Full study information: TutorChat, motivation, and learning

Why are we doing this study?

Artificial intelligence is changing how people study and learn. It can be useful, but we still know little about when it genuinely helps students think, stay motivated, and understand what they study—and when it does not.

A team from the University of California, Berkeley, and the University of Toronto, working with Humanitas, wants to study how TutorChat is used and how its use relates to motivation, participation, reasoning, and learning. The study evaluates the tool and the experience of using it; it does not aim to judge or diagnose you or decide whether you are a “good” or “bad” student.

You are invited because you are aged 14–17 and belong to a participating class.

What will you do if you participate?

We will invite you to complete three questionnaires: one at the beginning, one in the middle, and one at the end of the school year. Allow about ten minutes for each one. Questions cover your experience of studying, motivation, learning strategies, responses to difficult tasks and tests, and use of artificial intelligence.

To complete a questionnaire, you will need to answer every question. You may save your progress and continue later. Taking part in the study remains voluntary, and you may stop participating without affecting your tutoring, grades, teacher support, or school benefits. Answers will not be used to grade or diagnose you, will not be shown to teachers, and will not change TutorChat’s replies.

During 2026–2027, you will use TutorChat according to your class plan. Classes may use it differently or begin at different times. That class arrangement does not depend on your research decision. If you decline, you retain the same classes, tasks, teacher support, and TutorChat access assigned to your class.

What information will be used?

If you participate, the study may use your answers to the three questionnaires; coded TutorChat activity and learning data such as sessions, activities, and progress; predefined AI-generated research scores derived from conversations; and all official subject grades from the 2026–2027 school year.

The research team will not receive your name, email address, school identifier, or the key that connects your research code to you. From conversations it will receive only approved scores—never messages, transcripts, excerpts, quotations, or summaries. It will not receive voice recordings, medical records, disciplinary information, special educational needs, free-text teacher comments, or other school data that has not been expressly authorized.

How will AI analyze conversations?

If you agree, the study may use an authorized AI system to process conversations and create predefined research measures. These could measure observable behaviors such as asking questions, explaining reasoning, revising an answer, or continuing to work through difficulty.

To produce scores, the system must process conversation text inside the European Economic Area through the providers described in Your data and AI. The research team then receives only a numerical or categorical result. AI can be wrong, so scores are used only for research analyses and group results. They are not shown to teachers or used for grades, promotion, discipline, admission, diagnosis, or another individual decision. Conversations are not used to train AI models.

If you decline, you will not receive research questionnaires, research scores will not be created from your conversations, and your data will not enter the study dataset. TutorChat will still process messages needed to provide the educational service as explained in Your data and AI.

Who may access the information, and where is it stored?

Humanitas controls educational use of TutorChat and keeps the ordinary educational access described in Your data and AI. Teachers may access conversations and progress to help students, but they do not receive questionnaire answers, individual participation decisions, or research scores.

Primary research data is coded and stored encrypted with Amazon Web Services in Ireland. The key that connects a code to a student remains separate, under Humanitas’s control, inside the European Economic Area. Coding reduces identification risk but does not immediately make data anonymous.

Only authorized researcher or master-administrator roles may download coded data. Each download requires a recorded purpose, excludes identities and conversations, and is audit logged. Copies may be stored only on approved, institution-managed university devices in Canada or the United States.

Authorized oversight bodies may also confidentially inspect records needed to verify that your rights are protected. Exceptional technical access must be authorized, limited, and logged and does not permit research use.

Are there risks or benefits?

We do not expect physical risks. A question may feel personal, uncomfortable, or tiring; you may save the questionnaire and stop, but you will need to answer every question before completing it. AI may create an incorrect score, but it will have no individual consequence. Even with codes, encryption, and restricted access, a small risk of unauthorized access or identification remains; no system can promise zero privacy risk. You may also feel that a teacher or the school expects participation. They do not, and your decision does not affect grades, classes, TutorChat access, or your relationship with the school.

You may receive no direct benefit. What we learn could help improve the design of useful, safe, student-respecting educational AI. The study has no dedicated funding and offers no payment or reward. The principal investigator developed TutorChat; this relationship is disclosed so you can take it into account.

Participating, declining, and withdrawing

Participation is completely voluntary. You may decline, stop a questionnaire without completing it, or withdraw at any time without giving a reason. Your decision does not affect TutorChat, assignments, grades, teacher help, or another educational service.

To withdraw, email ceiroasolans@berkeley.edu. Questionnaires, reminders, and future collection then stop, and no new conversation research scores are created. TutorChat deletes saved or completed questionnaire answers and derived survey scores from active systems under the retention rules.

Ordinary educational conversations are not automatically deleted when you leave the study; they follow Your data and AI. Information already made irreversibly anonymous or included in a group statistical result cannot be separated later. A minimal withdrawal record may remain to prove that your choice was honored, without answers or conversations.

If permission from a parent or legal representative is also required, that permission does not replace your decision: you will not participate if you say no. If the study materially changes, we will explain the change and ask you to decide again.

How long will the information be kept?

Ordinary educational data follows Your data and AI. Research copies are kept for one year after the study ends. Recovery copies follow their technical deletion schedule. At the end of that period, identifiable or coded data is deleted or made irreversibly anonymous. Statistical results that can no longer identify anyone may be retained.

How will results be shared?

Results may appear in dissertations, academic articles, reports, or presentations as group findings. No student will be identified, and conversations, quotations, transcripts, or datasets that could identify you will not be published.

This document does not authorize individual data for other studies, model training, advertising, sale of information, or commercial purposes. Any new use requires new legal and ethical review and, where applicable, new authorization.

Legal basis and rights

Educational use of TutorChat has its own legal basis and does not depend on research participation. The proposed research basis is your affirmative consent under Article 6(1)(a) GDPR. Special-category data will not be deliberately collected without the relevant review and authorization.

You may request access to research data, correction, deletion or restriction, withdraw consent, and exercise other applicable rights. Verifying your identity may be necessary to locate your code without exposing information to unauthorized people. You may also complain to the Spanish Data Protection Agency.

Who can you contact?

Study questions or withdrawal:
Conrado Eiroa-Solans, Doctoral Student
ceiroasolans@berkeley.edu

Privacy and data protection at Humanitas:
Your teacher may pass on a request, or you may contact the Data Protection Officer at dpd@humanitastrescantos.com.

Data-protection complaints:
Spanish Data Protection Agency

Your decision

Reading this page or continuing to use TutorChat does not mean that you agree to participate. A decision is recorded only when you check the consent statement and choose “Yes, I want to participate” on the previous screen. “No, thank you; continue without participating” remains available without checking anything.

If you agree, TutorChat records your decision, checkbox state, date, class, a unique idempotency key, and the versions and digital hashes of the legal information, study information, and instrument. This lets the study prove and manage your consent. You may keep a copy of this page and change your mind later.