Improving work posture on a university campus
Working at a desk on a laptop with poor posture can be very damaging to neck and back health. Research has shown that years of computer use and hours of computer use per week were significantly associated with neck and back pain. That being said, the majority of the population nowadays is required to spend extended periods of time sitting in front of a computer on a daily basis, and this is especially true of students at technical universities.
To prevent chronic neck and back pain and related health issues, it is essential to build good posture habits early on in life.
We followed a user-centered design process, as shown in the summary diagram below.
Although computer posture issues affect the large majority of the population, for this project, we decided to focus specifically on the needs of undergraduate students.
Our first step was to carry out observational research on the university campus. This helped us gain a rapid understanding of the current ways students work on campus, and how that affects their posture. In particular, it revealed that postures adopted were very varied, and that most were neither healthy nor sustainable work postures. It also highlighted that the furniture (desks and chairs) selection on campus was very diverse, and more often than not blatantly incompatible with good posture maintenance.
For a quantitative overview of students' work habits, understanding of and attitudes towards posture, and the impact of poor posture on their lives, we deployed a survey on social media (Georgia Tech Facebook groups and Georgia Tech subreddit), and received over 125 replies from current undergraduate students.
The survey confirmed that students spent extended periods of time in front of their computer everyday and identified schoolwork as the activity required students to be on their laptop for the longest periods. It also confirmed that students experience neck and back pain on a regular basis, with 40% of respondents reporting experiencing pain at least occasionally. Finally, responses highlighted a gap between what students think and what they do. Indeed, 58% of respondents reported understanding what good computer posture is, 88% believed posture to be important to preventing back and neck problems in the future, and yet 68% reported not maintaining good posture when working at a computer. This showed students are aware of the importance of maintaining good posture but are unable or uninterested to take steps towards improving it, despite experiencing pain in a large number of cases.
Based on the survey results, we developed a short semi-structured interview guide to talk to a dozen of students and gain a deeper, more qualitative understanding of the problem of posture maintenance for students. We analyzed the data produced through affinity mapping.
From the interviews, we discovered that students find it difficult to focus on both their work and maintaining good posture. They reported that the longer they had been working for, the harder it was to pay attention to posture. We also found that students put work productivity first and are unwilling to make adjustments that slow them down when it comes to posture. Finally, common methods for alleviating pain included walking around and stretching for the most part. Some users more in tune with their back due to existing injuries also brought massage objects to campus.
The main user research findings from all methods combined were summarized in two formats: personas and an infographic, as shown below.
Problem Space Research
With a clearer picture of our target users, we focused on furthering our understanding of the problem of posture maintenance. To do so, capitalizing on our user research, we created a storyboard and a task analysis diagram of how the task of maintaining posture while working was currently carried out.
We then looked to external sources of information by carrying out a literature review and a competitive analysis. This allowed us to identify weaknesses of existing products for posture maintenance. Mainly, we found that existing products tend to quickly become a source of frustration for users as they are too present and disturbing to the primary task of focused work. Product users also did not like having to wear a device at all time, based on our analysis of customer reviews.
For our project, this meant that our solution should not be intrusive to the user's work environment or negatively affect the ability to focus on the main task of working. Further, users would be unwilling to carry a device on or with them to campus.
Research Phase Summary
The main takeaways from our research were:
- Users understand what it means to have good posture but struggle to maintain one while working.
- Unless they have a pre-existing back condition, users are not interested in bringing their own posture maintenance helping devices to their workspace.
- Users typically wait until they are in pain to make adjustments to their posture. They are not willing to make major changes to the way they currently work to ensure better posture.
Brainstorming for a Solution
The first idea is never the best idea, so we made sure to maximize our chances by generating over 80 design ideas. To get there, we structured our brainstorming sessions into 5-minute design sprints. During each sprint, we individually came up with as many concepts as we could. At the end of each sprint, we went through the new ideas generated to discuss them. We repeated the sprints for as long as we could come up with new ideas.
After divergence came convergence, and it was time to comb through our gigantic list of ideas. To identify front-runners, we went back to our research findings and generated the following list of design requirements for a viable solution:
- Low desk footprint
- Minimal attention needed for feedback
- Showing clear status changes for rapid feedback interpretation
- Inviting interaction in an interesting way
- Standalone product from the user's perspective
With those criteria in mind, we discussed the different concepts and clustered them to make the narrowing process easier by highlighting major themes.
With a more organized landscape of ideas and after a lot of discussion, we voted on our preferred ideas to eventually end up with three ideas we wanted to take deeper:
For ideas 1 and 3, posture data would be obtained through a pressure-sensing chair and sent to the device for feedback display. For idea 2, posture data would be obtained through the laptop camera.
Low-Fidelity Prototypes & Expert Feedback
To decide which of the three ideas to pursue, we turned to other user experience researchers and designers for their input. To aid communication, we created brief storyboards and low-fidelity prototypes. Our audience asked us questions, gave us feedback, and voted on the ideas.
For the laptop LED augmentation, people expressed concerns about sensing done through the laptop camera. Some were worried about security issues, and others did not think camera sensing would be able to accurately sense posture as it would only be able to see the user torso. There were a few concerns raised about the prototypes giving feedback through light: it could be disturbing other users of the environment with strong lights, or lead to feedback desensitization with too ambient lights. Experts interpreted the meaning of intensity, color and blinking of lights very different, which raised ambiguity in terms of conveying the right meaning for correcting the posture. Considering this idea received overwhelmingly neutral votes, we decided not to pursue it any further.
People preferred the desk object to the LED stripes because to them, both were essentially doing the same thing, but the desk art display was a lot "cuter and cooler". The position of the character on the desk is a direct representation of user’s posture at any given point in time. This delivers clear meaning to the user and allows them to correct the posture.
The chair raised a few questions about whether it would be safe, physiologically. The concern was that the cushion support might lead to muscular atrophy, as muscles would not have to do much work anymore. Some suggested that instead of the cushions inflating and staying so until a new change was needed, the cushions could inflate to indicate what changes need to be made and immediately deflate. The user would then have to decide whether to act on the feedback or not, as done with the two other ideas. However, overall people loved the idea of airbag chair since it would reduce the interruption in their work caused by reminders for posture correction. The chair would automatically sense and correct their posture or remind them to correct their posture. There are no lights or external objects to be used or placed on the desk. This reduces the disturbance caused to other students/people working around the user.
To decide between the desk object and the chair, we went through careful considerations. Eventually, we felt the chair was less true to our target user group characteristics because of the distraction it creates. The chair was also not as scalable as the desk object, making it a less viable solution for campus on the long term.
We therefore decided the desk art figurine to be the most fitting solution.
From Idea to Implementation: A Bumpy Road
Coming out of the design phase, the main specifications for the desk object were:
- Posture is sensed through a pressure sensing chair seat
- The character tilts to the side or forward if the user tilts or slouches forward
- The longer the user tilts or slouches for, the more the object's posture decays
- The object is equipped with LEDs that indicate when to take breaks from work
Creating a fully articulated desk figurine would have been difficult as it would have required each body part to be animated separately with its own motor. For the purpose of prototyping, we thought a stiffer figurine would be enough. The object would move in a joystick-like manner, going left, right, and forward. Reducing the range and granularity of motion meant we had to reconsider the type of feedback we had in mind for the object.
Originally, the figure was supposed to tilt to a side to reflect the user tilting, and the longer the user tilted for, the more the posture of the object deteriorated, up to a point of non-return when the figurine would crumble entirely. However, we were now looking at an object that could only move as a block in a gradual manner and could in no way tilt. This meant we had lost the ability to signify that important state change when the user has been in a poor posture for too long. In context, users do not constantly look at the object, so it was important that they be able to understand the feedback at a glance.
Another issue was that of building an aesthetically-pleasing object and deciding what shape it should take. We were concerned that modeling it after a human was likely to convey the wrong idea about the actual mapping granularity and fidelity between the user's posture and the object's orientation.
After a quick brainstorming of new ways to model the object, we started considering two fundamentally different ideas: a robot character or a minimalist art piece. We recognized advantages and drawbacks in both and could not make a compelling argument for choosing one over the other. We therefore decided to carry out both prototypes so we could test the two designs and compare them to make an informed conclusion.
To bring our idea to life, we 3D printed a base and two alternatives — RobotMan & The Orb — for the actual object to be mounted on a pan-and-tilt servo motor block living inside the base. The motors and LEDs fitted in the devices are controlled by 2 Arduino boards, which makes for a lot of wires!
To test our product, we had to resort to wizard of oz-ing since the device was not actually connected to a pressure-sensing seat. Based on observations, the wizard moved the device to the appropriate orientation, which is anything but reliable but was sufficient for the purpose of testing mental models and overall usability.
We divided the evaluation phase into three parts. First, we carried out a heuristic evaluation using the following heuristics, derived from Nielsen & Molich, with a few extras we believed were relevant in the context:
- Ease of access
User attentiveness requirements
Clarity of feedback
Ability to inspire action
- Match between system and real world
Impact on environment
Visibility of system status
Recognition not recall
The second part of testing was a 'quick and dirty' usability testing, where the goal was to evaluate users' abilities to understand the function of the object and its feedback, with as little information provided as possible. We also used this step to choose a single design for the final phase of testing, based on participants' feedback.
The process of posture degradation is a slow and variable process, which put major constraints on time and the ability to gather data from many participants. To keep each test short here while still being able to cycle through the different stages of the device feedback, we opted for a study design where participants observed an actor sitting in different postures on a pressure-sensing chair with the device on the desk. This made the connection between posture and the device discoverable without being artificial or inconsistent.
By offering free cups of coffee in a high-traffic area on campus, we were able to test with 10 participants. We tested with the RobotMan for half of the participants, and with The Orb for the remaining half.
To begin with, participants were rather confused about the purpose of the product and its relation to the chair or the actor. However, after going through various state changes, participants in most cases uncovered the true function of the device.
Participants testing with RobotMan struggled a lot more than participants testing with The Orb. The humanoid aspect of the robot encouraged a very constrained mental model that was inaccurate and raised the expectation on the level of mapping between the object and the person sitting. Overall, participants preferred the design of The Orb and appreciated the additional LED feedback it provided. We therefore concluded The Orb was the winning design.
For the third and final part of testing, we wanted to test whether the device had an actual impact on posture and was effective at inspiring correcting action when necessary. To do so, we tested with 3 participants who worked for an hour with the device on their desk while we observed. Two of us operated the device, while the rest tracked a variety of metrics like facial expressions, number of posture changes, or number of glances. To be able to track glances and facial expressions without having to stare at participants and make them uncomfortable, one of us was on Google Hangout with participants. We were thus able to set up a realistic work environment by looking very involved in our observation tasks in a non-obvious manner and communicating as a research team online, thus not disturbing our participants.
The tests however turned out to be rather unfruitful because typically posture does not start deteriorating in the first hour of work. Only one participant reached the alert state, and the other two maintained very good posture, not making it past the first stage and never even glancing at the device.
From post-test debriefing with participants, we realized our tests suffered from the observer effect, whereby participants were not true to their posture habits for they knew we were observing them.
Summary of Usability Testing Results
- Participants were able to understand the function of the device within 5 minutes of observation
- Initial and alert states were easily understood but intermediate states less so
- Lights and colors were more helpful than the tilting motion in decoding the feedback
- The size of the device may be an issue in smaller desk environments
- If users disagree with the feedback provided, they want to be able to communicate that to the object
- Users have strong mental models about objects, which shape their entire experience with them
Based on the results of our usability testing, we decided to explore a different design for our desk object. The most important change was in shape and size. We found that the tilting motion did not provide much information to the users that they did not know already, and that the device was unnecessarily tall. We therefore decided to remove the tilt feedback and focus on colors and lights.
We also wanted to include the possibility for users to actively disengage with the device. In our previous design, the device turned on when the user sat down, and off when the user left the seat.
Finally, we thought more subtle posture feedback could be conveyed. For example, in the third phase of testing, we found out some users sometimes experienced pain from having maintained the same position for too long, regardless of whether it was good or bad posture. We thus added feedback for users to correct this behavior as well.
What I Learned
- It is important to take the time to carry out proper user and problem space research before discussing ideas for solutions. Like we say in French, we cannot "burn the steps".
- Identifying the task that a user is completing is a seemingly simple and obvious thing to do, but can be a deceivingly complex endeavor, especially when researching and designing for a secondary task.
- Hardware prototyping is messy and somewhat unpredictable, especially with Arduino!
- Designing usability studies should be done very carefully and pilot testing should never be skipped!