Guide

How to use ClinCog

Everything you need to know to get through a week's case smoothly - five minutes, once, and you're set for the rest of the term.

1Set your name, once

The first time you open ClinCog, it asks for your first name. This personalizes the site and lets it recognize you the next time you visit - nothing more.

2Conceptualize: talk to the patient

Each case opens with a short written summary, followed by a chat where you interview the patient directly - ask about onset, duration, severity, anything relevant to building your own clinical picture. You have up to 20 exchanges per case. The patient can only discuss what's documented in their file; if you ask about something outside it, they'll say so rather than invent an answer.

3Evaluate: apply the three instruments

Once you're done exploring, move to that case's evaluation page - a categorical interview (ICD-11 diagnostic criteria), a dimensional questionnaire, and a cognitive task. For the cognitive task, you're the one being measured, not the patient; use your own result as a point of comparison, not a diagnosis.

4Your progress is saved automatically

Every conversation is saved as you go. Close the tab, come back tomorrow, and you'll find each case exactly where you left it. If you want a clean slate, Restart all progress on the dashboard clears every case at once; there is no way to wipe a single interview by accident.

This all happens in this specific browser, on this specific device. Switching browsers (e.g. Chrome → Safari) or devices starts that case fresh - so stick to one browser and one device for the whole term.

If you do need to change device, or you are about to clear your browsing data, go to My progress and use Download backup. That gives you a small file holding every conversation, rating and reflection; Restore from a file on the other machine puts it all back. The file never contains API keys, so it is safe to email to yourself.

5If something looks broken
  • Refresh the page before assuming something's wrong - most display hiccups clear on reload.
  • If a chat reply fails, it won't count against your 20 exchanges - just try again.
  • Still stuck? Write to the person who maintains this platform - say which case and which step you were on.
  • Before clearing your browsing data or reinstalling a browser, take a backup from My progress. Clearing site data erases everything ClinCog has stored, and there is no copy on any server.
6What happens to what you type

Your name, your ratings, your reflections and a copy of your conversations are kept only in this browser, on this device. None of it is stored on a ClinCog server.

What the server does keep is a usage count, never the words. For each reply a patient gives, ClinCog records when it happened, which case, which model, how long it took and how many tokens (pieces of text) it used - not what was said. On a course instance such as uvt.clincog.net that record is kept under your student number, together with how many exchanges you have used with each patient, so the limits hold on every device you use. On the public site it is kept under a random code your browser makes up, which says nothing about who you are. Only the person running the instance can see these records, and they are deleted after 90 days.

The chat is the exception. Each message you send to a patient - together with the earlier messages in that conversation and the case vignette - goes through ClinCog’s server to the company whose language model plays the patient. ClinCog’s server passes it on without keeping a copy, and your name is not sent. Which company that is depends on where you use ClinCog: on a course instance such as uvt.clincog.net, whichever of Anthropic (Claude), Google (Gemini) or OpenAI your instructor has chosen; Google (Gemini) on the public site, clincog.net; and, on a copy adopted by an instructor with their own key, whichever of Anthropic, Google or OpenAI that instructor chose.

Under the terms that apply to how ClinCog uses them, none of the three uses these conversations to train its models. Each keeps them for a limited time to detect misuse: OpenAI for up to 30 days, Google for a limited period, and Anthropic for up to two years if a conversation is flagged for breaking its usage policy.

So treat the chat as a place other people could read. Don’t type real names, details about real people, or anything about your own health. The patients are fictional; keep everything you write about them fictional too.

Two smaller things. To block automated abuse, requests pass through Cloudflare, which checks your IP address and a bot-protection token; neither is passed to the language-model companies. And messages sent from the Contact page are delivered by email to your instructor.

7For instructors: adopting ClinCog

This demo runs on a shared key with modest limits, meant for trying it out. If you'd like to use ClinCog with a real class, you can adopt it: enter your own Anthropic and WHO ICD-11 credentials, and every conversation from your browser runs on your own account from then on, with none of the demo limits. Takes about two minutes.

Adopt with your own keys →