Algorithms in the workplace

With algorithms playing an increasingly fundamental role in our lives, Lee Nair, Managing Associate and Jasmin Stevens, Trainee Secondee from Lewis Silkin LLP identify potential workplace issues around bias, data protection, trust and good work in the context of the rise of algorithmic management and highlight the suggested recommendations from recent studies by the Institute for the Future of Work and ACAS.

Algorithms play an increasingly fundamental role in our lives, and the workplace is no exception. Accelerated by the Covid-19 crisis, the adoption of data-driven algorithmic systems that control how, when and where we work has rapidly increased. While two studies, by the Institute for the Future of Work and ACAS, recognise the benefits algorithms and AI can offer in the context of the workplace, they also identify a number of unintended consequences. In this article we look at the sorts of workplace issues that arise and the suggested recommendations set out in the studies.

Algorithms, AI and machine learning

The term “algorithm” encompasses a variety of different types of systems and technologies with a range of capabilities. Some algorithms are relatively simple in terms of data collection, but other types of AI and machine learning enable systems to learn for themselves how to achieve a set goal, rather than merely following pre-programmed steps.

The rise of “algorithmic management”

“Algorithmic management” is a relatively recently coined phrase to describe the way in which AI can be used in the workplace. The ACAS study My boss the algorithm: an ethical look at algorithms in the workplace notes that the term “algorithmic management” reflects a diverse set of technological tools and techniques to remotely manage workforces, relying on data collection and surveillance of workers to enable fully-automated (without human assistance) or semi-automated (which informs and shapes human decision making) decision making. The extent to which decision making is automated or semi-automated varies by system and business approach to deployment.

The IFOW study The Amazonian Era – How algorithmic systems are eroding good work notes that the impacts of automation through algorithmic management are varied and identifies categories:

Bias and algorithmic management

Algorithms have the potential to reduce subjective and sub-conscious bias involved in decisions made by humans but there is a growing debate around whether algorithms increase or diminish bias and unlawful discrimination in employment decisions. Using algorithms (both “off the shelf” and bespoke programmes incorporating machine learning) in the employment context can pose significant risks for employers, and there have been some high-profile examples of the risks of getting it wrong. For example, a few years ago Amazon abandoned its AI-developed recruitment tool (that was four years in the making) as it reportedly favoured male candidates. The ICO has published guidance on AI and data protection setting out ways of mitigating risks of discrimination when using AI systems.

Data protection and algorithmic management

Algorithmic systems at work can be deployed to perform a number of traditional HR and managerial tasks, such as recruitment, task-allocation, performance management and monitoring. These systems can offer transformative benefits, providing valuable data driven insights to the workforce. However, handled the wrong way they can risk direct conflict with fundamental principles of data minimisation and transparency embedded in the General Data Protection Regulation (GDPR). Employers must also implement suitable safeguarding measures under the GDPR if relying on purely automated decision-making which affects employees. In order to comply with GDPR principles while taking full advantage of new technologies, employers need to consider each case on its facts and balance (i) the interests of the data subjects protected by the GDPR; and (ii) the employer’s interests in investing in new data driven, people analytics technologies within the workplace.

Job satisfaction, trust, and algorithmic management

Algorithms are particularly prominent in the gig economy, where individuals may receive the entirety of their instructions and feedback via a platform. This technology can promote flexibility to accept jobs which are convenient because, for example, they are close by and won’t take long to complete, and can be juggled with jobs on other platforms. This flexibility is a significant draw to the gig economy for many individuals.

The IFOW’s study recognises that as the practices and business models of the gig economy are being extended beyond gig work to many other sectors, resulting in a more widespread restructuring of workplace behaviours, relationships and jobs which could have negative consequences for the workforce.

AI systems can have a significant impact on human interaction and job satisfaction. Where employers adopt algorithmic systems to take on increasing aspects of a manager’s role or where work is redefined in narrow, measurable terms, this can increase work pressure, undermine the value of human skill and judgement and significantly impact on the interpersonal relationships between managers and their direct reports. Algorithms, particularly those measuring workforce productivity, also risk eroding employees’ trust if not used in an upfront, considered and proportionate manner.

Good Work and algorithmic management

The IFOW’s study highlights the potential impact of algorithmic management on the following headline areas of what, in aggregate, make up ‘Good Work (work that promotes dignity, autonomy, equality, has fair pay and conditions and where people are supported to develop their talents and have a sense of community):

Access, fair pay and conditions: where management platforms are used not just to manage performance, but also to determine employment terms (including tasks, shifts, pay and working time) by way of algorithmically predicted performance, it can diminish fair, open and consistent standards of work and damage morale.
Dignity, autonomy, and equality: a shift from managerial trust and transparent dialogue towards micro-management through intense monitoring and surveillance, by collecting data about every aspect of working life, can undermine the dignity, autonomy and trust of workers. It can create a culture of proof, as opposed to a culture of trust. In addition, workers may fear engaging in conversation with colleagues due to a lack of privacy, which dehumanises the job further.
Learning and development: algorithmic systems can be used to capture knowledge, in a bid to create a “GPS” type manual of work, a template for future digital instruction, so that anyone can do the job. This can save time, lead to faster decision making and increase productivity. However, channelling staff to just one way of doing things can also limit opportunities for innovation, human skill and judgement.

Recommendations - algorithms in the workplace

Both the IFOW and ACAS studies make clear that the adoption of algorithms, AI and machine learning in the workplace must be done with care. The studies make a number of recommendations which include:

Agreed standards on the ethical use of algorithms around bias, fairness, surveillance, and accuracy.
Using algorithms to advise and work alongside human line managers, but not to replace them - a human manager should always have final responsibility for any workplace decisions.
Line manager training on how to understand algorithms and how to handle an ever-increasing amount of data responsibly.
Greater transparency for employees (and prospective employees) about when algorithms are being used and how they can be challenged, particularly in recruitment, assignment of work and performance management.
A new Employment Bill with a dedicated schedule of ‘Day 1’ digital rights providing new protections for all staff, irrespective of employment status, including rights to security, knowledge, involvement, and human contact. The Bill could also offer a right to disconnect – an idea gathering momentum following Ireland’s introduction of a statutory Code of Practice and the UK Government’s calls to ban bosses from contacting workers out of hours.
A new Accountability for Algorithms Act in the public interest requiring early algorithmic impact assessment and adjustment when adverse impacts are identified. This assessment could extend to equality impacts and the physical, mental, and financial risks of labour intensification.
Investigation, research and guidance by the Health and Safety Executive on the health risks from the intensification of work under management by algorithmic systems.
New mandatory disclosure obligations requiring regular reporting on the fact, purpose and outcomes of algorithmic systems shaping access, terms, and quality of work.
Collective bargaining covering the use of algorithmic systems and new collective rights for involvement and review when algorithmic systems are introduced.
Employee contracts, collective agreements, technology agreements and employee privacy notices to include explicit commitments about the employer’s collection and use of employee data through algorithmic systems.
Company reporting on equality and diversity, such as around the gender pay gap, to include information on any use of relevant algorithms in recruitment or pay decisions and how they are programmed to minimise biases.

Conclusion - algorithms in the workplace

AI continues to permeate many aspects of our lives, transforming work and working lives across a myriad of different sectors, working models and workforces. Adopting data driven AI technologies has profound implications for the experience, value and role of work and calls for a responsible, transparent – and human – approach.