NewsFlow

A research study on personalised news recommendations

Information Sheet

Jheronimus Academy of Data Science (JADS), a collaboration between Eindhoven University of Technology (TU/e) and Tilburg University, invites you to take part in this research project. This study is conducted as part of a Research-in-Action course at JADS and may inform a future MSc thesis project. The aim is to investigate how people interact with and experience personalised news recommendation systems. The study takes approximately 10 to 15 minutes. To participate you must be 18 years of age or older and comfortable reading news in English.

Your participation is completely voluntary and does not pose any physical, legal, or economic risks. You are not obliged to answer questions you are uncomfortable with and you can withdraw at any time before submitting your final consent decision without explaining why. Declining or withdrawing will not have any negative consequences for you.

Who are we?

Data controller	Eindhoven University of Technology (TU/e), De Groene Loper 3, 5612 AE Eindhoven
Researcher	Fehmican Aydemir — [email protected]
Supervisors	Dr. Martijn Willemsen (TU/e / JADS) — [email protected] Jasmin Kareem (JADS) — [email protected]
Privacy	[email protected]
Data Protection Officer	[email protected]

What will taking part involve?

Completing a short background questionnaire about you and your news reading habits
Selecting news topics of interest
Selecting articles to create an initial reading history
Browsing a personalised news feed
Interacting with available features depending on the version of the system you are assigned to
Completing a post-task questionnaire about your experience
Reading a full explanation of the study design before your data is submitted
Deciding whether you still consent to the use of your data after reading the explanation

What personal data do we collect?

Demographic data	Age, gender, education level — used to describe the study sample
News habits	Reading frequency and preferences — used as control variables
Behavioural logs	Article clicks and reading time during the session — used to test research hypotheses
Survey responses	Your ratings and opinions about the experience — primary outcome measures
Session identifier	A random code with no link to your identity — links behavioural data to survey responses

We do not collect your name, email address, or IP address. All data is pseudonymous from the moment of collection. During the study, data is stored on a GDPR-compliant EU server hosted by Hetzner Cloud in Germany. After data collection is complete, data will be transferred to TU/e OneDrive for secure institutional storage. Data is retained for 10 years after the research is complete, after which it will be deleted or fully anonymised.

Your rights

You have the right to request access to, rectification of, objection to, or erasure of your personal data. For questions or complaints contact the researcher at [email protected] or the TU/e Data Protection Officer at [email protected]. You may also file a complaint with the Dutch Data Protection Authority: Autoriteit Persoonsgegevens (www.autoriteitpersoonsgegevens.nl).

TU/e processes your personal data to conduct scientific research, which is the university's public task as stated in Article 1.3 of the Dutch Wet Hoger onderwijs en Wetenschappelijk onderzoek.

Consent

By selecting "I agree and begin", I confirm that:

I have read and understood the information sheet above and had the opportunity to ask questions.
I take part voluntarily.
I understand that I can stop at any time before submitting my final consent decision without explaining why and without any negative consequences.
I understand that my personal data will be collected and used as described above, in pseudonymous form.
I understand that I will receive a full explanation of the study design before my data is submitted and will be asked whether I still consent to the use of my data.
I am 18 years of age or older.
I am comfortable reading news in English.

A few quick questions

About you. Takes about 1 minute.

What do you like to read?

Select all topics that interest you. Pick at least 3.

World News Sports Finance TV & Film Health Travel Music Food & Drink Autos Lifestyle Weather Entertainment

Select at least 3 topics to continue

Quick warm-up

Which of these would you normally click on? Select any that interest you.

Please select at least 10 articles you would genuinely read. The more you select, the more personalised and accurate your recommendations will be. Aim for 15 or more if possible.

You're almost ready

Please read carefully before starting

01

You will see a personalised news feed built from your interests. Browse the articles as you would normally read news.

02

Click "Read article" on articles that interest you. Try to read at least 3-5 articles.

03

You can click the Why? button to see an explanation of why each article was recommended to you.

03

You can remove articles from your reading history using the button to update your recommendations in real time.

03

Please spend at least 2 minutes browsing before proceeding to the questionnaire.

Almost done

Please answer these questions about your experience

About your reading habits

How would you describe your news reading style (in general, not just during this study)?

I read quickly: headlines and abstracts only I read deeply: full articles on topics I care about I browse broadly: a bit of everything I mix depending on the topic or my mood

What type of news content do you prefer most?

Mostly breaking news and current events Mostly analysis, opinion and commentary Mostly features, investigations and longer reads No strong preference

How important is it that news recommendations match your interests closely?

Please indicate how much you agree with the statements below. There are no right or wrong answers. We are interested in your honest experience.

A few quick questions about your browsing session

During the browsing phase, I saw explanations for why articles were recommended to me.

Yes No Not sure

During the browsing phase, I removed one or more articles from my reading history.

Yes No

The initial topic selection affected what I saw in my feed.

According to the explanation, removing a history item would most likely:

A. Have no effect B. Change future recommendations related to that item C. Delete the recommended article D. Change the website layout

According to the system, removing an article from your reading history would most likely:

A. Delete the article permanently from the website B. Change the recommendations you see in your feed C. Send feedback to the news editors D. Have no effect on your recommendations

A few final questions

Have you ever actively changed what a platform recommends to you? For example: removing a video from your YouTube watch history, telling Netflix you are not interested in a show, disliking a Spotify song to stop seeing similar ones, or adjusting your preference settings on a news app.

Yes, I do this regularly Yes, I have done this a few times No, I have never done this I was not aware this was possible

I enjoy thinking in depth about complex problems.

I prefer simple explanations over detailed ones.

Open feedback

Is there anything you would like to tell us about your experience? (optional)

Before we finish

Please read the following carefully

What this study was really about

This study investigated how explanations and user control in news recommendation systems affect perceived transparency, perceived agency, trust, and satisfaction.

The study used a 2x2 between-subjects design crossing two features: counterfactual explanations (showing why articles were recommended) and reading history control (allowing you to remove articles from your history to change recommendations). Participants were randomly assigned to one of four conditions: no features (baseline), explanations only, control only, or both features combined. We could not tell you this beforehand because knowing the purpose might have influenced your responses.

Your participation helps us understand how to design more transparent and trustworthy AI-powered news systems. This research is conducted as part of a Research-in-Action course at JADS, a collaboration between Eindhoven University of Technology and Tilburg University, and may inform a future MSc thesis project.

Please note that the news articles you read were generated using an AI language model (Claude 3 Haiku by Anthropic) based on real article titles and abstracts from the publicly available Microsoft MIND dataset. The articles were not written by human journalists and do not represent real published news stories.

Retrospective consent

Now that you know the full purpose of this study, we ask for your consent to use your responses in our research. You are free to withdraw your data. This will have no negative consequences for you.

Thank you

Your responses have been recorded

Your completion code

Please save this code as proof of participation.

If you wish to withdraw your consent and have your data deleted at a later point, please send an email to [email protected] with your completion code as reference within 24 hours of participation. Your data will be deleted before data analysis begins.

Data deleted

Your responses have been removed

Your data has been deleted and will not be used in this research. Thank you for your time.