|
Thank you for participating in my user study on XRL-DINE, an approach for explaining actions made by a reinforcement learning agent in the context of a self-adaptive system. XRL-DINE uses a dashboard to visualize the actions and the respective reasons.
This user study is part of my master's thesis, supervised by apl. Prof. Andreas Metzger, and examined by Prof. Klaus Pohl and Jun.-Prof. Stefan Schneegass. The master's thesis is written at the chair for Software Systems Engineering, paluno institute, University of Duisburg-Essen.
In this user study, you will be asked to adopt the role of a person who monitors and seeks to understand the decisions made by a reinforcement learning agent (e.g., a reinforcement learning software engineer, or an IT administrator responsible for system maintenance & monitoring).
You will receive a scenario description and a short explanation of what you can do with the XRL-DINE dashboard. Both scenario and dashboard descriptions are available throughout the entire questionnaire.
Afterwards, you have to solve tasks using the XRL-DINE dashboard and the scenario description. The tasks are in the style of a question. The time to solve a task is measured. Whenever time is measured you will be informed before you face the task. After each task, time measuring stops and you are asked to self-assess your previously given solution.
This survey will take around 20min to complete. It is not possible to go back after answering a question. Please make sure to use a computer/laptop instead of a smartphone/tablet.
Your responses will be recorded anonymously.
Thank you for participating!
Jan Laufer
|
