Chatbot Building

Explore & Discover

Chatbots are everywhere — on school websites, in games, on customer service pages. Before you build one, become a detective: spend 20 minutes talking to five different chatbots. Try Google's Bard, the chatbot on your school district's website, and a simple bot on a shopping site. Ask each one the same weird question: "What's the meaning of life?" Notice: which ones feel smart? Which ones feel like they're reading from a script? What makes a response feel helpful versus annoying? Read "How Chatbots Work" at khanacademy.org/computing/computers-and-internet for the technical background. Chatbots have been around since the 1960s — look up ELIZA, the very first one. You're ready for the next step when you can describe three specific differences between a smart-feeling chatbot and a frustrating one.

Learn the Basics

Chatbots need a brain — usually a decision tree or a language model. Learn both. Draw a decision tree on paper for a chatbot that helps a new student at a Utah middle school: it asks if they need help with schedule, lunch, or clubs, then branches based on the answer. Map at least 15 nodes. This is called a flowchart, and it's how rule-based chatbots work. Then watch "How Large Language Models Work" (3Blue1Brown, YouTube) to understand the other kind — the kind that generates responses instead of following a script. Notice the tradeoffs: rule-based bots are predictable, LLMs are flexible but can say weird things. You're ready for the next step when you have a complete paper flowchart with at least 15 nodes for your school helper bot.

Build Your First Project

Build your first real chatbot using Scratch.mit.edu — no account needed to try, but make a free one to save. Search "chatbot starter" in the Scratch community, remix a simple one, and extend it with at least 10 new responses. Use the "ask and wait" block to get user input and "if/else" blocks to match keywords. Build a bot that can answer questions about your favorite game or topic — make it specific, not generic. Add at least one easter egg: a hidden response that only triggers for a specific secret phrase. Test it with three different people and watch where it gets confused. Document every bug in your notebook. You're ready for the next step when your Scratch chatbot handles 10 different inputs and surprises at least one tester with its easter egg.

Experiment & Iterate

Level up to a text-based chatbot in Python on Replit.com. Create a free account and start a new Python Repl called "my-chatbot". Use a dictionary to store question-answer pairs: key is a keyword, value is the response. Write a while loop that keeps asking for input, searches the dictionary for a matching keyword using the "in" operator, and prints the right response (or a default "I don't know yet" response). Add at least 20 keyword-response pairs about a topic you actually know well — Utah Jazz stats, Minecraft biomes, your favorite book series. Use Python's .lower() method so capitalization doesn't break things. You're ready for the next step when your chatbot correctly handles 20 different inputs and never crashes on unexpected ones.

Advanced Techniques

Upgrade your Python chatbot with context memory — real chatbots remember what was said earlier in the conversation. Add a dictionary that tracks the user's name (ask for it at the start) and use it in responses. Add a conversation_history list that appends each exchange. Use simple sentiment detection: if the user's message contains "sad" or "frustrated" or "help," change the bot's response style to be warmer. Explore Hugging Face's free model hub at huggingface.co/models — look for small conversational models you can learn from, even if you're not ready to run them yet. Read their beginner guide at huggingface.co/learn. You're ready for the next step when your chatbot addresses the user by name, remembers at least one earlier thing they said, and changes tone based on emotional keywords.

Final Project Showcase

Design and deploy a chatbot that solves a real problem in your school or community. Ideas: a bot that helps new students find clubs at a Utah middle school, a quiz bot for a subject you're studying, or a bot that recommends hiking trails near Salt Lake City based on difficulty. Build it in Replit, write a README explaining who it's for and how to use it, and publish the Repl publicly. Record a two-minute screen-capture demo (free with Loom at loom.com) showing a real conversation with your bot. Share the demo link on the Scratch forums or with your school's computer science teacher. Collect feedback from five real users and write one paragraph about what you'd improve next. You're ready for the next step when five people have used your chatbot and you've documented their feedback.