{"id":"study-overfitting-practice-test-analogy","audience":"college","category":"explaining","technique":"audience","track":"study","title":"Explain overfitting with the memorized-practice-test analogy","free":false,"task":"Your AI and Society course (no programming) asks you to explain overfitting. Get exactly 3 numbered sentences using only the professor's practice-test analogy (what training data is, what overfitting is, why it fails on new data), then one line starting 'Check:' with a question to test yourself, under 80 words, plain text. No code, no math and none of the words 'parameters', 'loss' or 'variance'.","materialLabel":"Your lecture notes (attached to your prompt automatically)","material":"STS 2100 - AI and Society (no programming required) - Lecture 6 (Prof. Lindqvist)\n\nHow a machine learning system \"learns\":\n- It studies lots of examples (training data) and finds patterns.\n- Then it's tested on NEW examples it hasn't seen.\n\nPROF'S ANALOGY: the practice test\n- Training data = a practice test with the answers.\n- A student who UNDERSTANDS the material learns the ideas behind the answers, so they do fine on the real test.\n- A student who MEMORIZES the practice answers (question 4 = \"C\") aces the practice test but fails the real test, because the questions are different.\n- Overfitting = the system memorized its training examples, including random quirks, instead of learning the general pattern.\n\nExam question: \"Explain overfitting to someone who has never studied computer science.\"\nProf. L: \"If you write 'high variance' or 'low training loss' without explaining it, no credit.\"\n\nTextbook terms we are NOT using: parameters, loss function, variance, regularization.\nFinal paper due Dec 8.","rules":[{"type":"numberedItems","count":3},{"type":"maxWords","max":80},{"type":"plainText"}],"mistakes":[{"title":"Asking the general question","why":"The AI explains with technical terms like variance and loss, which your professor won't credit.","fix":"Say: \"Use only the practice-test analogy from my notes.\"","checks":["r0"]},{"title":"Not saying how many sentences","why":"The AI picks its own number, and it usually picks more sentences than you need.","fix":"Be exact: \"Exactly 3 numbered sentences.\"","checks":["numberedItems"]},{"title":"Not banning jargon","why":"AI defaults to the textbook terms.","fix":"Add: \"No code, math, parameters, loss or variance.\"","checks":["r1"]},{"title":"No way to test yourself","why":"The AI explains and stops, so you never check whether it stuck.","fix":"Add: \"End with one line starting 'Check:' with a question, no answer.\"","checks":["r2"]},{"title":"Trusting the AI to count words","why":"AI is bad at counting. Ask for a total and it often lands 20% over.","fix":"Cap each sentence: \"each under 20 words.\"","checks":["maxWords"]},{"title":"Forgetting where it will be read","why":"AI formats for a screen, with bold and headings that show up as stray symbols on the exam.","fix":"Say: \"Plain text, no bold, no headings.\"","checks":["plainText"]}],"promptHints":[{"label":"Names the analogy","patterns":["practice test","memoriz"]},{"label":"Says how many sentences","patterns":["\\b(3|three)\\b.{0,25}sentence"]},{"label":"Bans jargon","patterns":["variance","parameters","loss","no (code|math|jargon)"]},{"label":"Asks for a check question","patterns":["check:","check question","test (me|myself)","quiz me"]},{"label":"Sets a length limit","patterns":["\\d+\\s*words","word (limit|count)","under \\d+"]}],"hint":"Tell the AI the analogy, the three ideas to cover, the terms to avoid, and that you want a question to test yourself at the end.","rubricLabels":["Uses only the practice-test analogy: training data is like a practice test with answers, overfitting is memorizing the answers instead of learning the ideas, and that fails on the real test with new questions","Contains no code, no math and none of the words parameters, loss, variance or regularization","Ends with one line starting 'Check:' that asks a question about overfitting, and does not answer it"]}