Monday, February 24, 2020

Here’s What Happens When an Algorithm Determines Your Work Schedule

In the retail, hospitality, and food service industries, employees often work long and unpredictable hours—increasingly scheduled by algorithms that analyze seasonal sales patterns, customer trends, and even the weather to determine their shifts. These algorithmic systems have come under heavy scrutiny in recent years, pressuring employers and software vendors to present workers with more choice and flexibility.

But the workers who are beholden to these automated systems tell a different story. Many of the core issues with the software are still widespread, along with a lack of transparency on its inner workings. Despite the negative attention, automated scheduling systems are affecting an increasing amount of the retail and food industry workforce—which includes roughly one in five workers in the US, according to the Bureau of Labor Statistics—and it’s making them miserable.

Many large retailers and restaurant chains in the U.S. and abroad have deployed software that uses a variety of data to fine-tune and automate their employee’s schedules, according to press releases, public statements, and employees interviewed by Motherboard. These systems are designed to track employees’ time and activity to cut labor costs and maximize profitability for shareholders.

"Because there isn't a way to guarantee week-to-week that I [could have] certain days off depending on what the system dictates, it makes it really hard to plan anything outside of work,” Kyle, a former full-time worker at Target, told Motherboard. Motherboard granted Kyle and other sources in this story anonymity because they feared retaliation from their employers. “As far as a social life is concerned, all of my activities became spontaneous. Planning anything for me became months and months in advance.”

For decades, giant retailers have deployed software that tracks employee hours from companies like Kronos, a longtime leader in HR management systems. In addition to hourly workers, Kronos and its competitors like Zoho & Salesforce sell products that are also used in salaried jobs, as well as police departments, municipal governments, hospitals, and university systems. In recent years, this trend has expanded to custom apps marketed towards small-to-medium sized businesses, with names such as Deputy, Harvest, and ZoomShift. This has led to an explosion of revenue for these companies, and the industry is expected to grow from $250 million in 2019 to $530 million in 2024.

Navigating the system can be especially difficult around the holidays, because it often fails to consider things that a human manager would. “The software doesn’t look at the other 51 weeks in a year and know that you haven't seen your family all year,” said Kyle. “A human in charge of scheduling can.”

A Target spokesperson told Motherboard that the company offers a “variety of schedules to meet the diverse needs of our team” and “match their availability to the needs of our guests.”

Research has shown that erratic scheduling practices are associated with worse sleep quality and higher rates of depression, even for full-time workers who receive benefits. For part-time workers, this can affect their lives in troubling ways.

“I felt like my life source was being tapped out of my body. I felt like I should just suck it up and be better,” Callen, a student and Starbucks worker, told Motherboard.

They said that management used Starbucks’ scheduling software to prevent shorter shifts during the week in favor of longer weekend shifts.

“I live three hours away from [my family] and I can only return home on weekends. Last semester, I only saw them three times,” Callen said.

This illustrates how these systems rely on forcing workers to have a 7-day “open availability,” a common aspect of at-will employment contracts used in the service industry.

Scheduling software works by cross-checking employee availability against business needs for different tasks. The system then produces a schedule for each employee that is fine-tuned down to the minute, with break times pre-determined to maximize productivity. Customer-facing data is also used to schedule when workers are needed most, such as online appointment scheduling—something that has skyrocketed in recent years, with up to 66 percent of healthcare companies using digital self-booking to manage appointments.

These systems were the subject of a 2014 New York Times article, which detailed how algorithmic scheduling turned one Starbucks worker’s life into an intricate scheduling ballet, improvising childcare with “on-call” shifts being added same-day or dropped mid-shift. These types of practices were previously limited to specific industries such as emergency healthcare, but were widely implemented in retail after the Recession due to the increase in workers willing to perform cheap labor.

After the New York attorney general’s office questioned retailers’ use of on-call staffing practices in 2015, many retailers voluntarily changed their policies, and the software companies started to do damage control.

”The populist view is that scheduling is evil, in that it's causing erratic schedules for employees, and so forth,” Charlie DeWitt, Kronos’ vice president of business development, told Buzzfeed News at the time. “The fact of the matter is it's an algorithm. It does whatever you want it to do.”

DeWitt has a point: employers are ultimately responsible for the well-being of their employees. They could aggressively micro-manage their employees without software, and conversely, they could easily implement these tools in a more equitable way, as some companies have. However, because these algorithms are effectively a black box, we still don't know today what considerations are made in favor of workers, if any.

When asked for comment, Kronos told Motherboard that it cannot comment on “specifics regarding customer implementations” of its products. No major software company has publicly subjected their algorithms to an outside audit to identify any potential issues—and claiming they can’t control how their clients use their products is an increasingly common way of dodging responsibility for intentional decisions built into their tools.

“The inner workings of these systems are largely shrouded in corporate secrecy, which only puts more power in the hands of corporations at the expense of workers and the public overall,” Andrea Nill Sanchez, Executive Director of the AI Now Institute, a group that researches the social implications of AI, told Motherboard. “Worse yet, corporate secrecy is a structural characteristic of the AI industry in general that undermines our ability to assess the full extent of how the technology is being used and hinders redress, oversight, and regulation.”

With this lack of transparency, white-collar developers are effectively building systems that allow companies to exercise greater control over blue-collar workers, with people of color, and especially women of color, being at higher risk of schedule uncertainty, according to a study by researchers at the University of California, Berkeley. (Ironically, Kronos itself has consistently been rated one of the best places to work and has an unlimited vacation policy for its own employees.)

While workers at small businesses can work with management to figure out a schedule that works best for them, it’s often less equitable within large companies. That’s because when shifts are scheduled this tightly, any request to change one employee’s schedule can cause a snowball effect that throws the system off balance. Workers then negotiate amongst themselves to swap shifts using internal apps, but since most workers’ availability is already coded into their schedules, many will be busy with other responsibilities such as childcare or education, making it even more difficult for hourly workers to control their free time.

This can lead to calling out and losing pay, or using sick time unnecessarily.

“I wasn’t able to easily swap shifts with coworkers,” Callen said. “We were understaffed and my coworkers weren’t always available. The system could [easily] accomodate a shorter shift.”

A Starbucks representative told Motherboard that the company cannot comment on the specifics of the technology it uses, but said that managers use software to create “predictable and consistent” schedules.

Lost pay can also come from sudden scheduling changes that are often tied to sales figures. “The main [reason] is the micromanaging of hours that the software does,” James, a worker at Schuh, a UK clothing retailer, told Motherboard. He noted that hours are automatically calculated based on how many pairs of shoes the store sells, “so if we have lower footfall one day, we lose time from our shifts.”

This has a palpable effect on workers’ quality of life. Software firm Shiftboard’s own research shows that nearly half of U.S hourly workers would cut their own pay for more influence on their schedules.

Lack of labor organizing and regulations are obvious contributors to these problems. Although the retail workforce comprises less than 5 percent of U.S. workers unionized today, some progress has been made in the last decade. Oregon, Seattle, and New York City have passed and implemented scheduling predictability laws as recently as 2018. But one year after the laws took effect, The Shift Project, a Berkeley & UCSF research project surveying over 30,000 service workers, found mixed results in implementing and enforcing scheduling laws in Seattle.

Regulations also differ depending on the industry. In New York, a law passed in 2017 requires employers to give consenting fast food workers an extra $100 in wages for working back-to-back shifts with 11 hours or less in between, but this rule doesn’t extend to retail workers. If the law cannot protect vulnerable workers, then the responsibility falls onto employers and software companies—which, despite minor improvements in some places, many industries are still failing to do.

“Unless more companies using AI management technology also consider the dignity and needs of workers, working people—especially low-wage and non-union workers—will face greater risk of exploitation, economic instability, and other harms in the years to come,” Sanchez said. “At a bare minimum, [companies] should provide workers with the information and ability to contest and overturn determinations that technical systems are making about their livelihoods. Workers should have the right to collectively agree on workplace standards that are safe, fair, and predictable, including ones that address exploitative and invasive uses of AI in the workplace.”

Zoho, Deputy, and Schuh declined to comment for this story. Salesforce did not respond to requests for comment.



from VICE https://ift.tt/32x8Vz3
via cheap web hosting

No comments:

Post a Comment