Steer Your Recommendations

Interactive Recommendation Study

Estimated time: 10-15 minutes Minimum screen resolution: 1280x720 3 iterations per approach Questionnaire included Contact: vaclav.stibor@student.cuni.cz

What is this study about?

In this study, you will compare two feedback styles in a movie recommendation system.

  • Explicit feedback means you directly adjust steering sliders (feature controls) to express what you want more or less of.
  • Implicit feedback means you do not use sliders; the system learns from your interactions with shown movies.

Study parts

Preference elicitation

Choose a few movies you like so the system can estimate your starting taste profile.

Implicit feedback approach

Refine recommendations without sliders; the system learns from your interactions with shown movies.

Explicit feedback approach

Refine recommendations by directly adjusting steering sliders representing interpretable features.

Questionnaires

When an approach or the full study ends, you will continue to the relevant questionnaire automatically.

What will you do?

  1. Select movies you like so the system knows your baseline preferences.
  2. Try the implicit feedback approach over 3 rounds.
  3. Try the explicit feedback approach over 3 rounds using steering sliders.
  4. Fill out short questionnaires about your experience.

Good to know

  • There are no right or wrong answers (except for attention checks).
  • Your data is anonymous and used for research purposes only.

Informed consent

Please check the informed consent declaration and only continue with the study if you agree with it:

  • I familiarized myself with the aim and targets of the research project and I do not mind to contribute on it (e.g., the research topic does not go against my beliefs etc.). I agree that authors of the project may utilize my (anonymous) responses and my demographic data while presenting research outcomes - e.g. in scientific papers. I do not require any further monetary compensation for my participation in the research beyond what has been offered via the Prolific.com service.
  • We (study authors) will never publish data that could breach your identity (we do not have such data anyway:-). We also neither share your contact details to third parties nor disclose it publicly. In the prospective paper, we would like to publish an anonymized dataset of user responses. The dataset should contain following information:
    • randomly generated ID of the participant (i.e. UID = 468201)
    • demographic data of the participant (e.g., age group, education, movie consumption frequency)
    • participant's interactions (clicks, steering settings)
    • participant's responses to questionnaires
  • In the released dataset, there will be no mapping between the ID of the user and their true identity (we do not have such data anyway).
  • You can stop your participation at any time (just close the browser window:-).
  • You can always revert your consent to use your data - just write to us an email with your details (Prolific identifier, time when you started participation etc.).
  • Any disputes (hopefully, none should arise:-) will be governed by the law and jurisdiction of Czech Republic.
Confirmation required before you can continue