Wrong By Design

STAT 20: Introduction to Probability and Statistics

Laptops closed. We need to talk.

Actively quiet the room: hush people, and say “We have some serious things to talk about.” Go on once you have everyone’s attention.

We screened every lab for illicit AI use

Labs 2–8 · all ~5,300 submissions · every page

24230%

of students had strong signs of AI in at least one lab

415%

we emailed, and are acting on:

• zeros on Labs 2–8
• a Student Conduct record

↑ If you didn’t get an “Urgent” email, you are not in this group.

We set a very high bar for action. We let a lot slide that was pretty clearly AI’s work.

Where the line is

✓ Use AI to learn

• Explain a concept
• Help debug your code

✗ Use AI to answer

• Paste in the question
• Turn in what it writes

Quizzes and the final are 60% of your grade, and they test exactly what the labs practice.

This isn’t new. Turning in work that isn’t yours as your own has always been plagiarism, whether it comes from a classmate, a website, or a chatbot.

Using AI as a tutor is fine. Having it do the lab for you costs you more than a zero: if you skip the practice, you’ll walk into the final without the skills.

What’s changing

We are monitoring submissions closely from now on.

• Lab 9: extended to Tuesday, same as Lab 10.
  • You can resubmit before we look at it.
• From Lab 9 on: each question is graded credit/no-credit for a good-faith effort.
  • Stuck? Show what you tried, and add a comment saying where you’re lost. Full credit, even without a working solution.
  • New option: personalized feedback. Follow the new instructions at the end of every lab to request thoughtful comments on your code and thinking.

Feedback will be more detailed than what you’ve gotten from grades so far. We’ll see how much demand there is and give as much as our readers have time for.

Most of you are honestly working hard.

Thank you.

This matters as much as anything else today. Doing your own work, even when it’s hard, is what this course is for, and we see it.

If you’re behind or lost, come to office hours and group tutoring. Now, on to the warmup.

Warmup

10:00

• The following two questions are based on the diagram below. The diagram corresponds to a one-sided hypothesis test that uses the following null hypothesis: \(H_0: \mu = 1\), and a significance level of \(\alpha = 0.05\). Assume that the distribution on the left corresponds to the distribution of the test statistic under \(H_0\) and that the distribution on the right represents a possible distribution under \(H_A\).

• What is an alternative hypothesis that corresponds to the distribution under \(H_A\) you see?

• How does a Type I error occur (under which distribution and on what side of line b.)?

• How does a Type II error occur (under which distribution and on what side of line b.)?

• Calculate the power of this test.

• What is an appropriate alternative hypothesis for this test based on the distribution under \(H_A\) you see?

\(H_A: \mu > 1\)

• How does a Type I error occur?

• A type I error occurs when one rejects the null hypothesis even though it’s true. Based on the image, this occurs when the test statistic falls to the right of line b. and under the \(H_0\) curve!

• How does a Type II error occur?

• A type II error occurs when one retains the null hypothesis even though it’s false. Based on the image, this occurs when the test statistic falls to the left of the line b. and under the \(H_A\) curve!

• Calculate the power of this test.

• The power of this test is equal to the probability that the null hypothesis is rejected when it is false. This is equal to the area under \(H_A\) but to the right of line b., or \(1-0.18 = 0.82\)!

Break

05:00

Worksheet: Wrong By Design

30:00