Från prototyp till användartest - UX-testning med syntetiska och mänskliga användare : En studie av UX-testmetoder med syntetiska och mänskliga användare i utvärderingen av webbprototyper

Sammanfattning

The use of artificial intelligence in user-centric design steadily grows as it becomes a more common tool. This study examines whether synthetic users based on human users can replace human users in user tests for UX testing of web prototypes. The goal is to evaluate the differences between human and synthetic users and the role synthetic users can have in UX Testing. It also assesses the reliability of synthetic users as testers, especially for micro-companies with limited resources. The study used an iterative "Design and Creation" strategy. First, a web prototype was created using the design program Figma. Two prototypes with different looks were created, which were later evaluated using an A/B Test survey. After analysing the result from the A/B Test a new interactive prototype was compiled. The prototype was then evaluated using a usability test. Both human and synthetic users participated in both the A/B Test and the usability test. The synthetic users were created using personality traits that were collected from the human users. The attributes created two sets of personas, four based on the survey and five based on the usability test. Each persona was tasked with acting as a synthetic user in the tool ChatGPT. The results showed that human users often gave more varied and emotionally based answers and interactions, while synthetic users followed a more methodical and consistent pattern. The synthetic users often identified structural, content- related problems, but missed the emotional and contextual aspects that the human users raised. The conclusion was that synthetic users could not replace human users in usability tests. However, they functioned as an effective complement in early design phases for quickly identifying shortcomings, especially in projects with limited access to test subjects. The study also showed a need for continued research on synthetic users' credibility, variation and representativity.