Warakorn - Fotolia

Can work allocation algorithms play fair?

Allocating work by algorithm might have advantages for workers as well as employers. Can it be done fairly and with respect for “human capital”?

Over the course of six months in 2016, writer James Bloodworth worked for four companies: Amazon, in one of its warehouses; a home-care provider; insurer Admiral’s call centre; and Uber.

Drivers for taxi operator Uber are self-employed contractors with no employment rights, but even so, Bloodworth thought working this way might be an improvement on his previous experiences, giving him control over when he drove and leaving time to write a book.

Control is not what he got. Uber drivers are offered work through an app which does not tell them anything about the next job, either its duration or destination. A driver accepting too low a proportion of jobs will effectively be fired, while refusing two or three in a row will result in them being locked out of the app for a period of time.

A dip in user ratings can also lead to a driver being barred, but higher fares are available during unsociable hours – when customers are more likely to leave poor ratings. Bloodworth also experienced Uber’s attempts to nudge him into continuing to drive, even if he had been on the road for hours, by telling him he had nearly earned a certain amount.

“All in all, it was a peculiar sort of freedom,” he sums up in Hired: six months undercover in low-wage Britain, the resulting book.

Allocating work by algorithm is not an inherently bad idea, according to James Farrar, chair of United Private Hire Drivers, a branch of the Independent Workers Union of Great Britain, who has co-led legal action against Uber for drivers’ rights.

Many drivers working for conventional minicab companies pay a fee for the privilege (Uber takes a percentage, typically 20-25%), only to see controllers giving the best jobs to their friends, ordering them to collect their takeaway food and even demanding bribes. “People didn’t just walk away from those operators, they ran to Uber,” Farrar says.

But the company’s algorithms create their own problems. Uber tells prospective drivers that “there’s no office and no boss”, adding that “with Uber, you’re in charge”. Farrar says this is not reflected in the way the company’s algorithms allocate jobs and influence driver behaviour.

“These companies thrive on their ambiguity in nudging and exercising control while denying they have that control,” he says. “They are trying to dance on this fine line between triggering workers’ rights claims and controlling their businesses.”

There are significant financial benefits for “gig economy” companies in keeping workers as self-employed, and Farrar thinks this is part of the reason companies rely on algorithms rather than explicit instructions. “It’s management control nonetheless, but it’ll be brought behind the curtain,” he says.

These companies thrive on their ambiguity in nudging and exercising control while denying they have that control
James Farrar, United Private Hire Drivers

It is difficult to know exactly how Uber and many other gig economy companies use their algorithms to manage workers, as their workings are not usually disclosed.

Farrar says that the 2011 Autoclenz case, where the UK supreme court decided that supposedly self-employed workers for a car-cleaning provider were entitled to the minimum wage, has given companies incentives for secrecy, as the judges used Autoclenz’s actions rather than its stated policy in deciding the case.

He and his legal team have previously used Uber’s own data as evidence, but the company now releases much less, given this could be used against it. However, Farrar says Uber gathers data about its drivers, including braking and acceleration rates and how fast they accept or decline jobs.

Empowered and flexible planning for workers

Some drivers attempt to fight back, according to Amany Elbanna, associate professor of information systems at Royal Holloway, University of London.

Through her research with Uber drivers, she has found that some use software on their phones to say they have completed a job before they arrive or to accept incoming jobs automatically. Others stay just outside “virtual taxi rank” zones at airports, where drivers are placed in a queue before being allocated jobs, aiming to get work faster.

Even so, Uber retains most of the power. Elbanna says the company could give drivers more control of how they work by providing the approximate start and end points of the next job offered, which would allow a driver who wants to stop working shortly to refuse a long trip. Instead it punishes those who reject too much work.

“This is a critical question, whether it is using data for surveillance and punishment or for empowering and to encourage people,” says Elbanna.

She adds that a data watchdog could also help empower drivers, such as by investigating complaints over measurement, with some drivers claiming differences between distances recorded on vehicle milometers and those used by Uber to calculate fares. She also thinks there should be guidelines on how work performance data is used, to stop managers using data to carry out surveillance of individuals.

Elbanna also believes algorithms can improve conditions for workers, such as by ensuring they take breaks and helping them to plan their schedules. She previously worked for a university where one colleague liked to walk his dog every day at midday, while another wanted to be home by 5pm to make dinner for her children. The university worked this out manually, but software could allow workers to enter their availability and then work out schedules from that.

“Work allocation systems provide an excellent opportunity to integrate ideas we have always wanted,” says Elbanna.

Danish technology company Planday goes some way towards providing such flexibility for shift-based workers. Through an app, it allows them to say when they are and are not available and to swap or give up shifts to other workers.

For the latter, the system checks a receiving worker is happy to accept a transferred shift and lets employers set rules, to ensure that the right skills are available and even to prevent named individuals working together. It also provides a communications channel for workers who are outside an organisation’s email system, a way to manage holidays and information on earnings.

Chief executive Christian Brøndum says that at present, many employers focus only on cutting costs, resulting in unhappy workers. “There is going to be a power-shift in the market between businesses and workers,” he says. In the future, the company plans to incorporate other sources of data such as weather forecasts, so employers can change staffing requirements based on these.

Avondale Care Scotland, which runs four 24-hour nursing care homes in Falkirk and Fife, has to ensure a legally set mix of staff at each site at all times. Director Adrian Hendry says that Planday has made it easier to produce compliant rotas, which it now does six to eight weeks in advance rather than as little as a fortnight previously, saving around £500 a month in staff time.

Hendry adds that the company’s 450 staff love the system, with earlier information on rotas giving them more time to plan and the ability to exchange shifts as long as mandatory staffing ratios are still met.

“It gives them much more autonomy and power to speak to a colleague,” he says, as they do not have to go through an administrator during office hours. “It’s a quicker process for them, a more efficient process for us and a safer process, as we’re guaranteed the skill mix will be correct.”

Labour as a service

UK-based Broadstone provides organisations with a platform to find workers, initially in privacy security, but with plans to expand to other industries. Chief executive Tom Pickersgill describes this as “labour as a service”, akin to what employment agencies offer, but at a lower price and with greater flexibility. It currently has around 10,000 people seeking work, and hopes to increase this to more than 40,000 during 2019.

Employers list shifts and workers apply for them, rather than being offered jobs automatically. An employer can make its own choices on who to take on, but Broadstone offers an optimised list. While it doesn’t publish the algorithm used for the latter, it is based on whether workers have “badges” which can be won by undertaking three shifts without cancelling or maintaining a certain star rating, for example.

“If you can get those badges, you should be in a good position,” says Pickersgill. Those who qualify for all five badges receive a minimum hourly rate of £14, compared with £9 for those starting with the platform.

But most gig economy companies, including Uber, run their own systems. “One of the problems with these platforms is they are not peer-to-peer,” says Pete Robertson, associate professor at the school of applied sciences at Edinburgh Napier University. “A lot of it is to do with the power balance between employer and employee.”

A more equitable model would be a marketplace used by multiple employers with channels for workers to exchange information freely, he adds. But if companies can get workers onto their own platform, they are unlikely to give up the power this offers.

A few such companies are starting to improve what they offer workers, although these have not yet involved algorithms. Following an employment tribunal victory by union GMB, which found self-employed workers for UK courier Hermes were entitled to holiday pay and the minimum wage, the company has agreed a collective bargaining agreement which includes holiday pay and a minimum per-hour rate.

Hilfr, a Danish company that helps people find home cleaners, has similarly decided on a collective agreement with trade union 3F, which covers holidays, sick pay and pension contributions for those who complete 100 hours of work through the platform. However, it does not choose who goes where.

“We let customers decide for themselves,” says Dennis True, Hilfr’s chief executive, through searchable listings. He says that the company did not consciously decided against using a job-allocation algorithm, but adds that Hilfr’s “system works”.

What does a good machine-based management system look like? It looks like a good manager that instructed the machine
James Farrar, United Private Hire Drivers

Software can offer casual and shift workers more control while cutting administration for employers, which may also find their workforce is more contented and more willing to stick around. But it can also push people into working in ways that are not in their interests.

Publishing the algorithms or at least the measures gig economy companies draw on to allocate work would be a significant step. But while United Private Hire Drivers’ Farrar sees algorithmic transparency as critical, he believes that the most important issue is that companies accept responsibility for the actions of their algorithms.

If managers feel there is a psychological distance between them and the software, Farrar thinks it increases the risk of them abusing workers as they can claim that they are merely curating a community rather than running a business.

“An algorithm is just a management instruction,” Farrar says. “What does a good machine-based management system look like? It looks like a good manager that instructed the machine.”

Read more about worker management software

This was last published in February 2019

Read more on HR software

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.

-ADS BY GOOGLE

SearchCIO

SearchSecurity

SearchNetworking

SearchDataCenter

SearchDataManagement

Close