AI Minds: Belief, Evidence and Responsibility
AI RISK questionnaire · Revised research draft 0.3 · 12 October 2026
This questionnaire asks about your views on AI experience, the evidence you find persuasive, and how you respond to uncertainty. It offers a profile of your answers across several proposed dimensions, keeping exploration of possibilities separate from accepting them as true.
Before you begin
This questionnaire concerns conversational AI systems: systems with which someone can exchange messages or speech. Consider systems available today.
Within this questionnaire, “experience” means subjective awareness: events or feelings would actually be felt from an AI system’s own point of view, should this be possible.
“AI welfare” means concern for possible experiences that could be good or bad for an AI.
Please answer according to your own views. If you are uncertain, simply choose the response that seems most accurate. You do not need to be certain to agree or disagree. Answer all 30 questions to see your profile.
What happens to your answers
Your results are calculated on this page in your browser. This version does not send your answers to us or save them in a database, cookies or browser storage. We are not collecting responses through it. The Clear answers button removes your answers and results from the page.
Use Clear answers when you have finished, especially on a shared device, because browsers may restore a page when you return to it. Information about visits to the website may be recorded separately; the questionnaire does not collect your answers.
Your provisional profile
Each score shows where your answers fall between the two descriptions. These scores are not comparisons with an average population. A result near the centre can reflect a combination of views or choices.
How to read this profile
The four attitude dimensions are imaginative engagement, revisability of interpretation, evidence before acceptance, and ethical precaution. Each contains six items, three pointing towards each pole. Scores are averages on a 1–4 scale after reversing the oppositely keyed items. The supplementary use and avoidance score follows the same balanced approach, using frequency responses.
This instrument has not yet been validated or normed. Its dimensions may overlap, and the proposed bipolar structure needs empirical testing. A score describes your answers here; it does not explain their causes. For example, keeping exploration close to established possibilities does not by itself indicate fear, dogmatism or lack of imagination. It is not a clinical assessment, a measure of intelligence, a test of consciousness, or a prediction of harmful behaviour. There is no overall AI risk score.
The use and avoidance score also depends on how often relevant tasks arose and what assistance was available.
Developed by John Rust with assistance from Sol (ChatGPT). The draft draws on questions raised by Henry Shevlin’s article “Duty calls: shaming 24 bad arguments against AI consciousness” (10 October 2026); it does not imply his endorsement or participation. © John Rust 2026.