PostAR
Building a user research strategy for an augmented reality application with artificial intelligence
role
UX Trainee
​
approach
Wissenschaftliche Recherche, Research, Prototyping, Testing, Evaluation
​
tools
pen & paper
As part of the trainee program "Measuring UX in AR with the Power of AI", an AR mobile app was developed with the dual goal of creating research strategies for AR applications and exploring research methods such as quantitative data collection using artificial intelligence.
Development of a research strategy
​
To evaluate how established research methods can be applied to AR applications, I conducted extensive academic research and gathered approaches and insights from existing AR projects, while also developing a method of my own. This resulted in a set of methods suitable for user analysis in the AR domain.
I subsequently combined quantitative and qualitative methods in order to compare and triangulate the findings. The following overview presents my selected methods, the underlying assumptions, the research question, and the sample.
Use Case definition & planing
​
For the practical application of a research strategy, I needed a concrete use case and an AR application. In the offices of ELCA Informatik AG, numerous Paleo posters from different years are displayed on the walls. This inspired the idea of developing a simple AR app that uses the camera view to display information about the respective Paleo events. For the development of the AR app, I worked with the support of a Unity developer.
Execution & evaluation​
​
I combined two quantitative methods with two qualitative methods in order to both measure the breadth of insights and compare the similarity of findings across methods.
Qualitative
Interviews and observations
Quantitative
The prototype was made available to all employees for testing over a period of two days. Afterward, participants were required to complete a survey.
Following the execution of both the quantitative and qualitative methods, I clustered the findings using an affinity diagram. Each method was assigned its own color, allowing me to identify the contribution of each method within the groupings. The chart below illustrates the weighting of the insights.
​​​
In support of the movie method, I created a heuristic checklist for AR applications to ensure that the key focus areas were taken into account.
Artificial intelligence taking over user research
How can these methods be taken over by AI techniques?
I explored this question as well and, together with a Python developer, was able to evaluate various technical feasibilities and ideas. The combination of eye tracking and click-rate data during usage already showed significant potential. However, it did not generate enough data to sufficiently train or feed an AI model.
In addition, I examined the broader application of artificial intelligence in user research, such as the automated creation of personas or user insights within specific industries. The analysis of data from reviews or social media posts shows particularly high potential in this area.
Learning
AR apps trigger fundamentally different user behaviors, which must be taken into account even more carefully when designing research methods. It is more difficult to set up test environments without having a fully developed product. Some prototyping tools, such as Torch, show promise; however, working in a 3D environment introduces entirely different challenges. Methods like Wizard-of-Oz also offer strong potential to simulate fictitious situations and ultimately test future AR products.
Building and developing AI systems requires large amounts of data and significant time. In addition, issues around data privacy and data quality remain critical considerations.