This site is a work in progress and has not been widely shared. Content may contain errors. Feedback is welcome.
This site is undergoing review. Some annotations were human-generated, some AI-generated — all are being verified.
Back to papers

An Occupational Safety and Health Perspective on Human in Control and AI

Niehaus, Hartwig, Rosen, Wischniewski

2022Frontiers in Artificial Intelligence30 citations
Observational labor marketInterdisciplinaryTheoretical model
Automation / RobotsAI (General)Decision-makingRoutine task changeHealthcare
Abstract

The continuous and rapid development of AI-based systems comes along with an increase in automation of tasks and, therewith, a qualitative shift in opportunities and challenges for occupational safety and health. A fundamental aspect of humane working conditions is the ability to exert influence over different aspects of one's own work. Consequently, stakeholders contribute to the prospect of maintaining the workers' autonomy albeit increasing automation and summarize this aspiration with the human in control principle. Job control has been part of multiple theories and models within the field of occupational psychology. However, most of the models do not include specific technical considerations nor focus on task but rather on job level. That is, they are possibly not able to fully explain specific changes regarding the digitalization of tasks. According to the results of a large-scale study on German workers (DiWaBe), this seems to be the case to some extend: the influence of varying degrees of automation, moderated by perceived autonomy, on workers' wellbeing was not consistent. However, automation is a double-edged sword: on a high level, it can be reversely related to the workers' job control while highly autonomous and reliable systems can also create opportunities for more flexible, impactful and diverse working tasks. Consequently, automation can foster and decrease the factor of job control. Models about the optimal level of automation aim to give guidelines on how the former can be achieved. The results of the DiWaBe study indicate that automation in occupational practice does not always happen in line with these models. Instead, a substantial part of automation happens at the decision-making level, while executive actions remain with the human. From an occupational safety and health perspective, it is therefore crucial to closely monitor and anticipate the implementation of AI in working systems. Constellations where employees are too controlled by technology and are left with a high degree of demands and very limited resources should be avoided. Instead, it would be favorable to use AI as an assistance tool for the employees, helping them to gather and process information and assisting them in decision-making.

Summary

Niehaus et al. use linear regression analysis of a large-scale German survey (DiWaBe, n>8,000) alongside review of occupational psychology models to examine how automation technologies giving instructions to workers affect job control, work intensity, and mental health indicators

Main Finding

Higher levels of instructions from automation technologies (both ICT and machines) are associated with significantly lower job control and higher task repetition; instructions by machines additionally predict higher physical exhaustion and information overload, while instructions by ICT predict higher physical stress but not mental health outcomes

Primary Datasets

DiWaBe (Digitalization and Change in Employment) Survey 2019

Secondary Datasets

None

Key Methods
Linear regression analysis of cross-sectional survey data examining associations between technology-provided instructions (ICT and machines) and working conditions (job control, work intensity, burnout indicators)
Sample Period
2019
Geographic Coverage
Germany
Sample Size
8,000+ employees from approximately 2,000 German companies; regression analyses use 2,341 to 5,592 observations depending on technology type
Level of Analysis
Individual
Occupation Classification
None
Industry Classification
None
Notes
Frontiers in Artificial Intelligence, vol. 5 [Claude classification]: This is primarily a theoretical/review paper that synthesizes occupational psychology models (Job-Demand-Control, Job Characteristics Model, Vitamin Model) with automation theory, combined with empirical analysis from the DiWaBe survey. The paper examines how automation affects worker autonomy and the 'human in control' principle. The survey assesses technology-provided instructions rather than AI exposure per se, but is framed in context of AI-based systems. The theoretical models reviewed include formal frameworks (e.g., Parasuraman's levels of automation, Kaber & Endsley's ten-level model). [Claude classification]: This is primarily a theoretical/review paper that synthesizes occupational psychology models (Job-Demand-Control, Job Characteristics Model, Vitamin Model) with automation theory, combined with empirical analysis from the DiWaBe survey. The paper examines how automation affects worker autonomy and the 'human in control' principle. The survey assesses technology-provided instructions rather than AI exposure per se, but is framed in context of AI-based systems. The theoretical models reviewed include formal frameworks (e.g., Parasuraman's levels of automation, Kaber & Endsley's ten-level model). [Claude classification]: This is primarily a theoretical/review paper that synthesizes occupational psychology models (Job-Demand-Control, Job Characteristics Model, Vitamin Model) with automation theory, combined with empirical analysis from the DiWaBe survey. The paper examines how automation affects worker autonomy and the 'human in control' principle. The survey assesses technology-provided instructions rather than AI exposure per se, but is framed in context of AI-based systems. The theoretical models reviewed include formal frameworks (e.g., Parasuraman's levels of automation, Kaber & Endsley's ten-level model). [Claude classification]: This is primarily a theoretical/review paper that synthesizes occupational psychology models (Job-Demand-Control, Job Characteristics Model, Vitamin Model) with automation theory, combined with empirical analysis from the DiWaBe survey. The paper examines how automation affects worker autonomy and the 'human in control' principle. The survey assesses technology-provided instructions rather than AI exposure per se, but is framed in context of AI-based systems. The theoretical models reviewed include formal frameworks (e.g., Parasuraman's levels of automation, Kaber & Endsley's ten-level model). [Claude classification]: This is primarily a theoretical/review paper that synthesizes occupational psychology models (Job-Demand-Control, Job Characteristics Model, Vitamin Model) with automation theory, combined with empirical analysis from the DiWaBe survey. The paper examines how automation affects worker autonomy and the 'human in control' principle. The survey assesses technology-provided instructions rather than AI exposure per se, but is framed in context of AI-based systems. The theoretical models reviewed include formal frameworks (e.g., Parasuraman's levels of automation, Kaber & Endsley's ten-level model). [Claude classification]: This is primarily a theoretical/review paper that synthesizes occupational psychology models (Job-Demand-Control, Job Characteristics Model, Vitamin Model) with automation theory, combined with empirical analysis from the DiWaBe survey. The paper examines how automation affects worker autonomy and the 'human in control' principle. The survey assesses technology-provided instructions rather than AI exposure per se, but is framed in context of AI-based systems. The theoretical models reviewed include formal frameworks (e.g., Parasuraman's levels of automation, Kaber & Endsley's ten-level model). [Claude classification]: This is primarily a theoretical/review paper that synthesizes occupational psychology models (Job-Demand-Control, Job Characteristics Model, Vitamin Model) with automation theory, combined with empirical analysis from the DiWaBe survey. The paper examines how automation affects worker autonomy and the 'human in control' principle. The survey assesses technology-provided instructions rather than AI exposure per se, but is framed in context of AI-based systems. The theoretical models reviewed include formal frameworks (e.g., Parasuraman's levels of automation, Kaber & Endsley's ten-level model). [Claude classification]: This is primarily a theoretical/review paper that synthesizes occupational psychology models (Job-Demand-Control, Job Characteristics Model, Vitamin Model) with automation theory, combined with empirical analysis from the DiWaBe survey. The paper examines how automation affects worker autonomy and the 'human in control' principle. The survey assesses technology-provided instructions rather than AI exposure per se, but is framed in context of AI-based systems. The theoretical models reviewed include formal frameworks (e.g., Parasuraman's levels of automation, Kaber & Endsley's ten-level model). [Claude classification]: This is primarily a theoretical/review paper that synthesizes occupational psychology models (Job-Demand-Control, Job Characteristics Model, Vitamin Model) with automation theory, combined with empirical analysis from the DiWaBe survey. The paper examines how automation affects worker autonomy and the 'human in control' principle. The survey assesses technology-provided instructions rather than AI exposure per se, but is framed in context of AI-based systems. The theoretical models reviewed include formal frameworks (e.g., Parasuraman's levels of automation, Kaber & Endsley's ten-level model). [Claude classification]: This is primarily a theoretical/review paper that synthesizes occupational psychology models (Job-Demand-Control, Job Characteristics Model, Vitamin Model) with automation theory, combined with empirical analysis from the DiWaBe survey. The paper examines how automation affects worker autonomy and the 'human in control' principle. The survey assesses technology-provided instructions rather than AI exposure per se, but is framed in context of AI-based systems. The theoretical models reviewed include formal frameworks (e.g., Parasuraman's levels of automation, Kaber & Endsley's ten-level model). [Claude classification]: This is primarily a theoretical/review paper that synthesizes occupational psychology models (Job-Demand-Control, Job Characteristics Model, Vitamin Model) with automation theory, combined with empirical analysis from the DiWaBe survey. The paper examines how automation affects worker autonomy and the 'human in control' principle. The survey assesses technology-provided instructions rather than AI exposure per se, but is framed in context of AI-based systems. The theoretical models reviewed include formal frameworks (e.g., Parasuraman's levels of automation, Kaber & Endsley's ten-level model).