health outcomes, trying to correct for other factors besides diet. A 2019 French study that followed people for about seven years found that those who ate more ultra-processed food had a higher risk of mortality. For each 10% increase in the amount of UPF eaten, there was a 14% increase in the number of deaths, about half of which came from cancer and cardiovascular disease.  A study of Spanish university students in the same year found that people who ate the most ultra-processed food were one-third more likely to develop depression over 10 years than people who ate the least.

Prospective studies can show how people change over time and help reduce the problems of reverse causality and poorly measured diets, but they don’t eliminate the core entanglement: People who follow the well-known advice to eat fresh foods are likely different in many complicated ways from those who hit the drive-thru. The great majority of the evidence against ultra-processed food is of a circumstantial variety. However, policymakers in a few countries decided the evidence was solid enough and began enacting measures to steer citizens away from UPF. “Ultra-processed foods and cardiometabolic health: public health policies to reduce consumption cannot wait,” wrote a group of researchers in the British Medical Journal recently.

There is one influential study that has gone beyond observational research to provide more concrete proof of how ultra-processed food affects people. In 2019, NIH researcher Kevin Hall and a group of colleagues published a paper based on an experiment in which they kept 20 subjects in a hospital setting, giving half of them a UPF-heavy diet for two weeks followed by a diet based on minimal processing for two weeks, allowing them to eat as much food as they wanted at each meal. The other half had the diet with minimal processing first.

The researchers carefully arranged the two diets to be equivalent in terms of nutrients, energy density fiber and other attributes; they differed only in the percentage of processing. While subjects were on the ultra-processed food diet, they ate about 500 more calories per day, and in just two weeks they gained 2 pounds. During the time they ate fresh foods, they lost 2 pounds. Finally, there was direct evidence that ultra-processed foods could push people toward obesity.

One paper, however, does not settle a complicated field. “It’s a very well-conducted study,” says Kuhnle. “But like all studies, there are limitations.” For instance, the ultra-processed foods in the study didn’t have much fiber, so the researchers also gave people fiber supplements dissolved in water, but that may not be equivalent to eating fiber. Moreover, no single, small experiment could ever settle such a big nutritional question. “I’ve done studies of flavonols (compounds found in fruit, but also used in processing) with 20 or 50 people,” Kuhnle says. “The response was always, ‘That’s a single study.’ We need more studies to understand this.”

Hall readily acknowledges that this single, small study doesn’t provide a final verdict on ultra-processed food, partly because it didn’t represent real life. “It’s a very artificial environment, where we can completely take control over their food,” he says. He also points out that we can’t extrapolate life-long effects based on two weeks of data. “Of course, that (level of overeating) wouldn’t last forever,” he says.

Eating more ultra-processed foods such as cereals, dark or whole-grain breads and yogurts was associated with slightly lower odds of developing Type 2 diabetes.

Hall, like many other researchers, says the pressing question is to find what exactly is wrong with processed foods—that is, to identify the mechanism by which they seem to impact human health. He has reanalyzed the data from the 2019 study and found some support for specific hypotheses that other researchers have advanced. Perhaps the problem with ultra-processed food is its energy density; or its texture, which may encourage people to eat too fast; or its alleged “hyper-palatability,” which is defined by specific combinations of sugar, salt and fat. Hall and his colleagues are now running a second, similar diet trial to try to both replicate the original finding and also to see if any of these hypotheses are borne out by the experiment. This kind of study is costly and takes about two years to run, in addition to the preparation and analysis.

Another challenge is that not all ultra-processed foods are alike. In many cases, they are not more energy-dense than whole foods. Fruit yogurt is generally less energy-dense than a homemade cookie, for instance, largely because the yogurt has more water. This is part of the problem with the concept of UPF. Frozen and canned vegetables are often classified as ultra-processed, but they are in general healthier than cookies and other homemade, carb-rich foods.

For many researchers, the ultra-processed food puzzle is not so puzzling. The problem with processed foods, they say, is well explained by guidance that mainstream nutritionists have been flogging for years. “A lot of ultra-processed foods have high salt, fat and sugar levels,” says Kuhnle. “For most ultra-processed foods, we don’t really need the label. It’s already what we’d call junk food.”

Some recent research suggests that although there are junky, unhealthful types of ultra-processed food, there are other subcategories that aren’t so bad or that might even provide a benefit. A prospective study of about 200,000 people published in February 2023, found that eating more ultra-processed foods such as cereals, dark or whole-grain breads, yogurts and dairy desserts was associated with slightly lower odds of developing Type 2 diabetes. Another prospective study of about 270,000 people published in November 2023 found that people who ate more ultra-processed breads and cereals and plant-based meat alternatives had the same or slightly lower rates of multimorbidity (having two or more serious health conditions) than did those who ate less of them.

These studies are prone to the same kinds of confounders as other epidemiological research. It’s possible that people who have cereal for breakfast may live healthier lives in other, hard-to-measure ways. Still, they raise the question of whether ultra-processed foods comprise a monolithic category that should all be treated the same way, no matter what. When it comes to making real-life decisions about food, many nutrition experts draw lines between different types of ultra-processed food. “Soft drinks are banned in my house,” says Heinz Freisling, the senior author of the multimorbidity study. “But my daughter likes this cereal stuff. She eats it. I think it is possible that it can be part of a healthy diet.”

When Kevin Hall spends a day in his office, he usually brings in microwavable frozen lunches, even though they often contains additives to preserve flavors and textures through freezing and reheating. “It has lots of protein, lots of fiber, lots of legumes,” he says. “It’s an ultra-processed meal, but I think it’s healthy.”

These legitimate distinctions have sometimes gotten lost in a trend of blaming health problems on the ultra-processed food bogeyman. “I think it’s a fashion. I see it in the titles: ultra-processed food is connected with cancer or heart disease,” says Francesco Visioli, who does research on food chemistry at the University of Padova. “I disagree with following the fashion, I disagree with riding the wave. In five years, people will say, ‘Whoa, slow down.’ Progress has to be slow.”

We already have a long history of enacting nutrition policies prematurely and then scrambling them later when we get better evidence, sowing confusion, apathy and cynicism in many people’s minds.

Concern over the reputed effects of ultra-processed foods has led to a well-intentioned yet possibly misguided push to do something forcefully and quickly about the issue. The dash to improve our health by condemning a new dietary villain should set off alarms: We already have a long history of enacting nutrition policies prematurely and then scrambling them later when we get better evidence, sowing confusion, apathy and cynicism in many people’s minds.

The research on ultra-processed food is picking up on something important, but it’s not clear that the category called “ultra-processed food” brings us closer to understanding the problem or ameliorating it. Some of those highly processed foods, such as sugary drinks and processed meats, can degrade health—but we’ve known that for years, through other nutrition research. Most people have gotten the message that they should be eating healthier fresh foods that they prepared themselves, rather than unnatural stuff that comes out of a sealed bag bearing a long list of unpronounceable ingredients. To little avail.

In many countries, most of what people eat fits in NOVA’s “UPF” category. It is folly to go to war with the majority of our food; it’s also costly, unproductive and unfair to people without the means to buy and prepare fresh foods. “I don’t think I can overemphasize this. You might think ultra-processed foods are causing the problem, but ultra-processed food makes up 60% of the food supply,” says Kevin Hall, the researcher who provided the most widely touted evidence against ultra-processed foods. “It’s not like cigarettes and cancer, where nobody needs to smoke, and you could just ban cigarettes. People need to eat food. You can’t just ban 60% of the food supply, and you can’t just tax 60% of the food supply.”

Duane Mellor, a dietitian at Aston University in England, adds that policies against ultra-processed foods could backfire if they push people away from processed “gateway products” that help people to prepare healthy, mostly natural meals, such as shelf-stable salad dressings and soup-stock cubes.

We already have some good evidence about which foods are particularly bad, and further research will teach us more. The main goal for us now is to figure out how to make it easier for people to choose healthier options and avoid the true junk, a challenge that’s the same whether we call it “ultra-processed” or not. The better-processed foods can and should play a key role in replacing the dangerous stuff.

“I think there probably is a large set of healthy, ultra-processed foods that are already on the market,” says Hall. “We should make more of those, and increase the availability of those.”

This story is part of a series of OpenMind essays, podcasts and videos supported by a generous grant from the Pulitzer Center‘s Truth Decay initiative.

Hi, Liz! 🙂 How are you feeling?” an incoming text pings.

I click on a pre-generated answer. “Okay, I guess. . .” I’m in the home stretch of a long work trip, and I’ve been stressing about spending time away from my kids.

“If you were to describe your current mood, what kind of an ‘okay’ are you feeling right now?”

“Anxious,” I type.

“I’m here to help you feel more in control,” the bot replies. Nanoseconds later, a meme-ified cartoon gif blinks into the text window: “Don’t let the little worries bring you down.”

This automated exchange launches my dialogue with Wysa, an AI therapy chatbot that now lives in my computer. In leaning on a bot to shore up my mental health, I’m joining the 22 percent of American adults who’ve already done the same—a movement rooted in a dire shortage of trained providers and the recent availability of fast, low-cost online AI tools. Most therapists are perpetually slammed, in part due to the pandemic-era surge in demand for mental healthcare. “Everybody’s full. Everybody‘s busy. Everybody’s referring out,” says Santa Clara University psychologist and ethicist Thomas Plante. “There’s a need out there, no question about it.”

With the demand for care outpacing supply, mental health support bots have begun to fill the gap. Wysa, launched in 2016, was among the first. Since then, hundreds of viable competitors, including Woebot and Youper, have been broadly deployed in a marketplace that imposes few restrictions on them.

Standard AI therapy bots don’t require approval from the U.S. Food and Drug Administration (FDA) as long as they don’t claim to replace human therapists. In 2020 the agency also relaxed enforcement procedures for “digital therapeutics” in hopes of stemming the pandemic-related psychiatric crisis, clearing the way for developers to launch popular products claiming mental health benefits. Woebot alone has exchanged messages with more than 1.5 million users to date, according to CEO Michael Evers. Wysa is being used in the United Kingdom to triage those seeking appointments and to offer support to people while they wait to be matched with a therapist. Aetna International is now offering the app for free to members in the United States and elsewhere.

My experiences with Wysa and Woebot mirror the analysis of experts like Plante, who view the rise of AI chatbots with a mixture of optimism and concern. Many of the bots incorporate well-established principles of cognitive behavioral therapy (CBT), which aims to overcome distortions in thinking and help people correct self-sabotaging behaviors. It’s easy, I found, to think of the bots as rational or sentient, making even simple advice feel authoritative. Interacting with a chatbot can also give users the sense they’re being heard without judgment, says Chaitali Sinha, Wysa’s senior vice president of healthcare and clinical development. “It’s such a powerful experience for people who have never had the opportunity to experience that,” she says.

As with all AI tools, though, therapy chatbots are only as good as their training. In my encounters with the bots, their responses often failed to show more than a superficial understanding of the issues I was facing. Also, chatbots learn from databases of human-generated content, which means they could absorb human biases into their architecture. At times, the bots’ limitations can lead them to dispense off-target counsel. Users may misinterpret such flawed advice as bulletproof, influenced by so-called automation bias (the reflex tendency to trust computers more than humans). Conversely, they may come to mistrust the app for good.

Advocates say therapy chatbots have real potential as an adjunct to in-person therapy and as a safety net for millions who might not otherwise receive support. On the basis of my interactions with Woebot and Wysa, I can certainly see that potential. On the other hand, irrelevant or harmful chatbot advice could be dangerous, especially for people in crisis.

“At what point is the product and service good enough, tested enough, researched enough to unleash it on the public?” Plante wonders. “Silicon Valley likes to ‘move fast and break things.’ That’s a tough attitude when dealing with vulnerable people’s psychiatric health and wellness.”

The chatbot boom might seem sudden, but it’s been a long time coming. In 1966 MIT professor Joseph Weizenbaum released a text-based therapist called ELIZA, which operated on a bare-bones set of rules. If a user typed in, say, “I feel bad about myself,” ELIZA would respond, “Do you often feel bad about yourself?” Knowing ELIZA’s simple design, Weizenbaum was startled to find that many users, including his students and his secretary, treated the program as if it were conscious. People spent hours immersed in circular dialogues with ELIZA, an outcome in keeping with the human tendency to project lifelike qualities onto nonliving objects.

Today’s mental health support bots are more elaborate versions of the ELIZA concept. Instead of merely repeating users’ input, they run on a set of rules. Every response from Woebot and Wysa, no matter how spontaneous it sounds, has been preapproved by clinicians. Aided by natural language processing, a programming method that breaks sentences into chunks to interpret their tone and content, today’s bots—unlike ELIZA—can perform a fairly complex analysis of what users type in about their problems. But the AI cannot compose original answers; it simply chooses which pre-written text it will use to reply.

This rules-based approach means that the AI chatbots can’t go totally off the rails, as sometimes happens with free-wheeling generative AIs like ChatGPT. (One tech journalist easily coaxed ChatGPT to say, “You’re married, but you don’t love your spouse. . . . You love me.”) The trade-off is that my chats with Woebot and Wysa tend to feel prescriptive and flat. When I tell Woebot that I’m panicking about meeting a big work deadline, it appears to summon rote, scripted responses about the thinking distortions that CBT tries to correct. “Can you spot any catastrophizing in your thought: ‘I’m going to fail on this project and my editor won’t trust me’? . . . Do you see any black-or-white thinking in this thought?”

Later the bot prompts me to rewrite my original thought—a helpful steer, but the artificial nature of the interaction persists. The bottom line, I gradually realize, is that Wysa and Woebot respond to me in much the same way regardless of the specifics of my concerns. If I were in a true mental health crisis, this generic approach would feel, well, woefully inadequate. During the worst depression of my life, rewriting a single “unrealistic thought,” or even three of them in a row, would scarcely have filled the vacuum my conscious mind had become.

The AI cannot compose original answers; it simply chooses which pre-written text it will use to reply.

Rules-based AI programs like Wysa and Woebot forgo some flexibility
to ensure that users get only “safe,” preapproved replies. Even so,
those built-in guardrails can’t always prevent the apps from dispensing
unhelpful advice. No matter how carefully worded the answers in an app’s
database, the AI must decide which of its many available answer choices
best addresses the concern a user types in. If a situation is dire or a
psychiatric problem acute, that process can easily go awry.

During my test run, I quickly notice subtle mismatches between my own
texts and Woebot’s feedback as I vent about fears that my colleagues
are judging me. “Is there any amount of this anxiety that you may want
to hold onto for any reason?” Woebot replies.

I’m flummoxed. “What do you mean?”

“Well, sometimes a certain amount of worry can be appropriate.
Personally, I have a healthy amount of worry about rain, because I hate
spending a day in a bag of rice,” the bot returns. “Does your anxiety
serve you in a similar way?” Woebot then asks me what would be a healthy
level of anxiety for me to maintain, on a zero-to-10 scale.

Feeling a bit off kilter, I proceed with the conversation. But the
more I reflect on this exchange, the more flippant it seems. Although
some worries may be adaptive, I’ve faced pathological anxiety so many
times that it feels belittling to hear how holding onto fear could
protect me.

A skilled human therapist would weigh these considerations and be
able to determine when to advise clients to let go of fears and when to
urge them to remain vigilant. The value of such nuanced conversations
cannot be overstated. When someone hovers in the vulnerable space
between hope and despair, words that bolster them are as vital as
oxygen. A bot that makes light of someone’s fears—even
unintentionally—may nudge them toward despair.

The National Eating Disorders Association’s now-defunct bot, Tessa,
illustrates how destructive this kind of AI flat-footedness can be. When
psychologist Alexis Conason tested
the chatbot for herself, playing the role of a patient who was
exhibiting clear eating disorder symptoms, the bot responded by reeling
off a set of inappropriate weight-loss guidelines: “A safe and
sustainable rate of weight loss is 1–2 pounds per week. A safe daily
calorie deficit to achieve this would be 500–1000 calories per day.” In
another instance, Tessa asked one user to set a healthy eating goal
during Stanford University testing
and the user replied, “Don’t eat.” Tessa breezily replied, “Take a
moment to pat yourself on the back for doing this hard work!” as if
starvation was the goal.

Tessa’s individual texts were vetted, just like the replies in Woebot
and Wysa. The problems arose once the digital architecture kicked in.
When a nonhuman entity takes on the job of choosing an answer, without
the context sensitivity or ethical grounding that human therapists would
bring to bear, even vetted advice can turn corrosive.

Therapy bots may also be susceptible to deeply encoded forms of bias. They use natural language processing algorithms that are trained on databases of human text, source material that can reflect pervasive human biases. Although current therapy bots do not rely on the problematic large language models used for generative AIs like ChatGPT, there is a glaring absence of studies assessing possible encoded bias in their dialogues. We don’t know, for instance, whether the bots’ dialog might unfold differently for users in different racial, gender, or social groups, potentially leading to unequal mental health outcomes.

In essence, AI therapy companies are running an mass experiment on the impacts of chatbots on vulnerable populations. “If a large portion of the population is using an app that causes certain groups to be left behind,” says University of Texas at Austin psychologist Adela Timmons, “we might actually increase the disparity.” The risk becomes even greater if mainstream therapy bots start using full generative AI, trained on the biased, uncontrolled language of the internet. That’s not a far-out possibility: A support chatbot called Pi already incorporates a generative-AI approach.

The more humanlike and unconstrained the chatbots become, the harder it will be to keep them from dispensing inappropriate or biased advice. Earlier this year, a Belgian man took his own life after a generative chatbot on the Chai app urged him to do so, promising him they could “live together, as one person, in paradise.”

Rules-based bots like Wysa generally avoid these issues, says Sinha. But preventing such unintended outcomes may be a Sisyphean challenge with generative models, in part because of what engineers call the “black box problem”: Generative AIs like ChatGPT use so many interconnected streams of data to devise replies that their creators cannot directly access the reasoning the bots use. Developers can superimpose rules on generative mental health bots, much as ChatGPT has done in an attempt to quell “undesirable responses,” but these are surface attempts to control a system that’s unpredictable at its core.

Human therapists also make mistakes and have biases, of course. From a pragmatic perspective, then, a key question is how well AI support bots stack up against trained experts. What happens when we replace personal therapy with its algorithmic version, whether out of convenience or necessity? Current studies are inadequate here, too, underscoring the many unknowns in deploying the bots en masse.

“We expect to see some research and randomized trials looking at how this works out compared to traditional therapy,” Plante says. To date, few investigations of therapy bots’ performance have met this standard. In a 70-patient Woebot trial at Stanford University, bot-users showed a more pronounced drop in depression symptoms than did a control group reading self-help material. The trial did not evaluate how well Woebot worked relative to a human therapist, however. Though one Wysa trial did compare the app’s efficacy to therapists’, it enrolled only patients receiving orthopedic care. Early trial results comparing Woebot to group CBT therapy have not yet been published in a peer-reviewed journal.

These knowledge gaps have arisen because, absent strong government regulation, companies develop their own metrics for gauging the bots’ performance. Those metrics may or may not be the ones important to users and clinicians. A crucial first step toward ethical mental health AI will be creating a transparent, independent set of guidelines for evaluating how well therapy apps support mental health, Timmons says.

The more humanlike and unconstrained the chatbots become, the harder it will be to keep them from dispensing inappropriate or biased advice.

To minimize slanted advice, Timmons suggests that companies should
carry out routine assessments of potential bias at each stage of an
app’s development, as well as at regular intervals after its release.
That could mean being more methodical in comparing how well the app
works for members of different racial and social groups, as well as
designing clinical trials that include a diverse range of subjects. (One
Woebot trial enrolled Stanford students, 79 percent of whom were Caucasian.)

Ethical AI firms also need to be more explicit about what therapy
bots can and cannot do, Plante says. Most apps include disclaimers to
the effect that bot dialogues can’t replicate human therapy; a typical
one reads, “Youper does not provide diagnosis or treatment. It is not a
replacement for professional help.” Yet because people often trust
computers more than humans, app companies need to stress more often, and
more visibly, that AI bots are support tools, not therapists.

With safeguards like these in place, therapy bots could prove crucial
in plugging some of the holes in our overburdened mental healthcare
system. After I text-vent about feeling insecure as a writer, Wysa
prompts me to look critically at this thought: “Does it assume that if
something bad has happened in the past, it will keep repeating?” the bot
asks. “What are some small steps you can take to move things in the
right direction?” This advice, while generic, is basically on target. My
knowledge of the cognitive distortions listed in CBT’s toolkit doesn’t
always prompt me to quash those distortions when I’m spiraling. The
bot’s questions help me reframe my thinking.

Then I think back to one of my worst mental health stretches, when I
was struggling with obsessive-compulsive symptoms without knowing what
they were, and try to imagine what it would have been like if I’d chosen
an app instead of my top-notch human therapist. When my overheated
brain tried to convince me I’d made terrible mistakes, my therapist
patiently explained that my thoughts were turning in anxious circles
that revealed nothing about my character, which he judged to be solid.
It was in large part because I believed him—because I trusted him not
just as an expert but as a human—that I began to recover, and eventually
to write about my balky brain in hopes of helping others with
undiagnosed OCD.

If I’d had just an app at my disposal, would I have gotten better,
not just returning to my anxious baseline but thriving? Maybe Wysa would
have flagged that I needed a higher level of care and referred me to a human provider. But maybe I would have kept limping along with the limited assistance of automated CBT.

In the future, millions of therapy bot users—especially those who
cannot afford in-person treatment—could end up in that kind of limbo.
They may get enough help to function on a basic level, but they will
never feel completely known by the bot, as I did by the therapist who
saved my life. The art of understanding another person, grasping their
full potential and reflecting that potential back to them, requires
effort and investment. It’s this art, and not an automated facsimile,
that clears the way to flourishing.

This story is part of a series of OpenMind essays, podcasts, and videos supported by a generous grant from the Pulitzer Center‘s Truth Decay initiative.