Week 5 — Readings & Resources · Prompting III: Examples, Structure & Control

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Objective covered: Objective 2 — Use effective prompting techniques to produce high-quality, well-verified AI output.

How to use this page

Everything here is a link to an external resource — open it in your browser. Nothing needs to be downloaded or purchased.

This week's load is deliberately focused: two readings + one video + one interactive, grouped by the ideas from the lecture. Read or watch one item per group and you're ready for the quiz; do all of them and you'll be well ahead. Total time is roughly 30–40 minutes if you do everything, less if you pick one per group.

Order that matches the lecture: ① what few-shot prompting is → ② the control toolkit → ③ source-verification → ④ the tools.

The habit again this week: before you trust any source — in these readings, from a chatbot, or anywhere — ask: Is this claim verifiable? How would I check it?

① What Few-Shot Prompting Is

Maps to Lecture Segments 2–3. Zero-, one-, and few-shot prompting; using examples to teach voice and format.

Reading — "Prompt engineering" (OpenAI Cookbook / Docs)
🔗 https://platform.openai.com/docs/guides/prompt-engineering
Why it earns the click: the section "Provide examples" explains how including examples in a prompt (few-shot) helps the model match a pattern you've demonstrated — directly covering Skill 6. Skim the page for the examples-related sections; you don't need the entire guide this week. Free to access.
⏱ ~10 min (focused skim)

Reading — "Introduction to prompt engineering" (Microsoft Azure AI documentation)
🔗 https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/prompt-engineering
Why it's assigned: includes clear explanations of zero-shot, one-shot, and few-shot with worked examples — exactly the vocabulary drill from the lecture. Free, no account required.
⏱ ~8 min

② The Control Toolkit — Structure, Constraints & Regenerating

Maps to Lecture Segments 5–6. Specifying count and format, setting constraints, regenerating (and why it doesn't fix facts), asking for expansion, asking for guidance.

Reading — "Few-Shot Prompting" (Prompt Engineering Guide — DAIR.AI)
🔗 https://www.promptingguide.ai/techniques/fewshot
Why it earns the click: a clear, example-rich walkthrough of few-shot prompting — how it works, why multiple examples help the AI generalize better than a single sample, and the limitations to watch for (including the risk of the model over-fitting to surface features). Pairs directly with group ① above. Free, no account required.
⏱ ~8 min

Reference — "Zero-Shot Prompting" (Prompt Engineering Guide — DAIR.AI)
🔗 https://www.promptingguide.ai/techniques/zeroshot
Why it's here: the companion zero-shot page explains what happens when you give the AI a task with no examples at all — helpful for understanding where the zero/one/few-shot ladder starts. Read group ① first, then this and the few-shot page to complete the picture.
⏱ ~5 min

③ Requesting Sources & Verifying Them

Maps to Lecture Segment 6. AI can generate plausible-looking but fabricated citations; verification is always manual.

Reading — "Evaluating Sources in the Age of AI" (Purdue OWL)
🔗 https://owl.purdue.edu/owl/research_and_citation/evaluating_sources_of_information/index.html
Why it's assigned: the Purdue OWL guide on evaluating sources gives the underlying skill for what you do after an AI provides citations — not AI-specific, but the most important move you'll make with AI-generated references. Read the "Why Evaluate?" and "Criteria" sections; skip the discipline-specific appendices for now.
⏱ ~8 min

Reference — "How to check if an AI citation is real" (practical checklist — your own verification workflow)
There is no single authoritative link for this — but here is the workflow to internalize (no link to click; this is the skill itself):
1. Open every DOI or URL the AI provides in a new browser tab.
2. Confirm the article title and author match what appears at the publisher's or journal's site.
3. If it 404s, redirects oddly, or the author's name doesn't appear in the journal's record — it's likely fabricated.
4. Cross-check with Google Scholar (https://scholar.google.com) or your library's database.
5. If you're not sure, ask the AI: "Are you certain this citation is real? If not, please say so." A model that hedges is more useful than one that bluffs.

④ The Approved Assistants — Homepages

You only need one assistant this week (free tiers are fine). These are the official product pages.

• ChatGPT (OpenAI) 🔗 https://chatgpt.com
• Claude (Anthropic) 🔗 https://claude.com
• Gemini (Google) 🔗 https://gemini.google.com
• Copilot (Microsoft) 🔗 https://copilot.microsoft.com

Pick-one quick path (≈18 min total)

In a hurry? Do exactly these two and you'll be ready for the quiz:
1. Read the Microsoft Azure AI prompt engineering page (group ①) — it covers the zero/one/few-shot vocabulary cleanly.
2. Watch the "Advanced Prompt Engineering" video (group ②) — it covers the control toolkit including the regenerate ≠ fix point.

Heads-up (links rot): these point to outside sites that occasionally move or rename pages. If a link ever fails, search for the resource title directly using Google Scholar or the publisher's site search. Nothing here is downloaded or redistributed — all resources stay as links to their original sources.