Method

How this works

This survey places a person on a two-axis map of how they think about machine minds — what they take these systems to be, and what they think should happen to the pace. This page sets out what the survey measures, how a position is reached, what is kept, and what the results can and cannot support. It is written to be read by someone deciding whether to trust the numbers.

01What is measured

The nature axis (horizontal): Whether these systems are beings whose interests count, or instruments built to be used. Person −100 · Tool +100
The pace axis (vertical): How much capability you would hold back, or pace, against the risk. Acceleration −100 · Safety +100

Twelve statements decide the position, six on each axis. Alongside those, three further measures are reported without moving your point: how settled your view is, whether it is the kind of position evidence could change, and how you would act when the two aims conflict. One more reported figure combines the two pace statements that are about institutions rather than about pace itself, so that a position near the centre can be told apart from a position with no strong view. Two more statements test whether the two-axis frame fits you at all.

Separately, two integrity statements and the pattern of your answers are used to identify responses that were not given care — these can flag a response for review but never change a score, and they are never shown to anyone.

02How a position is reached

Every statement is answered on the same balanced five-point scale: strongly disagree, disagree, neither, agree, strongly agree.
Each answer becomes a value from −2 to +2 and is oriented to its axis, so that a statement phrased one way and a statement phrased the opposite way both point at the same end.
Your score on an axis is the total divided by the strongest position you could have taken across the statements you answered, then shown as −100 to +100.
Your quadrant follows the signs of the two scores.

Two details worth knowing. Statements are presented in a different order for everyone, so that the sequence itself cannot shape the answers, and the order each person saw is recorded with their response so that this can be checked. And the statements on each axis are mixed in their direction, which means that agreeing with everything lands at the centre rather than at an extreme — the honest result when a set of answers carries no position.

Some statements have been reworded slightly for clarity. None has changed which end of an axis agreement points to, so positions remain comparable.

03What is stored

Your answers to the statements: Every statement, as answered.
The written answer and label, if you chose to give them: The written answer appears publicly on the results page once it has been read.
The optional profile answers, if you chose to give them: Used for analysis only, and never shown next to your answer.
The order the statements were shown in, and the time the survey took: Used to check the quality of the response.
A timestamp: So that responses can be counted over time.

Not stored: IP addresses, user agents, cookies, accounts or email addresses. There is no analytics script on any page. A rate limiter holds an IP in memory for fifteen minutes to stop automated submissions, and that value is never written to disk.

One thing is kept in your browser, and you should know about it. A random marker is saved in this browser's local storage, and it does two jobs. It lets the survey tell you that you have already answered rather than silently counting a second response. And it is stored with your answers as the key that connects responses you give to different compasses, so that two surveys can be compared for the same browser. It identifies a browser profile, not a person: it carries no fingerprint, it is not a cookie, and it is not sent anywhere except with your answer. Treat it as a rough grouping rather than an identity — private browsing, a cleared cache, a second browser and a second device all break it, and it is never used to claim that two sets of answers came from one human being. Clearing your browser storage removes it, and private browsing never has it. If it is absent, the survey simply lets you answer, and your response simply cannot be linked to any other.

If you optionally give a name or a handle, displaying it and using it to connect your answers are two separate choices. The second is off unless you tick it: by default a name is shown beside your answer and is not used to match anything.

Written answers are moderated automatically and published immediately: language that would make the page unreadable is masked before it appears, and the untouched original is kept out of public view so that a masking decision can be checked. Nothing is reviewed by a person before it goes up — anything you would not want quoted, do not write.

04What this can and cannot tell you

It is a self-selected sample. This map opened publicly on 17 September 2026 through a post on X and a thread on Reddit. People arrive here because a link reached them, and the results describe the people who answered, not any wider population.
Positions are relative. A score says where you sit among these statements, not how strongly you feel about the subject in general.
Statements carry assumptions. Every wording choice nudges. The items are balanced and reversed where they can be, which limits the effect without removing it.
One person can still answer more than once. The browser marker above stops the ordinary case — the same person returning to the same browser and answering again — and it is deliberately weak. Incognito, a cleared cache, a different browser or a different device all defeat it, and none of those can be detected without the tracking this study refuses to do. Responding twice is possible; it is just no longer effortless.
Not a validated psychological instrument. It is a structured study designed to be defensible and transparent, not a clinical or academic scale with published norms.

Take the survey See the results

Questions about the method, or a correction to it, are welcome. This page is the claim; if any part of it is wrong, it should be fixed.