The idea of learning from the user what their preferences are and feeding personalized content back to them has existed for a long time in the world of web. Why can’t we take the same approach and apply it to tangible physical products? Could machine learning and data analysis be used in the context of physical product design? These were the questions I wanted to ponder over and answer with my thesis project.
For demonstrating the idea, I wanted to choose an object that we physically interact with a lot during the day. Something tangible, something important, something that would have a lot of potential for personalized data collection. I chose the workstation chair for this reason. The fact that I was designing workstation seating products before coming to ITP gave me a head start as far as the research of chair design was concerned.
Some info on chair design: Every single dimension of the chair is defined to fit a dimensionally average person with reasonable comfort. Anthropometric data is used for this purpose to choose the average or edge case condition for a particular dimension. To elucidate, the width of the seat is chosen to fit a 90 percentile person but the depth might be chosen to a 50th percentile person. The contours would also be determined through similar means. This is the method to choose different dimensions so that most type of people (dimensionally) could sit with reasonable comfort.
My approach is to use these dimensions and contours as the starting point of the form and evolve them to fine tune the design based the learning from the usage of the chair.
To evaluate how the user uses the chair, I used pressure sensors at strategic points of the seat and back. I collected data for different users by making them sit and use the chair for a considerable amount of time. By collaborating with a data scientist and friend KR Aravind, we made sense of this data to learn usage patterns to put the data to define various states of usage.
The intent was to generate a form for the seat and back, that would distribute pressure in an equal manner so that the load is supported in a distributed fashion. This process would ensure the user being comfortable in the chair for longer ( Herman Miller research). For achieving this, the contours and the weave pattern of the back and seat were modified. We also were reading the data realtime to slowly modify the form over time if the usage pattern indicated a significant change. This is how the chair learns from its usage to modify itself for the better!
I firmly believe that when shape changing materials are commonplace, every physical object would learn and modify its form among other things to adapt to and serve the user better. My thesis project is a small (but hopefully significant step) in that direction!