Which system is used. The assistant in this study is powered by the large
language model GPT-4o mini (gpt-4o-mini), provided through the OpenAI API.
The study application itself is hosted by Hetzner in Germany.
What is sent to OpenAI. To generate the assistant's replies, the messages you enter and the conversation needed for the current task are sent to OpenAI. Your participant ID, name, email address, or other direct identifiers are not sent to OpenAI unless you include them yourself in a chat message.
Where the information is processed. To generate the assistant's responses, your messages are transmitted to OpenAI and processed on OpenAI's systems. This processing may involve transfers to countries outside the European Economic Area, including the United States. OpenAI states that such transfers are protected using applicable safeguards, such as European Commission adequacy decisions or Standard Contractual Clauses.
Model training. OpenAI does not use inputs and outputs submitted through its API to train or improve its general models by default. The research team has not opted in to sharing the study's API inputs and outputs with OpenAI for model training or improvement.
Retention by OpenAI. OpenAI may retain the messages sent to the API, the generated replies, and associated safety-related metadata in abuse-monitoring logs for up to 30 days. OpenAI may retain this information for longer if required by law or where reasonably necessary to protect its services or third parties. According to OpenAI, the information is deleted from its systems after the applicable retention period.
What is recorded, and what could be inferred. This study records your chat messages, the assistant's replies, your decisions, and interaction logs including timestamps and how long you take to respond. All data are stored under a randomly assigned study ID only and are not linked to your name.
Deletion. The raw data will be deleted at 30.08.2027. Any e-mail address or student number collected for compensation is stored separately from the research data and deleted once compensation has been processed.
For tasks involving location-based recommendations, you do not need to specify your current location. If location access is requested, assume that the assistant would determine your location automatically. Please do not enter any personally identifying or sensitive information (for example full names, home addresses, specific locations, or health information) into the chat. If you prefer, you may use fictional or hypothetical examples instead of your own circumstances.
Questions? Contact M.Fedorenko@campus.lmu.de.
In this study you will complete a series of short tasks with an AI assistant. For each task you will see an instruction at the top of the screen. Write your own message to the assistant based on that instruction and your personal preferences.
When you are finished with a task, click Next task to continue.
There are – tasks in total. Please complete them in one sitting.
Important: Please do not enter any identifying or sensitive personal data (for example full names, addresses, specific locations, or health data) into the chat. Use general or hypothetical information instead. The tasks can be completed entirely with general or made-up preferences.
You have completed all tasks.
Your completion code: —
Please copy this code, switch back to the browser tab with the questionnaire, and enter it there to continue to the final questions.
Once you have entered the code in the questionnaire, you can close this tab. Please do not close the questionnaire tab before you have submitted your answers.