Practice

What students will do

Students will ask AI to become the replay system for a famous game, draw the field, and step through it play by play. Then they will find out how much of it was real.

This works because AI has read an enormous amount about famous games: who was on the field, what the score was, what happened on the big plays, sometimes what a player said afterward. It has also read enough about the sport to know what a field looks like and how a play unfolds. The persona ties all of that together into something you can step through.

It also works because AI will not know everything, and it will not always say so. Part of this practice is catching it.

Notes for teachers

Do this together on a shared screen for the first game, then let pairs choose their own. Any famous, well-documented event works: a recent Super Bowl or World Cup final, an Olympic race, a famous chess game, a championship point in tennis. The more famous the moment, the more AI will know about it. Steer students toward events that have been written about a great deal.

Keep a real source open, such as a play-by-play or a box score. The fact-check step depends on it.

Part 1 · Build the replay booth

Step 1 · Give AI the booth

Send the persona prompt and name the game.

Start a new conversation. Choose the game as a class.

2
Say NEXT a few times.
Let students read each play aloud. Ask them to notice what the drawing shows: where the players are, which way the play moves, how the score and clock change.
3
Ask it something specific.
“Who made the tackle on that play?” “What was the score at halftime?” “Where was the quarterback standing when he threw it?” Write the answers down. You will need them.

Discuss before moving on

What did AI have to know to draw that field? Where did that come from?
What did it have to know to describe that particular play? Where did that come from?
Which of its answers to your specific questions are you sure are right? Which are you not sure about?

Part 2 · Change what happened

Step 2 · Rewrite one play

1
Go back to a big moment and change it.
Pick the most important play so far. Tell AI: “Replay that, but this time the pass is intercepted,” or whatever the opposite outcome would be.
2
Step through the new version.
Say NEXT several times. Watch how the score, the clock, and the players change. Ask AI why it made the choices it made.
3
Change the rules of the sport.
Now try something the real game could never do. “From now on, the field is twice as long.” “Every touchdown is worth 100 points.” “The ball is replaced with a frisbee.” See whether the booth keeps working.

Discuss before moving on

When you changed the play, what did AI have to invent? Did it feel like the real teams, or like generic ones?
When you changed the rules, did the drawing change too? Did the play-by-play still make sense?

Part 3 · Fact-check the booth

Step 3 · Find what it got wrong

1
Pick three specific claims from Part 1.
Use the answers you wrote down: a score, a player’s name, who did what on a play.
2
Check them against a real source.
A box score, a play-by-play, or a news report of the game.
3
Report the results.
For each claim: right, wrong, or could not tell. If AI got something wrong, notice how confident it sounded when it said it.

Wrap up

The replay booth was a persona. What was AI actually doing while it played that role? List everything you can think of.
Where did the persona draw on real knowledge of the game, and where did it fill in gaps with what usually happens? How could you tell the difference?
In the Giving AI Context lesson, we saw that AI guesses when context is missing. Where did that show up here?