top of page

Advanced Tool Job Assessment

The Advanced Tool Job Assessment uses computer vision and machine learning to make assessing physical jobs for ergonomic risk faster, more accurate, and more precise. The user takes a video of the operator performing a job, uploads it to the system, selects privacy options, enters force data, and then recieves a detailed overview of the different areas of risk in the job.

Background

The Humantech Whole-Body Assessment is a measurement of ergonomic risk in a manual job that's broken down into various body regions and requires a significant amount of training and knowledge of ergonomics to be performed. Users go through a three-day workshop to learn to identify the different postures, forces, duration, and frequency for ergonomic risk, as well as how to apply the results of the assessment and identify direct causes of risk and improvements to mitigate that risk.

The Advanced Tool was developed by Kinetica Labs and removes the possibility of human error from job assessments. It was the first step in moving our assessments and improvement processes from a human-led effort that can be inconsistent and requires a lot of training to a machine learning process that will be consistent and accurate no matter who is performing the assessment.

Research

We've gone through multiple rounds of improvements to the Advanced Tool workflow. A major barrier to entry for many clients were privacy concerns - some operators very reasonably didn't want their face being recorded at work, and some clients had concerns about proprietary materials being recorded and uploaded into our system. So we worked with Kinetica to develop face blocking and background blurring. Initial research showed that a common issue was that the forces entry workflow was even more complex than it was in the Whole-Body Assessment given that the Advanced Tool has the ability to be much more granular than the WBA, so we updated the forces workflow to be much more intelligent and move the burden of understanding how object weight translates into ergonomic forces from the user to the system. 

Video Privacy

Our primary concern when developing the workflow for adding privacy to videos was that if privacy options were selected for a video, it would never show on the front-end or be stored in an unblocked state. I worked with the team, our consultants, and users to develop a flow that would give users an understanding of what their video would look like when processed without exposing any private information.

Forces Input

One of the bigger changes we made to the workflow was how forces are entered. The way that Humantech had handled forces previously was to present a list of all possible forces that could be entered into the job assessment and have the user select the ranges that their measured forces fell between from a dropdown list. This was the way that the algorithm handled forces, but not the way that users generally conceived of forces, since they would measure a particular force and have to enter that force in the notes section to reference while filling out the job assessment. This also involved some ergonomic knowledge on the user's part: for example if a force was applied to both sides of the body the user had to divide it and apply it twice, and users had to remember that a lifting force was applied to the hands, elbows, shoulders, back, and legs. Users often had to refer to notes to complete this process and it simply wasn't intuitive.

In the forces update, I moved the cognitive burden from the user to the system by allowing the user to simply enter the force, which side it was on, and the direction that the force was applied in. Then the algorithm would make all necessary adjustments and apply the force to all affected body regions. We also allowed the user to enter estimated forces in cases where they didn't have a force gauge on site by asking the operator to rate their exertion on a 0-11 scale and normalized that to the force bands in the algorithm.

Development and feedback

The product development phase went very smoothly, we met our target and only had to make minor changes from the designs. Customer feedback was very positive in my user research sessions as well as feedback that we received through customer support managers. Working with our principal solution strategist team ensured that the ergonomics logic made sense at every step of the process, and consistent user testing ensured that our customers would be delighted by this change and, most importantly, that it would allow them to do their jobs easier, resulting in safer workplaces.

© 2024 Owen Robison

bottom of page