"
Figure 1. Global biometric technology market revenue in $ billions, 2018-2027
Table 1. A taxonomy for biometric AI systems in the workplace
Table 1. A taxonomy for biometric AI systems in the workplace
Figure 2. Use of biometric authentication in enterprises, 2019
Table 3. Biometric AI for recruitment
Table 4. Biometric AI for workplace monitoring
![Monitoring employees is not a new concept. Yet, in contrast to direct supervision by a physically present superior, the digitalisation of work and the internet of things (IoT) enables continuous and comprehensive tracking of all of workers’ activities (Edwards et al 2018). Interest in using technology to monitor and control what workers do is booming. The COVID-19 pandemic and the shift to remote work has exacerbated a trend already present before the crisis. In 2018, Gartner found that more than half of large corporations had adopted non-traditional monitoring techniques, up from 30 percent in 2015 (Kropp, 2019)12. During the pandemic, demand for biometric-monitoring AI soared, and one out of four companies introduced technologies to track their employees’ behaviour passively (Kropp, 2021; Mascellino, 2020). Workplace applications centre on tracking attendance, activity or performance. The most frequent technological methods of workplace surveillance tend to be monitoring of work emails, browser histories and files, CCTV and the recording and logging of phone calls (however, no granular data on use of monitoring technologies by EU companies is available). Monitoring via wearable devices is more common in workplaces that require a lot of physical activity, such as warehouses or construction sites. Many workplaces re-apply biometric security devices for the purpose of worker monitoring. For example, fingerprint-based attendance tracking systems are widely commercially available. Advocates of the technology claim that such systems make attendance tracking more efficient while preventing some workers from clocking-in for others, improving productivity for both management and workers. However, these systems were ruled illegal in Germany in 2020, barring exceptional circumstances (Burt, 2020). Because the systems collect highly personal data, they run afoul of European GDPR laws. When biometric data is combined with productivity-centred algorithms, the technology can be used to push efficiency and accuracy, potentially at the cost of surveillance and lower job quality (Gutelius and Theodore, 2019). Headlines about the deeply automated tracking processes in an Amazon warehouse offer an exemplary description of the risks of algorithmic monitoring and management. According to one report13, workers wear a type of tracker that monitors their location and movements as well as their work activity. Based on historic data, an algorithm establishes standardised productivity rates and benchmarks to be attained by each employee. The tracking device also measures time-off-task and sends automatic alerts to workers if the period between measured work activities becomes too long. Reportedly, the AI system included an automated termination process: it would autonomously fire workers when quality or productivity benchmarks weren’t maintained. Since thresholds were set to near-unattainable standards, workers were put under such significant time pressure that they would skip bathroom breaks in order to fulfil their artificially set benchmarks. Discouraging and timing toilet breaks represents a questionable control over basic human needs and also raises issues around equality, illustrated by a number of reported instances in Europe where female employees (not of Amazon) were asked to wear specific clothing to signal when they were menstruating to receive permission to use the restrooms more often14. In office settings, a similarly comprehensive picture is painted by AI-driven sociometric devices: small, wearable badges capable of tracking individual and collective behaviours at work based on audio, movement, proximity and location data. In combination with corporate metrics on output and performance, AI can link specific behaviours, such as talkativeness, or whether a worker dominates conversations, to productivity, identify (un-) productive processes and make suggestions to improve organisational efficiency (Eveleth, 2019; Ito-Masui et al 2021). Although linking a badge to the wearer’s identity requires consent according to the developers, critics argue that surveillance opportunities remain within reach, in particular in small or medium-sized entities (Moore, 2020). Affective computing can also play a role in monitoring work performance. A US start-up called Cogito developed an AI system for call centres which assesses the mood of customers during phone calls and cues agents to adapt their way of speaking accordingly. Using voice analysis and natural language processing, the technology detects over 200 indicators of emotional state of both the customer and the agent in real-time. When it identifies a certain emotional state in a customer – for example frustration – it alerts the agent to speak more slowly, or display more empathy. Importantly, the AI serves not only as a tool to improve customer satisfaction, but also to monitor workers, as supervisors have “the ability to proactively listen to live calls with no extra setup required [and] are automatically alerted to calls in which a customer is having a poor experience”15. Automated monitoring may ensure that well-performing workers are identified and rewarded in a more consistent and objective manner. However, this comes at a cost of constant surveillance. The psychosocial risks associated with constant algorithmic monitoring are real and must be taken into account (Nurski, 2021). 2.4 Safety and wellbeing](images/u165875-71.png?crc=4225413572)
Table 5: Biometric AI for health, safety and wellbeing
THE GLOBAL TRADE AND FINANCE PLATFORM