13. Human-Centric Manufacturing, Ethics and Governance

Introduction and Context

Modern manufacturing is increasingly shaped by cyber-physical systems, collaborative robots, industrial artificial intelligence, connected equipment, digital work instructions, wearable devices and data-intensive decision systems. These technologies can improve productivity, quality, traceability and flexibility. They can also alter work substantially. A machine may reduce physical effort while increasing cognitive load; an algorithm may support a supervisor while making decisions less transparent; a connected tool may improve quality while creating new risks for privacy and cybersecurity.

Human-centric manufacturing begins with the recognition that manufacturing systems are social and technical systems. Equipment, software, processes and people jointly produce value. The objective is therefore not simply to automate as many tasks as possible, but to design work and governance so that technology strengthens human capability, safety, dignity, learning and responsible decision-making. This perspective is strategically important because productivity gains that damage health, trust or employability are unlikely to be sustainable.

For production managers, the challenge is to balance competing objectives: efficiency and autonomy, data-driven control and privacy, standardisation and worker discretion, automation and employment, innovation and compliance. Human-centric manufacturing provides principles and methods for making these choices deliberately rather than treating social and ethical consequences as afterthoughts.

Learning Objectives

After studying this chapter, the learner should be able to:

1. Explain the meaning and strategic significance of human-centric manufacturing.

2. Apply ergonomic, sociotechnical and human–machine collaboration principles to production design.

3. Analyse workforce augmentation, skills transformation and responsible automation.

4. Identify occupational, technological, ethical and cybersecurity risks in connected factories.

5. Design governance arrangements for data ownership, AI use, standards and regulatory compliance.

6. Evaluate manufacturing decisions through safety, fairness, accountability and sustainability criteria.

7. Recommend implementation practices for trustworthy and inclusive digital transformation.

Human-Centric Manufacturing as a Strategic Design Philosophy

Human-centric manufacturing places human well-being, agency and development alongside productivity, quality, flexibility and cost. It does not reject automation. Rather, it asks what should be automated, what should remain under human control, and how the combined system should be designed. A human-centric factory may contain extensive automation, provided that workers understand the system, retain meaningful authority, receive suitable training and are protected from unreasonable physical and psychological demands.

Three principles are particularly important. The first is human agency: people should be able to understand, question and influence decisions that affect their work. The second is human capability: technology should augment perception, reasoning, dexterity and learning rather than assume that workers are merely sources of labour to be substituted. The third is human dignity and well-being: safety, privacy, fairness and opportunities for development must be treated as performance conditions, not optional benefits.

This approach extends the traditional production objectives of quality, cost, delivery and flexibility. A contemporary performance framework also considers safety, inclusion, skill resilience, data responsibility, worker voice and long-term organisational capability. These dimensions are interdependent. A poorly designed interface can produce errors and rework; excessive surveillance can reduce trust and reporting; inadequate training can undermine both safety and equipment utilisation.

Important Concepts and Terminology

Human-centric manufacturing refers to the design and governance of manufacturing systems around human well-being, capability, agency and responsible technological use.

Human–machine collaboration is the coordinated performance of tasks by people and machines in which each contributes distinctive strengths.

Workforce augmentation uses technology to extend human physical, cognitive or sensory capabilities without assuming complete substitution.

Ergonomics is the scientific and professional discipline concerned with fitting work, tools, environments and organisations to human capabilities and limitations.

Cognitive ergonomics addresses mental processes such as attention, perception, memory, decision-making and workload.

Psychosocial risk is the possibility that work organisation, relationships, insecurity, excessive demands or lack of control will harm mental or physical health.

Responsible automation is automation designed and deployed with explicit consideration of safety, fairness, accountability, employment effects and human oversight.

Algorithmic management uses software, data and algorithms to allocate work, evaluate performance or support managerial decisions.

Data ownership concerns the rights and responsibilities associated with collecting, controlling, accessing, using and transferring data.

Data governance is the system of policies, roles, controls and practices that ensures data is accurate, secure, lawful and used appropriately.

AI explainability is the extent to which the reasoning or factors behind an AI-supported output can be understood by relevant users.

Human-in-the-loop means that a human participates directly in an automated decision or operation.

Human-on-the-loop means that a human supervises an automated system and can intervene when necessary.

Safety by design integrates hazard identification and risk reduction into the design of equipment, software and processes.

Cybersecurity protects digital systems, networks, data and operational technology from unauthorised access, disruption or manipulation.

Just transition refers to managing technological and organisational change so that affected workers receive fair opportunities, support and participation.

Ergonomics Beyond Physical Comfort

Classical ergonomics examines posture, force, repetition, reach, vibration, lighting, noise and workstation layout. These remain central in manufacturing. Poorly positioned tools, excessive lifting, repetitive motion and awkward inspection postures may cause musculoskeletal disorders, fatigue and absenteeism. Ergonomic design can include adjustable workstations, lift assists, gravity-feed racks, redesigned handles, appropriate seating, reduced reach distances and variation in tasks.

Modern factories also require cognitive and organisational ergonomics. Digital interfaces may present too many alarms, poorly prioritised messages or ambiguous instructions. A worker who must monitor several screens while responding to frequent alerts may experience mental overload even when physical effort is low. Conversely, excessive simplification can remove important context and encourage uncritical acceptance of automated recommendations.

An ergonomic assessment should therefore consider the complete work system: task demands, equipment, interface, environment, shift pattern, staffing, training, pace and recovery time. Useful methods include direct observation, worker interviews, task analysis, checklists, workload assessment, postural analysis, near-miss review and participatory design workshops. Measurements should not be interpreted in isolation. A low average cycle time, for example, may conceal high peak effort or unsafe recovery periods.

A strong practice is to involve operators during design rather than asking them to validate a finished solution. Operators possess practical knowledge of exceptions, maintenance difficulties, material variation and informal workarounds. Their participation improves both hazard identification and adoption. However, participation must be genuine: workers should be able to raise concerns without retaliation, and management should explain how their recommendations were considered.

Human–Machine Collaboration and Workforce Augmentation

Machines are generally effective at repeatability, high-speed calculation, heavy-force operations and continuous sensing. Humans contribute contextual understanding, ethical judgement, improvisation, communication, learning and the ability to handle unusual situations. Collaboration is valuable when tasks are decomposed so that these strengths complement one another.

Collaborative robots, or cobots, illustrate this principle. A cobot may present components, hold a part, perform repetitive fastening or support lifting while an operator performs inspection, adjustment or exception handling. Safe collaboration requires more than installing force-limiting equipment. Risk assessment must consider tooling, payload, speed, workspace, unexpected movement, programming changes, maintenance and foreseeable misuse. The operator must understand the robot’s states and know how to stop or isolate it.

Augmentation also includes vision systems that highlight defects, exoskeletons that reduce physical strain, augmented-reality instructions, speech interfaces and predictive maintenance tools. These applications can reduce error and fatigue, but they may create dependency, alert fatigue or new forms of exclusion. A wearable device that is uncomfortable, difficult to clean or unsuitable for different body sizes is not human-centric merely because it is technologically advanced.

The appropriate level of human involvement depends on risk and uncertainty. A fully automatic conveyor speed adjustment may require routine monitoring, whereas a system that rejects products with safety implications may require human verification and an auditable reason. The concepts of human-in-the-loop and human-on-the-loop should therefore be selected through risk analysis rather than applied as slogans.

Occupational Safety in the Connected Factory

Occupational safety includes traditional hazards and technology-related hazards. Traditional hazards include moving machinery, electricity, chemicals, heat, noise, confined spaces and vehicle movement. Connected production introduces additional concerns: autonomous equipment may behave unexpectedly after software updates; sensors may fail silently; remote access may alter machine parameters; and workers may be distracted by digital devices.

A practical safety system follows a hierarchy of controls. Elimination and substitution are preferable to relying on worker behaviour. Engineering controls include guarding, interlocks, safe speed limits, presence sensing, emergency stops and physical separation. Administrative controls include procedures, competence requirements, permits, maintenance schedules and incident investigation. Personal protective equipment remains important but should not compensate for poorly designed systems.

Safety management should incorporate both leading and lagging indicators. Injury rates and lost-time incidents are lagging indicators. Near misses, completion of corrective actions, safety-critical training, interlock tests, inspection quality and worker-reported hazards are leading indicators. Digital dashboards can assist, but they may distort behaviour if managers reward low reporting rather than genuine risk reduction. A fall in reported near misses may indicate improvement—or fear of reporting.

Management of change is essential when introducing automation or AI. A change review should ask: What new hazards are created? Which existing controls become ineffective? How will workers interact with the system during normal operation, faults, cleaning and maintenance? What happens if communication fails? Who is authorised to override the system? Safety validation must include realistic abnormal conditions, not only ideal demonstrations.

Skills Transformation and the Employment Question

Digital manufacturing changes the composition of skills rather than simply eliminating work. Operators may need competence in data interpretation, digital troubleshooting, robot interaction, cybersecurity awareness and process optimisation. Maintenance technicians increasingly combine mechanical, electrical, controls and software knowledge. Supervisors must interpret analytics while also managing trust, learning and cross-functional collaboration.

A useful skills strategy begins with a capability map. It identifies current skills, future task requirements, critical gaps and learning pathways. Training should combine formal instruction, simulation, supervised practice, peer learning and problem-based projects. Certification should demonstrate actual competence rather than attendance. Reskilling decisions should be based on realistic task analysis; not every employee must become a programmer, but every employee affected by automation needs adequate understanding of the new work system.

Responsible automation requires attention to distributional effects. Workers whose roles are reduced should be considered for redeployment, job redesign, phased transition or supported career movement. Consultation should occur before major decisions are irreversible. This is both an ethical obligation and a practical risk-control measure: workers often identify process consequences that technical teams overlook.

Managers should avoid measuring the success of transformation only by headcount reduction. Relevant measures include internal mobility, training completion linked to competence, inclusion across age and demographic groups, error rates during transition, employee voice, health outcomes and the quality of redesigned jobs. The goal is productive transformation with a capable and respected workforce.

Governance for Data, AI and Cybersecurity

Connected manufacturing generates production records, machine logs, quality images, maintenance histories, location information and sometimes biometric or behavioural data. Governance determines who may collect these data, for what purpose, for how long, and under what safeguards. Data should be classified according to sensitivity and operational importance. Access should follow least privilege, meaning that users receive only the access necessary for their roles.

Data ownership is often more complex than legal possession. A supplier may operate a machine platform, the manufacturer may generate process data, employees may be represented in workforce data, and customers may have rights in product information. Contracts should clarify access, permitted uses, retention, portability, confidentiality, intellectual property and responsibilities after termination. Organisations should not assume that the ability to collect data creates unlimited authority to use it.

AI governance requires a lifecycle approach. Before deployment, the organisation should define the purpose, intended users, affected groups, performance criteria and unacceptable uses. During development, data quality, representativeness, security and potential bias should be evaluated. During operation, outputs should be monitored, decisions logged, incidents investigated and models reviewed when process conditions change. A model that performs well in one product mix may become unreliable after a material or equipment change.

Explainability should be proportionate to impact. A recommendation concerning routine tool scheduling may require less explanation than an algorithm that influences disciplinary action, access to overtime or employment selection. In high-impact cases, affected persons should receive understandable reasons and a meaningful route for review. Accountability cannot be transferred to an algorithm or vendor. A named organisational owner must remain responsible for outcomes.

Operational technology cybersecurity requires defence in depth. Important practices include asset inventories, network segmentation, strong authentication, secure remote access, patch and vulnerability management, tested backups, incident response plans, supplier assessment and recovery exercises. Cybersecurity is also a human issue. Workers need training to recognise phishing, unsafe USB devices, suspicious remote requests and anomalous machine behaviour. Security controls that are impractical may encourage workarounds, so usability and consultation matter.

Ethics, Standards and Regulatory Compliance

Ethical governance concerns what an organisation ought to do, while compliance concerns what it must do under applicable law, regulation, contracts and standards. Compliance is a minimum condition, not a complete ethical programme. A practice may be legally permissible yet unfair, intrusive or inconsistent with organisational values.

A useful ethical review examines at least five questions. Is the purpose legitimate and necessary? Are affected people treated fairly? Is the system transparent enough for contestability? Are benefits and burdens distributed reasonably? Can harm be prevented, detected and remedied? These questions apply to worker monitoring, algorithmic scheduling, facial recognition, automated quality decisions and restructuring.

Relevant standards and regulations vary by jurisdiction and industry. Production managers should work with legal, safety, information-security and human-resource specialists to identify applicable requirements relating to machinery safety, occupational health, privacy, employment, data protection, product liability, environmental protection and critical infrastructure. International standards can provide structured guidance for quality, occupational safety, information security, AI risk management and human-centred design, but certification does not eliminate the need for judgement.

Governance should assign clear responsibilities. A cross-functional committee may oversee high-impact technology, while process owners remain responsible for day-to-day controls. An AI or automation register can record purpose, owner, data sources, risk classification, validation evidence, review date and incident history. Procurement processes should require suppliers to disclose system limitations, update practices, security responsibilities, audit rights and arrangements for data access.

A Framework for Responsible Automation Decisions

A production manager can evaluate a proposed technology through the following sequence:

1. Define the problem. State the operational need and the human outcome sought. Do not begin with a technology looking for a problem.

2. Map the work system. Document tasks, roles, dependencies, exceptions, hazards, data flows and affected stakeholders.

3. Compare alternatives. Consider elimination, process redesign, low-technology improvements, augmentation and full automation.

4. Assess risks and rights. Evaluate safety, privacy, cybersecurity, bias, employment effects, reliability and environmental consequences.

5. Design human control. Specify authority, override conditions, escalation routes, explanations and fallback procedures.

6. Validate with users. Test usability, accessibility, abnormal situations and representative operating conditions.

7. Deploy progressively. Use pilots, training, monitoring and clearly defined stop criteria.

8. Monitor and improve. Review performance, incidents, worker feedback, fairness and changing legal or operational conditions.

This framework prevents the common error of treating implementation as the end of the project. Responsible automation is a continuing governance process.

Illustrative Case: Introducing Collaborative Robots in an Assembly Plant

Background

A fictional medium-sized manufacturer of industrial pumps planned to introduce collaborative robots for repetitive screwdriving and component presentation. The expected benefits were reduced wrist strain, improved torque consistency and greater production flexibility.

Situation and Problem

The initial proposal focused on purchasing equipment and achieving a targeted cycle time. Operators were concerned that the cobot would accelerate work, reduce job security and make it difficult to stop the line during faults. Maintenance staff lacked experience with robot software, while the information-technology department had not assessed the proposed remote connection. Management also proposed using workstation data to rank individual operator performance.

Analysis

A multidisciplinary team mapped the process and found that the principal ergonomic risk was not only screwdriving repetition but also awkward replenishment of fasteners. A redesigned presentation rack would address part of the problem at low cost. The team also identified risks from unexpected robot movement, inadequate recovery after faults, unclear responsibility for parameter changes and possible misuse of performance data. Operators reported that quality exceptions required judgement that the proposed automatic sequence could not reliably handle.

Approach and Strategy

The plant adopted an augmentation strategy. The cobot would perform repetitive screwdriving under verified conditions, while the operator retained responsibility for exception handling and final visual confirmation. A physical and software risk assessment established speed limits, safeguarded zones, stop functions and maintenance isolation procedures. Remote access was disabled by default and allowed only through approved, time-limited, logged connections.

Operators and maintenance technicians participated in workstation trials. Training covered robot states, safe intervention, fault recovery, data practices and escalation. The plant decided not to use individual cycle-time data for disciplinary ranking. Instead, aggregated data would support process balancing and ergonomic review. A governance committee approved the pilot, reviewed incidents weekly and defined criteria for suspension if safety or quality deteriorated.

Outcome

The pilot demonstrated improved consistency and reduced repetitive effort, but it also revealed that the original digital instruction screen caused delays during product variants. The interface was redesigned, and variant selection was made more visible. Several workers moved into robot-support and process-improvement roles after completing structured training. The plant did not treat the pilot as proof that every station should be automated; it used the evidence to refine its decision rules.

Learning Points

The case illustrates that responsible automation begins with problem definition rather than equipment selection. Ergonomic improvement may require process redesign in addition to robotics. Human control must be specified operationally, not merely promised. Data collected for production improvement should not automatically be repurposed for employee surveillance. Finally, worker participation, cybersecurity and skills development are integral to implementation quality.

Challenges, Risks and Common Mistakes

Human-centric transformation can fail through technology determinism—the assumption that a technically possible solution is automatically desirable. Other common mistakes include measuring only short-term productivity, treating training as a one-time event, ignoring informal work, relying on personal protective equipment instead of engineering controls, and purchasing systems without clear data or cybersecurity responsibilities.

Organisations may also create a false choice between automation and no change. In practice, options include job redesign, assistive devices, better material presentation, decision support, flexible automation and incremental adoption. Another risk is consultation without influence, in which workers are asked for opinions after decisions have been made. This damages trust and reduces the quality of information received.

AI systems introduce risks of biased data, model drift, automation bias and false confidence. Human review is ineffective if reviewers lack time, authority or competence to challenge the system. Governance must therefore provide resources and incentives for meaningful oversight. Finally, global supply chains create accountability challenges when software, equipment and data services come from multiple vendors. Contracts, technical controls and audit rights must be designed before deployment.

Current Developments and Best Practices

Emerging manufacturing systems increasingly combine digital twins, edge analytics, wearable assistance, natural-language interfaces, computer vision and adaptive robotics. These developments may improve responsiveness, but they also increase interdependence among software, equipment and people. Generative AI may assist with maintenance instructions, troubleshooting and training; its outputs require verification because plausible language is not evidence of technical correctness.

Best practice is to establish a human-centric design charter, maintain an inventory of automated and AI-enabled systems, classify technology by risk, involve workers and safety specialists early, and integrate privacy and cybersecurity into design reviews. Organisations should use measurable acceptance criteria covering safety, usability, accessibility, reliability, learning and worker experience. Independent review is appropriate for high-impact systems. Periodic audits should examine not only whether controls exist but whether they work in actual operations.

Leadership behaviour is decisive. Managers who reward transparent reporting, learning and responsible intervention create safer systems than managers who demand uninterrupted output at any cost. Human-centric manufacturing is therefore not a single project or software feature; it is a continuing capability for governing technological change.

Chapter Summary

Human-centric manufacturing treats people, technology and organisation as an integrated production system. Its strategic purpose is to achieve productivity and flexibility while protecting safety, dignity, agency, privacy and long-term capability. Ergonomics must address physical, cognitive and organisational demands. Human–machine collaboration should allocate tasks according to the strengths and limitations of people and machines, with human control matched to risk.

Responsible automation includes workforce planning, reskilling, worker participation and fair transition. Connected factories require governance for data ownership, access, cybersecurity, AI explainability, accountability and lifecycle monitoring. Ethical practice goes beyond legal compliance by examining necessity, fairness, transparency, contestability and remedy. A disciplined process—problem definition, work-system mapping, alternative analysis, risk assessment, human-control design, participatory validation, staged deployment and continuing review—enables production managers to make responsible technology decisions.

Key Terms

Human-centric manufacturing: Manufacturing designed around human well-being, agency, capability and responsible technology use.

Ergonomics: The discipline of fitting work systems to human capabilities and limitations.

Cognitive ergonomics: The design of tasks and interfaces for human perception, attention, memory and decision-making.

Workforce augmentation: Technology that extends human physical, cognitive or sensory capability.

Responsible automation: Automation implemented with consideration of safety, fairness, accountability and employment effects.

Human-in-the-loop: A system in which a human participates directly in an automated decision or operation.

Human-on-the-loop: A system in which a human supervises automation and can intervene.

Psychosocial risk: Work-related conditions that may harm psychological or physical health.

Data governance: Policies, roles and controls governing data quality, access, security and use.

Data ownership: Rights and responsibilities concerning control and use of data.

Algorithmic management: Use of algorithms to allocate work, assess performance or support management.

AI explainability: The degree to which an AI output can be understood by relevant users.

Safety by design: Integrating hazard prevention into system design from the beginning.

Cybersecurity: Protection of digital and operational systems from unauthorised access or disruption.

Defence in depth: Use of multiple complementary security controls rather than one safeguard.

Management of change: Formal evaluation of risks and controls when processes or technology change.

Just transition: Fair support for people affected by technological or organisational change.

Automation register: A record of automated systems, purposes, owners, risks and review information.

Review Questions

Short Answer Questions

1. What is meant by human-centric manufacturing?

2. Distinguish between human-in-the-loop and human-on-the-loop systems.

3. Why must cognitive ergonomics be considered in a digital factory?

4. What is the purpose of an automation register?

5. State two leading indicators of occupational safety.

Long Answer Questions

1. Explain how ergonomics in modern manufacturing extends beyond physical workstation design.

2. Discuss the role of workforce augmentation and reskilling in responsible automation.

3. Analyse the major elements of data and AI governance in a connected factory.

4. Explain how cybersecurity and occupational safety are interconnected in cyber-physical production.

5. Describe and evaluate the framework for responsible automation decisions presented in this chapter.

Analytical / Application Questions

1. A factory proposes using algorithmic scores based on cycle time to allocate overtime and identify low performers. Analyse the ethical, data-governance and operational risks, and recommend safeguards.

2. A collaborative robot is being introduced into a high-mix assembly cell. Develop a human–machine task allocation and validation plan that addresses safety, training, exception handling and cybersecurity.

3. After installing digital work instructions, error rates fall but worker complaints about mental workload and surveillance increase. Diagnose the possible causes and propose a balanced evaluation and improvement programme.

22. Multiple Choice Questions with Answers

Question 1

Which decision best reflects a human-centric manufacturing philosophy?

A. Automate every task that can technically be automated

B. Select technology only on the basis of immediate labour-cost reduction

C. Design automation to improve performance while preserving appropriate human agency and well-being

D. Remove workers from all decisions involving production technology

Correct Answer: C. Design automation to improve performance while preserving appropriate human agency and well-being

Question 2

Why is cognitive ergonomics important in a connected factory?

A. It determines the tensile strength of machine components

B. It addresses workload, attention, perception and decision-making in digital work

C. It replaces the need for physical guarding

D. It focuses exclusively on employee compensation

Correct Answer: B. It addresses workload, attention, perception and decision-making in digital work

Question 3

In which situation is human-in-the-loop control most appropriate?

A. A system operates without any possibility of human intervention

B. A human directly reviews or participates in a high-impact automated decision

C. A machine is disconnected from all data networks

D. A supervisor receives a report only after production ends

Correct Answer: B. A human directly reviews or participates in a high-impact automated decision

Question 4

A manufacturer uses worker-location data collected for safety to rank employees for discipline without informing them. What is the principal governance concern?

A. Excessive machine speed

B. Purpose limitation and inappropriate secondary use of data

C. Insufficient product standardisation

D. Lack of ergonomic seating

Correct Answer: B. Purpose limitation and inappropriate secondary use of data

Question 5

Which is a leading safety indicator?

A. Number of lost-time injuries last year

B. Total workers’ compensation cost

C. Completion and quality of corrective-action reviews

D. Number of reported fatalities

Correct Answer: C. Completion and quality of corrective-action reviews

Question 6

What is the most appropriate first step when evaluating a proposed automation project?

A. Purchase the newest available equipment

B. Define the operational problem and the human outcome sought

C. Dismiss concerns about employment effects

D. Connect the equipment to the production network immediately

Correct Answer: B. Define the operational problem and the human outcome sought

Question 7

Which cybersecurity practice best illustrates defence in depth?

A. Using one password for all systems

B. Relying only on operator vigilance

C. Combining segmentation, authentication, backups, monitoring and response planning

D. Disabling all maintenance access permanently

Correct Answer: C. Combining segmentation, authentication, backups, monitoring and response planning

Question 8

Why should AI-supported manufacturing decisions be monitored after deployment?

A. Models can become unreliable when data, products or process conditions change

B. Monitoring guarantees that the model will never make an error

C. Monitoring eliminates the need for human accountability

D. AI systems do not require validation before deployment

Correct Answer: A. Models can become unreliable when data, products or process conditions change

Question 9

In the illustrative cobot case, why did the plant avoid using individual cycle-time data for disciplinary ranking?

A. The data could not be collected technically

B. The plant wanted to use data for process improvement without creating inappropriate surveillance

C. Cycle time is never relevant to manufacturing

D. The robot was not capable of recording production data

Correct Answer: B. The plant wanted to use data for process improvement without creating inappropriate surveillance

Question 10

Which statement best distinguishes compliance from ethical governance?

A. Compliance is optional, whereas ethics is legally mandatory

B. Ethical governance is limited to cybersecurity controls

C. Compliance establishes required obligations, while ethics also examines fairness, dignity and responsible consequences

D. There is no meaningful distinction between them

Correct Answer: C. Compliance establishes required obligations, while ethics also examines fairness, dignity and responsible consequences