Privacy Policy

Last updated: 28 August 2026

This policy covers dbcourse.app, the learning platform for COMS W4111 Introduction to Databases at Columbia University. The platform is operated by the course instructor, Oktie Hassanzadeh, for the teaching of that course. It is not a commercial service, it carries no advertising, and it sells nothing.

Questions about anything below, and any request about your own data: privacy@dbcourse.app. For anything else — the platform, the course, a broken page — write to admin@dbcourse.app.

What we collect

From your Google account, when you sign in. Your email address, your name, and the URL of your Google profile picture. That is the entire set — signing in requests only the openid, email, and profile scopes, which are Google's non-sensitive scopes. The platform cannot read your Gmail, your Drive, your calendar, or your contacts, and it never asks for permission to.

For a columbia.edu address we also store your UNI, which is simply the part of the address before the @. It is stored so that a registrar roster can be matched to a sign-in.

From the course roster. If you are enrolled, the instructor uploads the registrar's class list, which carries your name, email, UNI, and section. This may create your record before you have ever signed in — that is deliberate, so the gradebook shows the whole class from week one.

From your coursework, as you do it. This is the bulk of what is stored, and it is stored because it is the course:

• Which units you have opened and completed, and when.
• Every answer you submit, kept as the raw text you wrote — the SQL query, the relational-algebra expression, the essay — together with the score, the machine's diagnosis of it, and the random seed that determines which variant of a question you were asked. Answers are re-graded from the stored text on the server, which is why the raw text is kept rather than just a number.
• Quiz and exam sittings: which questions you drew, when you started, when you submitted, and any extended-time accommodation applied to you.
• Scores, XP, streaks, badges, skill mastery, and any grade override or comment a grader recorded.
• Issue reports you file on a question, and the variant you were looking at when you filed one.

During in-app timed assessments only, the platform counts two things: how many times the browser window lost focus, and how many times text was pasted into an answer. These are counts, not content — nothing about what was pasted or where you went is recorded. They are shown to course staff as signals to look into, never as an accusation and never as an automatic penalty, and they are not shown to you during a sealed assessment. Nothing else about your device, browsing, or location is tracked.

What we do not collect. No date of birth, no government or student ID number, no address, no phone number, no payment details. No analytics service, no advertising network, no tracking pixels, no third-party cookies, and no fingerprinting. The only cookie the platform sets is the one that keeps you signed in.

How your data is used

To run the course, and for nothing else: to show you your own progress, to grade your work and let you appeal a grade, to let course staff grade and moderate, to compute the class-wide statistics that tell the instructor which topic to spend more lecture time on, and to produce the final grades that go to Columbia.

Class-wide statistics are not about you. The analytics page reports per-question difficulty and per-skill distributions across the class. Your user identifier reaches that calculation only so that a question can be correlated against your own overall performance; nothing on the page names or identifies a student.

Course staff can see your work. Instructors and TAs can see your submissions, scores, and issue reports — that is what grading is. Other students cannot see anything of yours.

Automated and AI-assisted grading

Most questions are graded by running your answer, not by a language model: your SQL is executed against test databases and compared with a reference query, your functional dependencies are checked by an algorithm, and so on. No third party is involved in any of that.

For a small number of written-answer questions, the instructor may enable an AI-assisted grading suggestion, which sends your answer to Anthropic's Claude API. When that happens:

• Your identity is not sent. The request contains the question, its rubric, and your answer text. Your name, your email, and your user id are structurally incapable of reaching it — the function that builds the request is not given them.
• A model never sets your grade. The output is a suggestion in a staff grading queue; a human records the score.
• It is off by default, both course-wide and per question, and it is enabled only for a rubric the instructor has already hand-graded enough to trust.

If you would rather your written answers were never sent to a model, write to privacy@dbcourse.app and it will be turned off for your submissions.

Who else your data reaches

• Google — identity only, and the flow runs the other way: Google tells us who you are when you choose to sign in. We send Google nothing about your coursework. Your use of Google sign-in is also governed by Google's own privacy policy.
• Vercel — hosts the application (US East, Ohio).
• Neon — hosts the database (AWS us-east-2, Ohio). Encrypted in transit and at rest.
• Anthropic — only for AI-assisted grading, only where enabled, and only as described above.
• Columbia University — final grades, which go to CourseWorks as the institutional record.

That is the complete list. Your data is not sold, rented, or shared with anyone else, and it is not used to train anyone's model.

Where your data is held, and for how long

All of it is stored in the United States.

Coursework and grade records are kept for as long as they are needed for academic records, grade appeals, and course administration, in line with Columbia's record retention requirements. We do not set a fixed deletion date, because a grade can be questioned long after a term ends and an answer depends on the work still being there.

You can ask for yours to be deleted at any time, and we will do it unless we are required to keep it — see Your choices, below.

Withdrawing from the course deactivates your enrollment rather than deleting your work, so that a grade dispute can still be answered.

Your choices

You can ask the instructor to show you everything held about you, correct anything wrong, or delete your account and its data. Ask at privacy@dbcourse.app.

One honest limit: while you are enrolled, and while a grade of yours could still be questioned, coursework and grade records are part of your academic record and cannot be deleted on request. Everything else can go at any time, and you can stop using the platform whenever you like. If you ask and we have to say no, we will tell you which records and why rather than simply declining.

If you are a student at a US institution, your education records are also protected by FERPA, and Columbia's own student-privacy policies apply to them alongside this one.

Changes

If this policy changes materially, the date at the top changes and enrolled students are told in class or by email. The current version is always the one at https://dbcourse.app/privacy.