The survey, conducted in French, Bangla, Spanish, and English, reached workers through recruitment led by WJP staff. Of 805 workers approached, 618 responded and 561 eligible workers consented and completed the questionnaire. The researchers are careful to note that the findings describe the workers surveyed and should not be interpreted as statistically representative of the entire citywide workforce. Even so, they point out that several results align with the most recent data and anecdotal evidence on the composition and experience of New York’s delivery workforce, a labor pool dominated by immigrants who depend almost entirely on platform income to support their families.

The numbers on algorithmic control are striking. More than 80 percent of respondents said the food delivery apps mostly or entirely determine how quickly orders must be completed, which routes workers take, and how many hours they work. That level of control, the authors argue, translates directly into risky behavior on the street. Two-thirds of workers said they rush to deliver on time in order to maintain high ratings, 56 percent said they work in bad weather, and 30 percent said they work even when they feel very sick. Each of these behaviors, from speeding between restaurants and apartments to riding through storms while ill, represents a pathway by which platform-managed incentives become physical harm.

The injury statistics quantify the consequences. Fifty-three percent of respondents reported a work-related injury, a figure the researchers convert to 48 injuries per 100 full-time equivalent workers, more than five times the national rate for delivery messengers. Forty-six percent of all respondents reported an injury serious enough to require medical care or to cause them to miss work. Compounding the danger, 30 percent of the workers surveyed said they lack health insurance, meaning that a single serious crash or assault can translate into unpayable medical debt on top of lost wages. For a workforce that keeps the city fed through all weather and all hours, the occupational safety net is thin to nonexistent.

Mental health outcomes were equally severe. Forty percent of surveyed workers met the criteria for severe psychological distress, a state of overwhelming mental burnout that requires urgent mental healthcare. That rate is five times the rate among adult New Yorkers overall and ten times the rate among working-age adults nationally. The authors emphasize that this distress is not an incidental byproduct of hard work but is shaped by the structural conditions of platform labor: intense work demands, little control over how the job is organized, and constant uncertainty about whether access to work will continue from one day to the next.

Violence and theft add further layers of risk. Thirty-six percent of surveyed workers reported having been physically assaulted while working, and 86 percent reported theft of equipment or belongings, with losses averaging about 2,100 dollars. For workers earning their living primarily through app-based delivery, the loss of an e-bike or a phone is not merely an inconvenience; it can mean the temporary loss of the ability to work at all, deepening the financial precarity that the survey documents throughout.

Perhaps the most consequential mechanism the study identifies is the threat of deactivation, the practice by which platforms cut off a worker’s access to the app and therefore to income. Fifty-two percent of workers said they worried about being deactivated if they did not deliver fast enough. Seventy percent had contacted an app about a deactivation issue in the previous month, but only 37 percent of those workers said the issue was resolved satisfactorily. When asked what they most wanted to change about their jobs, 56 percent chose protection from deactivation, more than any other option. Workers were especially supportive of two specific reforms: resolution of deactivation cases within 24 hours and 14 days’ advance notice before deactivation takes effect.

The dependence on platform income explains why deactivation looms so large. Ninety-six percent of surveyed workers said app delivery was their only job in the previous month, 98 percent earned at least half of their income from delivery, and 53 percent said delivery was their family’s only source of income. In practical terms, a deactivation notice is not a disciplinary footnote; it is the sudden removal of a household’s livelihood, often with little explanation and no timely avenue for appeal. The researchers argue that this precarity should be understood as an occupational exposure in its own right, one with measurable consequences for mental and physical health.

The study’s authors frame the findings in explicitly occupational health terms. “The toll we documented shows up both in injuries and in mental health,” said Mustafa Hussein, assistant professor at CUNY SPH and lead author on the brief. “More than half the workers we surveyed had been hurt on the job, and 40 percent have severe psychological distress. Understanding what drives those outcomes, including how the apps organize and monitor work, is essential to protecting the people who keep this city fed.” Senior author Nevin Cohen, professor at CUNY SPH and director of the CUNY Urban Food Policy Institute, added that when an app can cut off a worker’s income with little explanation and no timely way to resolve it, that constitutes a health issue as well as a labor issue, and that workers told the team clearly what they need first: deactivation cases resolved within 24 hours and advance notice before they lose access to their jobs.

Senior author Sherry Baron, professor at the Barry Commoner Center for Health and the Environment at CUNY Queens College and an occupational health physician and researcher, noted that from an occupational health perspective the risks these workers face are not limited to what happens on the street, and that intense work demands, little control over how the job is organized, and constant uncertainty about keeping access to work all shape workers’ health and deserve the same attention as more visible workplace hazards. Ligia Guallpa, executive director of the Worker’s Justice Project and a co-author, said delivery workers have long described how much is at stake when an app can limit or end their access to work, and that the findings show how widespread those concerns are and why workers want a clear, timely process when deactivation occurs. The brief was prepared by a team including Francesca Michielli, Mustafa Hussein, Sherry Baron, Nevin Cohen, Maria Figueroa, Ligia Guallpa, Emilia Vignola, Dana Calacci, Rositsa Ilieva, Yadira Sanchez, Zoey Laskaris, and Jacquelyn Sullivan, and represents a collaboration among CUNY SPH, the CUNY Urban Food Policy Institute, the Barry Commoner Center at Queens College, York College, the Worker’s Algorithm Observatory, the Worker’s Justice Project, and the Rockefeller Institute of Government, with support from the Robert Wood Johnson Foundation. As cities and states continue to debate minimum pay standards and due-process protections for gig workers, the study offers a detailed evidence base connecting algorithmic management to concrete health outcomes, and it gives worker advocates a specific, testable set of demands: fast resolution of deactivation cases and advance notice before access to work is terminated.

Subject of Research: Occupational health and algorithmic management of app-based food delivery workers in New York City

Article Title: NYC delivery workers injured at more than five times the national rate under intense labor control by the apps, survey finds

Article References: NYC delivery workers injured at more than five times the national rate under intense labor control by the apps, survey finds. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: gig economy, delivery workers, occupational health, algorithmic management, deactivation, psychological distress, workplace injury, New York City, platform labor, CUNY SPH, Worker's Justice Project, labor rights

The research team, led by Min Seok Kwon and corresponding author No-Kuk Park of Yeungnam University, together with colleagues from the Institute of Clean Technology and Inwoo Eco Corporation, set out to understand precisely why manganese-oxide catalysts lose their appetite for ozone and then to compare two fundamentally different regeneration strategies. Their choice of material was deliberate. Manganese dioxide sits among the most effective and inexpensive catalysts for splitting ozone into ordinary oxygen at ambient conditions, which makes it attractive for small-scale VOC treatment systems where energy budgets are tight and elaborate heating equipment is impractical. Yet manganese oxides carry a well-known weakness: they are highly susceptible to moisture, and their performance decays steadily during prolonged use.

To build their catalyst, the researchers employed a potassium permanganate-assisted precipitation route. Aqueous manganese nitrate was combined with potassium permanganate in a 3:2 molar ratio at 80 degrees Celsius, a slow addition that drove oxidation and precipitation of manganese oxide from solution. After roughly 24 hours of aging, the precipitate was filtered, washed three times with deionized water to strip away residual nitrate and unreacted permanganate, dried at 110 degrees Celsius overnight, and pulverized. The powder was then shaped by extrusion with methyl cellulose and colloidal silica binders into cylindrical pellets a few millimeters across, later ground and sieved to particles between 300 and 800 micrometers for testing. X-ray fluorescence confirmed a product dominated by manganese oxide, roughly 93.2 weight percent MnO with about 4.6 percent SiO2 carried in from the binder. Crucially, because the synthesis avoided high-temperature calcination, the resulting catalyst remained only partially crystalline, a structure rich in lattice defects and oxygen vacancies that previous work has linked to exceptional ozone-decomposition activity.

The deactivation experiments told a vivid story. When about one gram of the catalyst was loaded into a tubular reactor and fed roughly 15 ppmv of ozone in air at room temperature and atmospheric pressure, the outlet ozone concentration plummeted to zero. Complete decomposition held for approximately 40 minutes before ozone began to creep back into the effluent, ultimately stabilizing near 8 ppmv, equivalent to about 50 percent conversion. This classic breakthrough curve is the fingerprint of a catalyst whose active sites are being progressively occupied. The mechanism behind ozone destruction on manganese oxide proceeds in three steps: ozone adsorbs onto the surface and splits off an adsorbed atomic oxygen species, a second ozone molecule reacts with that atomic oxygen to form two adsorbed molecular oxygen species, and finally those molecular species desorb as gaseous oxygen, freeing the site for another cycle. At low ozone concentrations, that third step, desorption, becomes the bottleneck. When desorption lags, oxygen intermediates pile up on the surface like cars in a parking garage with no exit, sealing off the very sites the reaction depends on.

Temperature-dependence experiments reinforced this interpretation. When the team gradually heated a deactivated catalyst, ozone removal climbed from partial conversion at 30 degrees Celsius to essentially complete destruction at 40 degrees Celsius, with no ozone detectable at the outlet above that point. Thermal energy, in other words, helps the trapped oxygen intermediates escape, temporarily re-exposing active sites. But definitive evidence came from X-ray photoelectron spectroscopy, or XPS, which probes the chemical state of the outermost atomic layers. On the fresh catalyst, the oxygen 1s spectrum was dominated by a lattice oxygen peak near 529.9 electronvolts. On the deactivated sample, a feature near 532 electronvolts, corresponding to surface-adsorbed oxygen species, swelled to encompass roughly 72 percent of the total spectral envelope. The poison was not some exotic contaminant; it was the reaction’s own intermediate, hoarded on the surface until the catalyst choked on its product.

With the deactivation mechanism established, the team tested the conventional remedy: thermal treatment. Deactivated catalyst samples were heated in an electric furnace at 100 to 600 degrees Celsius in 100-degree intervals for four hours each, at a controlled ramp of 5 degrees per minute, and then re-examined by X-ray diffraction and re-run in the ozone reactor. Between 100 and 500 degrees Celsius, the XRD patterns remained broad and weak, the signature of amorphous or partially crystalline phases studded with defects and vacancies, including traces of tunnel-structured epsilon-MnO2 that readily hosts oxygen vacancies. But above 600 degrees Celsius the picture changed dramatically: sharp diffraction peaks of alpha-Mn2O3 appeared, marking crystallization into a stable cubic phase with few vacancies and a high desorption energy for oxygen. In effect, the intense heat welded the catalyst into a form that no longer wanted to work.

The activity measurements tracked this structural narrative precisely. The fresh catalyst had a breakthrough time of about 40 minutes; samples treated at 100, 200, 300, 400, 500, and 600 degrees Celsius held out for approximately 8, 15, 23, 26, 22, and 3 minutes respectively. Expressed as recovery of initial activity, treatments at 100 and 200 degrees Celsius managed only about 20 and 37.5 percent, while the 300-to-500-degree-Celsius window consistently delivered around 57.5 percent. Beyond that, performance collapsed: catalysts treated at 500 and 600 degrees Celsius allowed steady-state outlet ozone concentrations of 9.5 and 11 ppmv, far above the roughly 8 ppmv seen after milder treatments. XPS confirmed that thermal treatment did strip away much of the adsorbed oxygen responsible for poisoning, but the accompanying loss of lattice defects at high temperatures meant that even a clean surface had fewer places to work. Thermal regeneration, the study showed, is a compromise: hot enough to free the surface, but not so hot as to crystallize the active phase.

The more surprising result came from the alternative strategy. Instead of heating the catalyst to hundreds of degrees, the researchers flowed a dilute mixture of 10 percent hydrogen in nitrogen over the spent catalyst at just 40 and 60 degrees Celsius. Embedded thermocouples revealed the chemistry happening in real time: the catalyst bed temperature fluctuated by up to 0.15 degrees Celsius at 40 degrees and within a 0.65-degree band at 60 degrees, transient swings produced when adsorbed oxygen species oxidized the incoming hydrogen to water in small local exotherms. Those gentle temperature excursions are direct evidence that the adsorbed oxygen, the poison itself, was being selectively consumed. After hydrogen exposure at 40 degrees Celsius, the catalyst regained about 25 percent of its initial activity; at 60 degrees Celsius, recovery reached approximately 55 percent, achieved with a fraction of the energy demanded by thermal treatment.

Spectroscopy sealed the case for selectivity. After hydrogen reduction, the 531.6-electronvolt XPS peak assigned to adsorbed oxygen dropped sharply, falling to an area fraction of 26.31 percent at 60 degrees Celsius and 29.61 percent at 40 degrees Celsius, while the lattice oxygen and oxygen vacancy signals remained essentially untouched. The hydrogen had removed the poison without disturbing the underlying active structure, avoiding the crystallization and defect loss that plague high-temperature regeneration. Intriguingly, catalysts regenerated by hydrogen outperformed some heat-treated samples even when their residual adsorbed-oxygen levels were comparable, because the hydrogen-treated surfaces retained their defect-rich, vacancy-laden architecture. The authors also noted that the reaction rate in ozone decomposition is governed by the interplay between adsorbed atomic oxygen and ozone forming adsorbed molecular oxygen, and that hydrogen regeneration short-circuits the poisoning cycle by converting the trapped species directly to water. For an industry searching for ways to extend catalyst lifetimes in compact, low-energy air-purification systems, the message is clear: sometimes the gentlest treatment, a warm whisper of hydrogen rather than a blast of furnace heat, is precisely what a poisoned catalyst needs to breathe again.

Subject of Research: Regeneration of deactivated manganese-based ozone decomposition catalysts by thermal treatment and hydrogen reduction

Article Title: Optimization of regeneration conditions for ozone decomposition catalysts via thermal treatment and hydrogen reduction

Article References: Kwon, M. S., Jang, J. G., Kim, M., Kwon, B. C., Park, N.-K., Kong, J. S., & Kong, S. W. (2025). Optimization of regeneration conditions for ozone decomposition catalysts via thermal treatment and hydrogen reduction. Advances in Industrial and Engineering Chemistry, 1(1), Article 29. https://doi.org/10.1007/s44405-025-00030-z

Image Credits: AI Generated

DOI: 10.1007/s44405-025-00030-z

Keywords: ozone decomposition, manganese catalyst, catalyst regeneration, hydrogen reduction, thermal treatment, XPS, adsorbed oxygen, oxygen vacancies, VOC treatment, air purification, MnO2, deactivation