How AI-Driven Algorithms Improve an Individual’s Ergonomic Safety

With the use of AI-driven (Artificially Intelligent) algorithms, the pressure of personal worker safety is relieved from organizations and transferred to individuals. Workers are empowered by using personalised feedback and learning about their actions.

• By Toni-Louise Gianatti
• May 14, 2020

The physical demands of Material Manual Handling (MMH) workers are immense and are to be congratulated, handled with care, even. It is fair to say that employees are using their own invaluable asset (the body) to perform tasks that benefit organizations.

Musculoskeletal injuries at work cost the individual, the organization and society. According to the Bureau of Labour Statistics, in 2019, US companies lost more than $1 billion per week due to workplace injuries. Overexertion was the number one cause, relating to injuries from lifting, pushing, pulling, holding, carrying or throwing.¹

When it comes to workplace manual handling and training, one-size-fits-all doesn’t always match, and there are considerable evidence-based reviews supporting the idea that the effectiveness of classroom training is limited, and the principles are not applied in the working environment.² Traditional training also fails to address the compounding factors of lifting technique, posture, task repetition and intensity, which are often the cause of lower back pain onset and musculoskeletal disorders—a singular instance of poor manual handling.

Predictive analysis and AI are becoming the leading resource to help prevent injuries. Using big sets of personalized data to recognize and provide recommendations about how a worker is behaving can help to train the worker in a more personalized fashion.

What is Artificial Intelligence (AI)?

First, it is safe to say that AI is not just another technology trend. Gartner, a leading research and advisory company, is known for describing the process trends of emerging technologies. Gartner calls this the ‘hype cycle’ which is simply an explanation of the typical cycle of a new technology that comes onto the market and helps to distinguish hype from the real deal.³ The five stages look a little like this:

1. There is early proof of concept/product

2. The technologies first fulfil great expectations

3. The technologies begin to fail quickly

4. After time and research, the technologies rise again

5. Finally, the technologies meet initial expectations

Towards the end of 2018, all the technologies near the plateau stage (No. 5, meeting expectations) of the Gartner hype cycle had association with AI. This shows that AI is no longer a hype, especially also given that expenditure has increased 768 percent since 2016 and is set to reach $46 billion this year and $97.9 billion in 2023.⁷

There are a lot of different definitions of exactly what AI is, but at a simple level of interpretation describes AI as the collection and evaluation of extraordinarily large data sets (big data). It is also important to know that ‘machine learning’ algorithms are considered a subset of AI and can be defined as the ability for the machine to ‘learn’ from human behaviour and improve its analysis by using the algorithms. The algorithms work by taking large sets of data to recognize patterns and then training the machine to make recommendations. With continual use, the repetitions enhance modifications and the machine is then able to provide predictions about behaviour based on input it receives.

One of the earliest forms of this was developed by IBM, a piece of software that could play checkers autonomously. It gathered enormous amounts of data and then began to see all available options and was able to then make decisions based on what it had learned.

How Does this Help the Individual Manual Worker?

Anastasia Vasina is a Medical Doctor and Physiotherapist at Soter Analytics and is responsible for managing the research and development of technology for musculoskeletal (MSK) safety. She explains:

"Many health and safety organizations such as NIOSH have established rules and guidelines specifying what 'typical' humans can withstand with relation to frequency and duration of labour-intensive activities.⁴ These standards, however, face criticism as they rely on statistical averages and anthropometric data which does not withstand the immense variety of population observed in today’s society. Having the added technical tools that enable objective data rather than solely observational techniques, provides an increased ability to help workers to stay safe."

Personalized Experience

Taking the above into consideration, with machine learning algorithms, movement data collected from the worker using a device or sensor can personalize the safety of an individual and be used to calculate his or her ergonomic risk. This is game changing when it comes to musculoskeletal safety since no two bodies body are the same or have the same strength, fatigue levels, emotions, injuries or any of the factors that influence movement behaviour.

Personalizing the learning experience through AI-driven algorithms can make a significant impact on injury rates when all of these factors are taken into consideration.

Weight of Exertion

These algorithms have potential to not only evaluate exposure for individuals, but depending on the configuration, they can also recognize a person’s personal capabilities. For example, the weight of an object can be measured, and the algorithm notes how difficult it is for the worker to lift the object. With continuous collection of data by way of machine learning, AI algorithms can determine what would be the ‘norm’ of exposure for the individual user and if, on any given day, this is exceeded, one could see a prediction of fatigue.

Fatigue Management

When it comes to fatigue management, this is a superpower, especially if a user has an existing injury or is feeling unwell or has stress. The algorithms will recognize factors and give recommendations or warnings. Workers start to be conscious of when they are feeling fatigued, and a new pattern of behaviour through awareness is ignited. The more information the machine has about the user’s movements the more accurate it will be.

Job Profiling

Gathering data around different job roles is also beneficial. Each role has a contrasting risk, and with more data, it is possible to create job profiling and identify the risks within that particular task or job. This aids in understanding the exposure of each job and what should be considered “normal.”

Worker Acceptance

An area that has brought about some debate is how difficult it is to obtain worker buy-in due to privacy concerns around data collection. There is now evidence that majority of employees do not have aversions to data collection, so long as the use and who will see the data has been communicated in the right manner.

Applied Ergonomics Journal published a study regarding employee acceptance and noted that the way in which data was presented to the employees had a significant impact on their reception.⁵ Data collectors and managers should:

• Include employees in the decision making
• Provide enough evidence that data collection will yield desired results
• Communicate the exact data that will be collected, who will see it and if it will or can be aggregated and anonymized
• Communicate if the workers have control over their own data
• Foster a favourable safety environment and culture

Conclusion

If there is the ability for individuals to self-manage movement and safety by way of AI-driven algorithms providing objective data, then change can provide lasting outcomes. By transferring the responsibility to the worker and looking at the safety issues for one person and not a standard, organizations can benefit from fewer injuries, more accurate diagnosis of environmental issues and an overall increase in worker health. Using AI-driven algorithms based on data collected from workers is not a stand-alone solution but is a powerful tool to help within the safety ecosystem to greatly impact workers personal safety.

REFERENCES

¹ Bureau of Labor Statistics, U.S. Department of Labor. (2019). Workplace Safety Indices by industry: insights and methodology. Retrieved from https://business.libertymutualgroup.com/business-insurance/Documents/Services/DS200.pdf

² Gartner, Inc. (2020). Gartner Identifies Five Emerging Technology Trends That Will Blur the Lines Between Human and Machine. Retrieved from https://www.gartner.com/en/newsroom/press-releases/2018-08-20-gartner-identifies-five-emerging-technology-trends-that-will-blur-the-lines-between-human-and-machine

³ Bernard BP, editor. U.S. Department of Health and Human Services, Centers for Disease control and Prevention, National Institute of Occupational Safety and Health. Musculoskeletal disorders and workplace factors: a critical review of epidemiologic evidence for work-related musculoskeletal disorders of the neck, upper extremity, and lower back. July 1997. DHHS (NIOSH) Publication No. 97-141.

⁴ Jesse, V. Jacobs., Lawrence. J. Hettinger., Yueng-HsiangHuang., Susan Jeffries., Mary. F. Lesch., Lucinda. A. Simmons., Santosh, K, Verma., Joanna, L. Willetts. (2019). Employee Acceptance of Wearable Technology in the Workplace. Applied Ergonomics, 78, 148-156. https://doi.org/10.1016/j.apergo.2019.03.003

⁵ Nath, N., Chaspari, T., & Behzadan, A. (2018). Automated ergonomic risk monitoring using body-mounted sensors and machine learning. Advanced Engineering Informatics, 38, 514–526. https://doi.org/10.1016/j.aei.2018.08.020

⁶ Bini, S. (2018). Artificial Intelligence, Machine Learning, Deep Learning, and Cognitive Computing: What Do These Terms Mean and How Will They Impact Health Care? The Journal of Arthroplasty, 33(8), 2358–2361. https://doi.org/10.1016/j.arth.2018.02.067

⁷ IDC (2019) Worldwide Spending on Artificial Intelligence Systems Will Be Nearly $98 Billion in 2023, According to New IDC Spending Guide https://www.idc.com/getdoc.jsp?containerId=prUS45481219