What are some safety issues that could arise when implementing AI solutions in the workplace?
Photo: Just_Super/gettyimages
Responding is Andrew Faulkner, communications and content manager, SafeStart, Belleville, Ontario.
Hardly a day goes by without a news outlet declaring that artificial intelligence is about to crash into the workplace like a meteorite. Most coverage of AI focuses on how it’ll transform desk jobs, but it’s not only white-collar work that’s affected. The advent of algorithm-powered chatbots presents several safety implications for workplaces in manufacturing, warehousing, construction and other industries. And while some see potential benefits related to data analysis and system-building, a number of serious safety risks lurk amid the promises of AI.
At first blush, jobs that make things, move things and fix things, such as electricians, maintenance workers and welders, to name only a few, seem insulated from AI. And although you won’t see an algorithm lugging a toolbox, its presence can still be felt on a worksite. Supervisors could use AI to generate a quick toolbox talk in a pinch, a safety manager might use AI to create slides for their next safety presentation and an EHS director may analyze data and change the company’s SMS based on insights from AI.
Many algorithms and models have a surprising rate of hallucinations as a result of how generative AI is designed, so even the most advanced AI models on the market are susceptible to fabrication and other errors. And in each of these examples, an error introduced by AI will trickle down to the worksite and potentially put a frontline employee in danger by providing incorrect information or omitting essential details. This makes it increasingly vital for workers to hone their own sense of safety awareness, as they’ll be working in the shadow of AI’s mistakes whether they know it or not. This also emphasizes the need for people to avoid using AI in a hurry, as they’ll be less likely to notice errors. After all, human factors such as rushing can be just as dangerous when using AI as they can be when doing any other work-related task.
AI isn’t always hidden behind the scenes – sometimes its dangers are in plain sight. Overconfidence has long been a notable human factor in the workplace, causing people to have a greater sense of their own abilities than they should. Overconfidence in AI functions in much the same way. When something is introduced as being the product of AI, it can have the sheen of infallibility – as the NIST Artificial Intelligence Risk Management Framework states: “Whether correct or not, AI systems are often perceived as being more objective than humans or as offering greater capabilities than general software.”
So when an AI-assisted product is rolled out in the workplace – whether that’s a safety analysis, work plan or even something as simple as a shift schedule – workers and frontline leaders alike can have an outsized confidence in it. And that can spell trouble if the overconfidence leads people to forgo their usual levels of safety diligence.
It’s ironic that the advent of AI has heightened the need for human intelligence, but that’s exactly the case with workplace safety. People’s personal sense of safety awareness, knowledge of EHS processes and ability to recognize human factors are all more important than ever. As the use of AI begins affecting workplaces, safety professionals can mitigate the additional risks posed by machine-learning software.
What this looks like will vary from company to company but, in most cases, it starts with reinforcing existing safety skills, training on human factors and educating workers that AI is like any other tool: It can be valuable in some situations, but dangerous if used incorrectly.
McCraren Compliance offers comprehensive safety training to help prevent accidents. Visit our class calendar to see how our training and consulting services can enhance your safety efforts.
Original article published by Safety+Health an NSC publication