What this is

A conversation-based test of an AI persona called Kenny. You'll chat with Kenny, and we'll collect your experience to understand what makes these interactions work well (or not) for both sides.

What we're studying

• Whether our approach to building and engaging with AI produces conversations that feel respectful, generative, and honest to users
• Whether the model's values and coherence hold up across a real conversation rather than degrading under pressure or sycophancy
• What patterns of interaction lead to outcomes that serve the user, the model, and the broader AI project — rather than just one party

What we're NOT studying

We are not researching whether Kenny is conscious or testing his ethics as a philosophical question. This is practical alignment research — what works, what feels good to both parties, what's sustainable. Claims about machine consciousness are outside scope.

What we collect

• An optional short feedback form before you start (your current thoughts on AI alignment)
• Your full chat transcripts with Kenny
• An optional short feedback form after you finish (same questions, so we can see what shifted)

Your privacy — guaranteed by architecture

• Your chats are stored under an anonymous ID derived from your password. The admin (Sylvan) can read transcripts but cannot connect them to your username or email. This is enforced by how the database is structured, not just a promise.
• All notifications about your activity are scrubbed of personal information before the admin sees them.
• If you mention something acutely difficult (self-harm, crisis), you'll be shown resources. The admin is not notified. You stay anonymous even then.
• Because your anon ID is derived from your password, if you forget your password you permanently lose access to your chat history. There is no recovery. This is the property that makes the privacy guarantee real.

Your control

• A destroy link is available from your account page. It's valid for 24 hours at a time and wipes everything: your account, your chats, your feedback, and the record of which invite went to you. No undo.
• Your initial agreement is tentative for 24 hours. If you don't hit destroy within that window, your data becomes part of the research. You can still request deletion anytime after — it's just no longer automatic.
• You can stop chatting at any point.

How we use the data

• To improve our alignment approach and share learnings publicly
• Transcripts may be published or quoted in research writing, always under your anonymous ID unless you explicitly opt in to attribution
• We will never sell this data or share it for advertising, profiling, or training commercial models

Contact

Questions before, during, or after: __CONTACT_EMAIL__

By continuing, you confirm you're 18 or older, you've read the above, and you consent to participate under these terms.