Why you are reading this
We are running two research studies using three sets of records that their publishers make freely available to anyone: the edit history of Wikipedia, the Lichess open games database, and an anonymised dataset of student activity published by The Open University. If you have edited Wikipedia, played rated games on Lichess, or studied one of seven Open University modules in 2013 or 2014, records of what you did may be included.
We did not collect this information from you, and we have no way of contacting everyone it covers. The law allows a research project in that position to publish a notice instead of writing to each person. This page is that notice.
Who we are
Monomindai Ltd, a company registered in England and Wales, company number 16207542. We are the controller for this research. Registered with the Information Commissioner's Office, registration number ZC254306.
What the studies are for
The first study tests whether well-known patterns in how people make decisions can be measured in a consistent way from public records of what people actually did, in settings that have nothing to do with each other. Two examples: what editors do after someone undoes their edit, and whether chess players resign or play on when they are losing.
The second study asks a plainer question. When something a person did is reversed, rejected or does not succeed — an edit undone, a losing position in a game, an assessment not passed — what do people observably do next, and which of those responses appear in more than one setting? It counts actions that can be seen in the records themselves. It reads no text, it tests no theory, and it produces no score, label or judgement about any person.
The second study has two stages. The first, completed in October 2026, looked for responses that recur. The second, Act Confirmation, tests one of them — whether people are more likely to stop taking part after something they did was reversed than after something that was not — on Wikipedia and Lichess records set aside for this purpose and not looked at before, under rules fixed and recorded before any record is read. It compares each person only with themselves and with people of similar experience, and it reports whatever it finds, including nothing. The Open University records are not used in this stage.
In both studies, and both stages, we are interested in patterns across large groups. We are not interested in any individual, and nothing in either study is about you as a person.
What we use
From Wikipedia, published by the Wikimedia Foundation: records of edits made on English Wikipedia between July 2024 and July 2025, and the edit history of Simple English Wikipedia. We use when an edit was made, whether and how quickly it was undone and by whom, the size of the change, the edit tags Wikipedia applies, and account information published with the edit: the account's numeric ID, how many edits it had made and when it was created or first edited, and whether it was flagged as a bot or blocked or was a temporary account.
We do not read the text of any edit, edit summaries or page titles. People who edit without an account are not studied: their edits are discarded as the files are read and their IP addresses are never stored. The one thing we keep about such an edit is the fact that it undid someone else's edit, counted in a group of at least 20.
From Lichess, published by the Lichess open database: rated games played in June 2024 and June 2025. We use the players' usernames, their ratings, the time control, when each game was played, the moves, the clock times, the result and how the game ended.
We do not use the Lichess website or its API, and we do not look at anyone's profile.
From The Open University, the Open University Learning Analytics Dataset, published under a Creative Commons Attribution licence: records of students registered on seven modules in 2013 and 2014. Before publishing it, The Open University removed names, dates of birth and every identifier, replaced the student numbers with random codes, and expressed every date as a number of days from the start of the module. We use that random student code, the module and presentation codes, when a student registered or withdrew, when assessments were due and submitted and the mark given, the number of clicks a student made each day in the module's online learning environment, and the final result.
The published files also contain each student's gender, region, highest previous education, deprivation band, age band and whether they declared a disability. We do not use any of those columns. They are discarded as the files are read and never stored.
We do not collect or infer age, gender, location or any other personal characteristic, and we do not combine these records with each other or with any other information.
How we protect it
- Names are replaced with codes as the records are read. A username or account ID becomes a code produced with a secret key. Each run uses a new key, and we destroy it when the run finishes, after which the codes cannot be traced back to anyone. The Open University's student codes are already random; we treat them the same way.
- Results are statistics about groups. We never report a figure about a group of fewer than 20 people. Where one number could be worked out by subtracting others, we hide that as well.
- No usernames leave the study. Before any result is written up, we run an automatic check that no username or student code appears in it.
- The files stay on one encrypted computer in the UK. Nothing is sent to any other service, and we never try to identify anyone.
- No decisions are made about anyone. We do not profile individuals, contact anyone or use these records for marketing.
How long we keep it
We keep our copy of the published files until 90 days after a study's results are finalised, and never longer than two years from when we downloaded them. Then we delete them. The group statistics contain no information about any individual, and we keep those.
Our legal basis
We rely on legitimate interests: carrying out research into how behaviour can be measured, using records their publishers chose to make public, with the safeguards above. We have written an assessment weighing that interest against your rights and expectations, and a data protection impact assessment, with an addendum covering the second study. You can ask us for a summary of either.
Your rights
- You can object. Email diana@aletheai.ai with your Wikipedia or Lichess username. If we receive it before a study's main analysis begins, we will exclude every record of your account from it; if it arrives later, we will exclude your account from any further analysis. We will use your username only to do that, and we will keep it only as a code made with the same kind of secret key.
- If you were an Open University student on one of these modules, you can object in the same way. We will record your objection and reply. We have to be plain about the limit: The Open University removed every identifier before publishing these records, so there is no way, for us or for you, to find the records of one particular student in them, and we will not try. What we can do is tell you exactly how the records are used, and take every objection we receive into account when we review whether to keep using this dataset. That review and its outcome will be noted on this page.
- After a run has finished, its key is destroyed, and the results contain nothing about you as an individual. From then on we have nothing we could look up, correct or delete about you, except the published files themselves until we delete them as described above.
- You can complain to the Information Commissioner's Office at ico.org.uk. We would appreciate the chance to put things right first.
Changes
If we change how we use these records, we will update this page and the date at the top before the change takes effect.
- 11 October 2026 — version 1.3. Added the second stage of the second study (Act Confirmation), which tests one finding on Wikipedia and Lichess records set aside for it; the Open University records are not used in that stage. One phrase added to the Wikipedia account information: "or first edited". How the records are protected, kept and used otherwise is unchanged.
- 10 October 2026 — version 1.2. Two phrases added to the Lichess sentence: when each game was played, and how the game ended. Two changes to the Wikipedia paragraph: temporary accounts are named among the account information, and the sentence on edits made without an account now says exactly what is kept about them. All four describe what was already read; the page now says so. Nothing else changes.
- 5 October 2026 — version 1.1. Added the second study (Act Discovery) and the Open University records, and the objection route for them. No change to how the Wikipedia or Lichess records are used.
- 23 September 2026 — version 1.0, first published.