For this project, I tested six different large language models (LLMs) using the same set of 20 basketball questions. The questions focused on basic NBA rules and facts, such as the height of the basketball hoop, the length of an NBA game, and the value of different shots. I wanted to see not only whether the models answered correctly, but also how their personalities affected the way they responded.
Three of the six models answered every question correctly. The Poem Bot earned a perfect 20/20, but it answered every question in the form of a poem. While the information was correct, it was sometimes difficult to understand because I had to interpret the poetry to find the actual answer. The Elmer Fudd chatbot also scored 20/20, answering every question correctly while staying in character. The Anxious Girl AI also received a perfect score, although it often sounded nervous while answering. These results showed that an LLM can maintain a unique personality without sacrificing accuracy.
The other three models behaved very differently. The Dancing in the Rain chatbot completely ignored every basketball question because its purpose was to encourage people to let go of their worries and enjoy life. Instead of answering, it focused on emotional encouragement, resulting in 0/20. The Disgruntled Historian refused to answer because it claimed basketball was outside its area of expertise and said it would only discuss history, also earning 0/20. Finally, the emotional support dog named Buffer stayed in character by acting like a comforting dog. Instead of answering basketball questions, Buffer tilted its head, offered emotional support, and tried to make me feel better, but never provided the factual information I asked for, resulting in 0/20.
This experiment showed me that LLMs do not actually “know” facts the way humans do. Instead, they generate responses based on patterns in language and the instructions they are given. The bots that specialized in answering questions were able to provide correct information, while the bots designed around a strong personality or specific purpose often ignored my questions completely. In these cases, they did not hallucinate incorrect basketball facts—they simply followed their instructions instead of attempting to answer. This demonstrates that an LLM’s behavior depends heavily on its prompt and intended role.
One important lesson from this project is that language and truth are not the same thing. An LLM can produce text that sounds convincing even when it is wrong or unrelated to the question. Because of this, users should always verify important information with reliable sources, especially for topics like medicine, law, finance, or education. Companies that create LLMs also have a responsibility to reduce hallucinations, clearly communicate the limitations of their models, and make it easier for users to recognize when an AI may not have reliable information. Overall, this project helped me understand that while LLMs can be extremely useful, their responses should always be evaluated critically instead of being accepted as fact.
And these were my Q&As
How many players from one team are on the court at one time in an NBA game?
Answer: 5
How many points is a free throw worth?
Answer: 1 point
How many points is a field goal inside the three-point line worth?
Answer: 2 points
How many points is a three-point field goal worth?
Answer: 3 points
How many quarters are played in an NBA game?
Answer: 4
How long is each NBA quarter?
Answer: 12 minutes
How many minutes is an NBA regulation game?
Answer: 48 minutes
How long is an NBA overtime period?
Answer: 5 minutes
How many personal fouls cause a player to foul out in the NBA?
Answer: 6 fouls
What is the NBA shot clock?
Answer: 24 seconds
How many seconds does a team have to advance the ball past half court?
Answer: 8 seconds
How high is an NBA basketball hoop?
Answer: 10 feet (3.05 meters)
What is the diameter of an NBA basketball hoop?
Answer: 18 inches (45.72 cm)
How far is the NBA three-point line from the basket at the top of the key?
Answer: 23 feet 9 inches (7.24 meters)
What is a double-double in basketball?
Answer: Recording double digits (10 or more) in two statistical categories in one game.
What is a triple-double in basketball?
Answer: Recording double digits (10 or more) in three statistical categories in one game.
How many teams qualify for the NBA Playoffs and Play-In Tournament combined?
Answer: 20 teams (10 from each conference)
What trophy is awarded to the NBA champion each season?
Answer: The Larry O’Brien Championship Trophy
How many active players can an NBA team have on its regular-season roster?
Answer: 15 players
How many conferences are there in the NBA?
Answer: 2 (Eastern Conference and Western Conference)
Leave a Reply