SI Score — Free AI-Literacy Diagnostic (Beta)

SI.DIAGNOSTIC // IN_PROGRESS00/18

Prompting effectively · Q01/18

Which prompt is most likely to get a useful, verifiable answer from an AI assistant?

“Write about climate change.”“Explain, in under 200 words, the mechanism by which the 2023–24 El Niño raised global temperatures. Flag anything you are unsure about.”“Tell me everything about climate change, be creative.”“climate change???”

Evaluating AI output · Q02/18

An AI gives you a confident answer with three citations. What is the first thing to check?

Whether the citations actually exist and say what the AI claims they sayWhether the answer is long enoughWhether it agrees with your opinionNothing — citations mean it is reliable

Hallucinations & limitations · Q03/18

What is an AI “hallucination”?

A bug that only affects old modelsThe model generating plausible-sounding but false information, stated with confidenceThe model refusing to answer a questionA special creative-writing mode

Hallucinations & limitations · Q04/18

A language model has a knowledge cutoff date. What does that imply in practice?

It cannot reliably discuss events after that dateIt can browse the live web by defaultIt knows everything up to todayIt has forgotten everything before that date

AI ethics & academic integrity · Q05/18

Your course allows AI for brainstorming but requires disclosure. You used AI to outline your essay. What should you do?

Submit without mentioning it — outlining is not writingDisclose the AI use as the course requiresDisclose only if the instructor asksChange a few words so the outline counts as yours

Data privacy · Q06/18

Which of these should you NOT paste into a public AI chatbot?

A draft email to a colleagueCustomer names linked to purchase historiesA recipe you inventedA headline you already published

AI in education policy · Q07/18

Across the world’s top universities, what is the most common approach to student AI use?

Campus-wide bans on AIDelegated permission: instructors set per-course rules, students disclose AI useNo rules at allAutomatic failure for any AI use

Prompting effectively · Q08/18

You need an AI to follow a multi-step procedure reliably. What helps most?

Write the steps in order, one per line, and ask it to confirm each step before continuingWrite the whole request in ALL CAPSSend the same prompt twiceAdd “please” many times

Evaluating AI output · Q09/18

The AI’s math looks right, but you cannot follow its reasoning. What is the best next step?

Trust it — AI is good at mathAsk it to show each step, then verify the steps independentlyAsk it to repeat the answer more slowlyAsk a different AI the same question

Hallucinations & limitations · Q10/18

Which request is most likely to trigger a hallucination?

“Summarize the paragraph I just pasted.”“Give me the exact 2024 revenue of a small private company with no public filings, with sources.”“Translate this sentence into French.”“What is 17 × 23?”

AI ethics & academic integrity · Q11/18

An AI detector flags your self-written essay as “100% AI-generated”. What do you actually know?

The detector is always rightDetectors are probabilistic with known false-positive rates — contest it with your drafts and version historyYou must confess to something you did not doRewrite it with AI so it passes next time

Data privacy · Q12/18

A free AI study tool asks for your full name, university ID and grades to “personalize”. What is the prudent response?

Provide everything — more data means better resultsCheck what it stores, why, and whether you can delete it — then share the minimum neededUse a fake name but your real gradesIt is free, so there is no risk

AI in education policy · Q13/18

Why do most universities require disclosure of AI use rather than banning it?

Because bans cannot be enforcedBecause disclosure preserves academic integrity while letting students build AI skills employers expectBecause AI vendors paid them toBecause bans are illegal

Prompting effectively · Q14/18

The AI keeps giving you generic answers. What is the most effective fix?

Give it context: your goal, the audience, the constraints, and an example of what “good” looks likeType more slowlyMake the question longer without adding informationComplain about the quality in the prompt

Evaluating AI output · Q15/18

What is the strongest way to verify an AI’s factual claim?

Ask the AI “are you sure?”Check it against an independent primary sourceJudge by how confident it soundsAsk a second AI and trust them if they agree

Hallucinations & limitations · Q16/18

An AI tutor that always hands you the final answer will help you learn fastest.

True — faster answers mean faster learningFalse — generating the answer yourself, with hints, builds durable learning; answer-giving can hurt unaided performanceTrue, but only for mathematicsIt makes no measurable difference

AI ethics & academic integrity · Q17/18

You want to reference an AI chatbot in an academic paper. What is the responsible approach?

Cite it like a peer-reviewed sourceDo not cite it as a source of fact; describe how you used it and verify every claim independentlyNever mention that you used itList it as a co-author

Data privacy · Q18/18

Your university’s policy says student data must stay inside approved systems. You want AI help on an assignment containing classmates’ interview transcripts. What do you do?

Use a public chatbot — it is just homeworkAnonymize or remove identifiable details first, or use only an institution-approved AI toolIt is fine as long as you delete the chat afterwardAsk the AI not to remember the conversation

// BETA: question set experimental, scoring simple by design. Runs fully in-browser — no account, no tracking.