설계된 불안, AI 업무 자동화 [이서윤의 인공지능&인권지능]

Depression and anxiety, feelings we all experience at times, become much more severe when they persist for long periods. Although they are significantly less common in Western societies, the number of people in South Korea seeking mental health treatment due to depression and anxiety is steadily increasing each year. The reasons behind this trend are complex and varied, but one particular factor stands out: the erosion of autonomy in the workplace.

The ability for workers to have control over their own work and its impact is a fundamental element of ideal working conditions. The International Labour Organization (ILO) recently classified surveillance and the removal of autonomy as a new psychological and social risk. This "psychosocial risk" refers to any aspect of work design or management that increases stress associated with work.

A 2022 guideline from the World Health Organization (WHO) also highlighted the correlation between the degree of involvement in decision-making and mental health, with higher job control leading to reduced emotional exhaustion and burnout, and lower job control associated with increased mental issues and subsequent absenteeism.

While higher automation levels may seem to offer psychological comfort and stability, research shows the opposite to be true in most cases. When automation becomes too extreme outside of repetitive tasks, it leads to a loss of contextual awareness in daily work, and the inability to make judgments in the most complex and unexpected moments, thereby increasing mental workload and eroding situational awareness and control.

In such situations, automation may appear to alleviate workload, but it actually transforms into a different kind of burden. Interestingly, the study also found that handing over parts of decision-making to machines can actually decrease workload and increase situational awareness. This emphasizes the importance of the extent to which automation is implemented rather than the total amount of automation.

The key is to determine at which stage automation is introduced, as the stage of automation is more critical than the total volume.

Having substantial control over one's own work is crucial for individual well-being, going beyond the usual discussions around transparency and fairness in artificial intelligence. It's not just about having the final say, but also about knowing where and how much automation is appropriate. Implementing an automated system that allows humans to retain meaningful decision-making power requires sophisticated algorithm design.

Industries and sectors should engage in in-depth discussions to determine the appropriate level of automation for their specific processes.

The design of processes should involve labor participation and the inclusion of workers' opinions, as mandated by the Labor Participation Act. The question arises: is it fair to blame individuals for anxiety and depression when the system's design is at fault? As work design starts to be reshaped at speeds beyond human comprehension, creating space for human judgment in areas where it is still necessary becomes increasingly challenging. It's high time we had a more urgent discussion about this issue.