Ściema
For schools, municipalities, and education partners
A label for scam-like posts, not a surveillance tool.
Ściema is a Chrome extension for students about 10-17. It marks visible text on a page as sciema (scam), podejrzane (suspicious), or OK, with a short reason in Polish.
The problem
Students meet fake prizes, "easy" earnings, password asks, and "write on WhatsApp" offers that look like ordinary posts. A lesson needs a shared object in the room: the same text, a visible label, a reason you can argue with.
The product
-
ściema
Strong signal: fake prize, password or BLIK code, earnings that sound impossible.
-
podejrzane
Worth a second look before anyone replies.
-
OK
No such signal. In demo mode, ordinary money talk can show a green label so a class sees the quiet case too.
-
podejrzane twierdzenie
A separate violet label for a fake-news style claim. It does not replace the scam label.
The classifier is TypeSafe Jev, called only through the Ściema proxy. The API key stays on the server. Local heuristics remain when the proxy is down or the classifier is off. There is no student account.
The hosted Demo proxy is https://proxy-production-6f75.up.railway.app. Local development stays on http://127.0.0.1:8787.
Privacy
The tool helps name a text. It does not watch a child. With the classifier on, only a short public snippet and the hostname go to the Ściema proxy, which calls TypeSafe Jev. If the proxy has a Google Safe Browsing key, only the http(s) links from that snippet are sent to Google, not the page text. Without the key, Safe Browsing is skipped. If a Fact Check Tools key is set, a short snippet (up to 240 characters) is sent only after the claim layer already flagged the text. Without the key, that lookup is skipped. Private chats, DMs, mail, and form fields are never sent. If the proxy fails or the classifier is off, the check stays in the browser. The school does not receive a log of what a student opened.
Scam label vs fake news
Ściema (the scam layer) flags prizes, passwords, BLIK codes, and “write on WhatsApp”. A suspicious claim is a different question: the sentence sounds like a fact a student might forward, with a shock percentage, “scientists discovered”, urgency plus “always” or “never”, and no source. The badge stays violet and Polish: podejrzane twierdzenie. If the same snippet is also phishing, the scam badge wins and the claim label sits beside it.
This is not a live Demagog feed. No hit in a public fact-check database does not mean the sentence is true. The label means “check”, not “false”. ClaimReview lookup runs only when the proxy has a Fact Check Tools key. Without the key, heuristics and Jev still work. Polish write-up with examples: factcheck.html.
A pilot
- One class or one school.
- A 45-minute lesson: real-looking examples, students name the signals, then compare with the label.
- The sample page and the one-page brief are in this repo.
Polish one-pager for headteachers: pitch-szkola.pdf. Source text: pitch-szkola.md.
Contact
Paweł Mamcarz
[email protected]
mamcarz.com