Make a Game of It! Explore AI with Games

Explore game-based AI projects for older students and creative AI-powered game design activities for younger learners.

Flexible STEM in Summer

Summer break offers flexible time for STEM, including family activities and student-led deep dives. The out-of-school months invite students to experiment in STEM areas of personal interest, either pursuing topics that may not come up in the typical school year or exploring more deeply than the school schedule allows. This post is the first in a series about experimenting with AI and machine learning over the summer.

Games as an AI Gateway

Game-based projects are a fun and accessible way for students to explore AI and machine learning. From inventing house rules with a chatbot to training an agent to navigate a maze, games and puzzles offer approachable ways to learn how AI systems generate ideas, make decisions, and improve through feedback.

Explore Machine Learning and AI with Game Projects

For older students, the game-focused projects highlighted below offer a starting point for exploring machine learning algorithms, working with Google Colab, and developing smart agents that can solve a maze or play tic-tac-toe, Connect 4, or Minesweeper.

Step-by-step directions and starter code for these projects are provided. No coding experience is required.

• 1. Tic-Tac-Toe

  How many moves ahead does an AI agent need to think to play tic-tac-toe well? In this project, students use the Minimax algorithm to explore decision trees and investigate how adjusting the AI agent's search depth affects gameplay. Questions to consider: How does looking farther ahead change the decisions an AI agent makes? What are the considerations or trade-offs when setting the depth of exploration of the game tree?

  

  https://www.youtube.com/watch?v=JWRxdoPrk7M
  
  

• 2. Connect 4

  Can you create a hard-to-beat Connect 4 opponent? Connect 4 has too many possible moves to analyze exhaustively. That's where understanding the game tree comes in. In this project, students explore how AI uses Minimax, alpha-beta pruning, and heuristic scoring to make smart decisions while analyzing only part of the game tree. Questions to consider: Why is a heuristic search algorithm useful for this type of game? What aspect of the Connect 4 game can be used to help refine the evaluation function?

  

  https://www.youtube.com/watch?v=rbmk1qtVEmg
  
  

• 3. Minesweeper

  Can you train an AI agent to solve a Minesweeper puzzle? Minesweeper is a classic game that uses deductive reasoning to figure out which squares are safe. In this project, students experiment with reinforcement learning to train an AI agent to play the game. Using a Deep Q-Network (DQN) algorithm to evaluate possible moves, students explore how hyperparameters and the number of training episodes affect performance. Questions to consider: What does a DQN do instead of memorizing every single board configuration? Why is the number of episodes used in training important?

  

  https://www.youtube.com/watch?v=EGp2LHOEWTI
  
  

• 4. Solve a Maze

  Can an AI agent learn from its mistakes to solve a maze? A maze is not technically a game, but this project explores reinforcement learning using the Q-Learning algorithm to train an AI agent that improves through trial and error as it solves a maze. Questions to consider: What is the role of rewards and penalties in reinforcement learning? What are the potential drawbacks of reinforcement learning?

  

  https://www.youtube.com/watch?v=QUNM-QyM5PA
  
  

Games, AI, and Younger Students

Students do not have to start with code to begin experimenting with AI. Working together with a parent or guardian and following the age requirements for the AI tool being used, younger learners and families can use chat-based AI tools like ChatGPT, Claude, or Gemini to invent new rules for a family favorite, remix familiar games, or develop game variants.

Using AI as a brainstorming tool, students and their families can talk to it about the games they play and explore changes, house rules, and variations to try. For families that already enjoy games, this can be an excellent, low-stakes way to explore AI in a way that supports creative thinking, reinforces engineering design, and directly translates into family time.

Change Up a Favorite Game

To get started, choose a favorite game and try one of the options below. Card games work especially well for testing variants. Whether you devise new rules for Uno, Sleeping Queens, Sushi Go!, Phase 10, or Crazy Eights, you might end up with a new variation that is just right for you and your family.

• Use AI to invent house rules. Custom rules can be used to add new variety to a game that has been played so many times that it's gotten stale or to make a game work better for your specific needs. Questions to ask: How can we make this game more strategic? What variations will add new life or novelty? What rules can we add to make this game less luck-based? What are three silly variations for this game that are still fair?
• Ask AI to help make the game more (or less) difficult. If a game is starting to feel "too easy," AI can offer ways to help increase the strategy. This doesn't have to involve cumbersome rules. With just a few tweaks, a game may take on more nuance. Ask AI to help with difficulty tuning for multiple ages for a favorite game. This can be especially helpful for families with players of different ages. Questions to ask: What rules can we add to make the game accessible, fun, and fair for both younger and older players? What rule changes would help younger players stay competitive without making the game less fun for older players?
• Ask AI to help add more math to a favorite game. This is a version of difficulty tuning, but if math is an angle that your students enjoy or need to practice, it can often be easy to find ways to integrate a unique math-based rule that changes the game. It doesn't need to be hard to be effective. Questions to ask: How can we add simple math to this game without slowing it down? What scoring rule would encourage players to practice multiplication or fractions? Can you suggest a few math-based rule changes for different age levels? How can we make the game more strategic using math?
• Use existing cards to devise a new game. If your kids have Pokémon, Magic: The Gathering, or other collectible cards lying around, they may have fun talking with AI about ways to create a game that uses the cards. It might be a version of another game that feels familiar but is able to take advantage of the specific cards they collect. Questions to ask: What kind of game could we play using these cards? Can you suggest three different game ideas that would use the cards we have in different ways? How could we use these cards with rules from another game we already know?

Reinforce STEM Learning

Coming up with new ways to play favorite games is fun and can lead to hours of experimentation as students test new rules and modifications. To add a STEM layer to the exploration, encourage students to collect and analyze data. Set up a table to log the changes they try and record the outcomes. Players might rate variations or rules on a scale of 1-5, for example. After a few games, students will have data showing which changes worked best and which ones they enjoyed most. If they label each rule or variation by the type of change it represents (e.g., difficulty, math, more strategy), they can also draw conclusions about the kinds of house rules that work best for their players.

As they gather data about what they liked and didn't like and what was too hard or too easy, students can practice communicating their results and using feedback to refine their ideas. By discussing the results with the chatbot, they can brainstorm and test additional iterations.

If your students try AI and machine learning projects or explore modifying favorite games this summer, we would love to hear about it! You can reach us at [email protected] to tell us about your game night.