AI is transforming the workplace, including health and safety. Here’s what OSH professionals need to know.
For some of us, artificial intelligence (AI) is the stuff of dreams, making our lives easier. For others, it is a dystopian nightmare, where robots replace humans and take over the world. But whatever your perception of AI, for decades now, it has been a part of our everyday lives at home and in the workplace.
Put simply, AI is ‘a general-purpose technology capable of mimicking human intelligence processes’, says Iván Williams Jiménez, policy development manager at IOSH. ‘The increasing deployment of AI technologies can not only perform tasks previously undertaken by humans, but also observe information and undertake analysis in different ways.’
More businesses are introducing worker management systems based on AI to increase efficiency and productivity, and to identify and manage OSH risks.
Dr Laura Bradshaw, research programme lead in technology at IOSH, says there are many ways that AI can be employed to support safety and health in the workplace. It can be used for near-miss detection, ‘but needs to be used cautiously, as this could result in complacency around reporting’, says Laura.
She says AI can detect when people do not follow safety protocols, and offers cost-effective learning opportunities in safety and health. It can be used for speech recognition and for incident reporting. It can recognise still images, situations and objects. By using virtual reality, AI can be used for practice drills, which are normally expensive to run.
AI has enormous potential to manage OSH. ‘AI can easily be used to replace workers in hazardous environments – drones can be used to perform more efficient workplace inspections in complex environments,’ says Iván.
AI sensors could lead to tracking of all aspects of worker activity
Laura says AI can free up workers from mundane repetitive tasks. It is also ‘quick, cost-effective and efficient, and doesn’t tire – it has 24/7 applicability’, she says.
AI is good for planning and training, says David Sharp, founder and CEO of International Workplace – a digital learning provider specialising in health and safety training. ‘We use machine learning to deliver our course materials, and algorithms to recommend courses and track learner performance,’ he says.
Case Study: AI motorway schemes are placed on pause
New all-lane running smart motorway schemes use AI to regulate traffic, but their roll-out has been paused by the government until five years of safety data is available. The technology, which uses AI and machine learning algorithms to identify potential issues, provides road managers with the precise real-time location and trajectory of vehicles. The aim is to help predict and prevent collisions or issues. ‘This is a work-related issue as many of those on the road drive for a living, such as truck, bus and taxi drivers, but also trades, delivery or sales personnel, self-employed or any other role that includes driving for work,’ says Laura.
Case Study: AI air disasters
The Maneuvering Characteristics Augmentation System, an AI system designed to activate and assist the pilot in flight stabilisation, resulted in two Boeing 737 Max aeroplanes crashing. A Lion Air flight plunged into the sea off Jakarta and an Ethiopian Airlines flight crashed after takeoff from Addis Ababa within four months of each other in 2018-19. The deaths of 346 people led to the grounding of the Max fleet, and global scrutiny.
Source: Rushe, 2022
Predicting future outcomes
US-headquartered AI company SparkCognition provides a variety of industries, including manufacturing, oil and gas, and aerospace, with a wide array of AI technology to help predict future outcomes, optimise processes and protect against hazards.
‘We provide organisations with AI that analyses camera footage to prevent near misses in the warehouse, on the plant floor and in difficult-to-access areas,’ says Stephen Gold, chief marketing officer. ‘AI can detect the likelihood of an incident in “real time”, so that managers can be alerted to shut down equipment to prevent injury.’
Ireland-based software company Protex AI helps environmental health and safety teams ‘to use AI as a prevention tool to identify behaviours that lead to accidents, rather than waiting for them to occur’, says CEO Dan Hobbs.
The company’s privacy-preserving software plugs into existing CCTV infrastructure to use its computer vision technologies to capture unsafe events autonomously in settings such as warehouses, manufacturing facilities and ports. Working with the likes of leading UK retailer Marks & Spencer, Protex AI has been able to decrease incidents in the workplace by 80%, says Dan.
But AI also brings risks and challenges. ‘Potential misuse of AI-enabled workplace sensors could lead to tracking of all aspects of worker activity,’ says Iván.
He says algorithmic decision-making in people analytics and performance management does not normally involve human intervention and ethical consideration. ‘People working under algorithmic management could be exposed to heightened physical and psychosocial risks and stress.’
The roll-out of AI in the workplace can cause anxiety around job losses and role retention, says Laura. ‘Trust can be eroded with poor engagement. There is a need to communicate with workers early in the implementation process, and engagement is vital.’
AI can increase bias. ‘There is a danger that data could be used for discriminatory purposes. We need to remember that AI data is not neutral or objective – it has human bias encoded into it,’ says David.
The pitfalls of the unintended and unethical use of using AI systems in the workplace, and the safety problems they might cause staff, are increasingly well documented. For example, Amazon has significantly invested in the automation and robotisation of its warehouses. Despite this, work-related injuries have been reported to be 50% higher at robotic facilities than in its conventional warehouses in the US (Reveal Center for Investigative Reporting, 2020). ‘This raises questions on work intensity levels, the pace of work and productivity quotas pressure on warehouse workers,’ says Iván.
While much has been reported about AI’s failures – and successes – there is much that is unknown about this technology. As a recent NIOSH science blog notes ‘research gaps exist regarding the use and impact of AI on the workforce’ (Vietas, 2021). So what do we not understand about the use of AI on workplace safety?
How to ensure AI is used safely
- Effective Ensure AI is the right tool to address the problem/concern.
- Explainable The logic of AI and its decisions should be communicated to stakeholders in a concise and useful manner.
- Accountable Organisations and individuals should be accountable for the outcomes of the AI systems that they develop and implement.
- Secure AI systems should be safe from outside interference.
- Fair AI systems should be aware of and appropriately address potential discrimination and bias.
Bridging the knowledge gaps
AI can present a picture through data, but it doesn’t always show the entire picture of what is happening. ‘AI will be a tool that can be used to enhance processes, but there will always be a need for human integration and connection in these processes,’ says Laura.
A particular concern about AI is its increasing use in monitoring and surveillance through algorithmic tracking, says Laura. ‘In a positive light, it can be seen as monitoring worker wellbeing, such as ensuring adequate rest breaks for drivers.
‘However, it can also calculate how long workers spend in restrooms or the number of breaks taken within a shift, surveying movements via trackpad, fingerprints, webcam monitoring or through wearables.’
These growing cases of high levels of surveillance and tracking are having adverse effects on workers’ stress, anxiety and depression (Vou, 2021), says Laura.
‘Robust legislation and safeguarding are required around algorithmic management, to ensure workers’ safety and privacy and to protect them from adverse effects or psychosocial issues,’ she adds.
With the increasing use of AI for workplace decision-making and assisted work, ‘it is essential that more robust evidence focuses around the potential OSH benefits as well as risks’, says Iván.
‘It would be interesting to explore more available data and evidence drawing the links between the adoption of AI technologies to reduce work-related injuries, or to alleviate work-related physical demands, repetitive and stressful tasks which cause musculoskeletal disorders or mental ill health,’ he says.
There is also much that OSH professionals need to learn about the adoption and integration of AI technologies in the workplace. Iván suggests AI knowledge gaps are partly due to ‘a lack of transparency’ around how these technologies are adopted and applied to workplaces. ‘To that extent, corporations in this digital age need to keep pace with requirements for enhanced public disclosure, stronger due diligence, and policies and practices for the governance of AI.’
Iván says practitioners also need to be ‘better informed and catch up with technology debates and consequently scaling up the role of OSH in this field’.
‘This also applies to day-to-day policies and procedures. For example, fit-for-purpose risk management strategies with regards to the implementation of new technologies might become more relevant than ever,’ he says.
AI companies have a crucial role to play in educating people about how it works in practice, to allay fears and address concerns. ‘SparkCognition has a 50-acre research facility, which brings together the physical and digital world to help organisations really experience this technology, how it works and how it will optimise their businesses,’ says Stephen.
‘The onus is on AI companies to inform people about what it means for their job,’ says Dan. ‘At Protex AI, we educate and reassure teams that AI is there to help.’
Could we hold back AI while more research is conducted? Laura says: ‘This is challenging because the pace of progress is so rapid, and research is already playing catch up. Unfortunately, we don’t always take the time to press pause and collect more information around areas of concern.
‘Legislation is also constantly evolving as we further understand emerging technologies and how they are being implemented and adopted.
‘Laws can be made for one use of technology, and it is then adopted in a way entirely different from the original intended use. This “process creep” is a concern for lawmakers who are trying to catch up with the rapid speed of emerging technologies.’
Looking to the development of AI technologies in the future, IOSH will continue to advocate that they ‘incorporate a more human-centred and ethical focus, that prioritises occupational safety and health and process safety’, says Iván.
‘We believe that before any AI-enabled devices or systems are introduced into a workplace, a thorough and more proactive OSH review of their benefits and risks should be carried out.
‘OSH professionals, researchers, employers and workers must continue considering how AI-enabled applications in the workplace might impact the workforce and workplaces – positively and negatively.’
Whatever changes it brings to the workplace, ‘we need to remember that AI is as much about humans, and human behaviour, as it is about technology’, says David. ‘So whether you view it as brilliant or terrifying, AI is really what you make it. Don’t let it disempower you. Use it how you want to use it.’
- IOSH’s response to the European Commission white paper on AI: bit.ly/white-paper-artificial-intelligence
- IOSH’s self-driving vehicles: new safety ambition consultation: iosh.com/about-iosh/our-influence/consultations/self-driving-cars-new-safety-ambition
Department for Transport. (2022) Smart motorway rollout to be paused as government responds to Transport Committee report. (accessed 10 November 2022).
Evans W. (2020) How Amazon hid its safety crisis. Reveal. (accessed 10 November 2022).
Fatemi F. (2019) 3 ways artificial intelligence is transforming business operations. Forbes. (accessed 10 November 2022).
Grand View Research. (2022) Artificial intelligence market size, share & trends analysis report by solution, by technology (deep learning, machine learning, natural language processing, machine vision), by end use, by region, and segment forecasts, 2022 - 2030. (accessed 10 November 2022).
McKinsey. (2021) The state of AI in 2021. (accessed 10 November 2022).
PwC. (2017) Sizing the prize: what’s the real value of AI for your business and how can you capitalise?. (accessed 10 November 2022).
Rushe D, (2022) Boeing to pay $200m to settle charges it misled investors over 737 Max crashes. The Guardian. (accessed 10 November 2022).
Vietas J. (2021) The role of artificial intelligence in the future of work. Centers for Disease Control and Prevention. (accessed 10 November 2022).
Vou A. (2021) Surveillance before safeguards: algorithms will only deepen workplace distrust. Voxeurop. (accessed 10 November 2022).