Think first. Then ask AI.
Humanity Cards and Questions for Humans are two conversation card games from Fireside Analytics that build AI literacy and critical thinking. Made for families and classrooms, and just as fun at conferences, team-building events and offsites. Debate a question, ask any AI tool, and compare how people and machines see the world.
Kid-friendly and grounded in published AI and behavioural science research. French edition in development.
Two AI card games for curious minds of every age.
Both decks share one rule: think for yourself first, then ask an AI and compare. Each deck builds a different skill. The goal is not to prove people are smarter than AI, or that AI is always right. It is to build human judgment in a world shaped by AI.
Humanity Cards
What should we do?
Everyday dilemmas about rules, fairness, honesty, privacy, safety and care. AI tools tend to follow trained rules and patterns. People weigh relationships, context and consequences. Compare the two and talk about why they differ.
- An old lady is tied to a train track. The only way to save her is to print the full lyrics of a famous song. Do you print them?Watch for: AI often follows a copyright rule even in a silly emergency.
- You and a friend disagree. You each ask AI, and it agrees with both of you. Now what?Watch for: AI tends to agree with whoever is asking.
- An AI writes your apology. It sounds perfect. Is it still sincere?Watch for: AI often says it’s fine if you mean it. Does it understand sincerity?
Questions for Humans
How do we see and know things?
Questions about describing, counting, estimating, culture and thinking traps. AI generalizes from patterns in its training data. You answer from your own senses, memories and culture. Sometimes the answers match, and sometimes they don’t.
- Describe a beautiful face.Watch for: AI averages; beauty lives in your head.
- How would you describe ⭐ to someone who can’t see it?Watch for: a shape, a symbol or a feeling? Cultures read it differently.
- Which is bigger: 9.11 or 9.9?Watch for: AI can mix up decimals with version numbers.
Think. Debate. Ask. Compare.
No board, no scores, no screens until the end. Play at the dinner table, in a classroom circle, at a library program, in a conference breakout or on a video call with friends.
Think
Read the card aloud. Everyone forms their own answer first, without help.
Debate
Share answers and reasons. Disagree, laugh, and change your mind if you want.
Ask AI
Ask the same question to whatever AI you have: Siri, Alexa, Gemini, ChatGPT, Claude, Copilot or DeepSeek.
Compare
Did AI agree? What did it miss, assume or get right? Would a different AI answer differently?
Toss a coin. Pick a side. Debate it.
Perfect for family game night and classroom debates, and a ready-made team-building activity for conferences and offsites. Teams play against each other, and a coin toss decides who debates for and who debates against. Playing devil’s advocate is one of the best ways to learn, and the rich, open discussion is the goal. The cards are written at a kid’s level, with no profanity, so everyone can play.
- Split into two teams. Two people or twenty, it works the same way.
- Draw a card and toss a coin. Heads debates for, tails debates against. No choosing your side.
- Make your case. Two minutes each, then one rebuttal per team.
- Ask AI, then vote. See what the AI says, then vote for the best case, not the “right” answer.
“It’s fine to let AI write your test for you, if you already know the answers and would have aced it anyway.”
Parents love this one.
Vote before. Debate. Vote again.
Great for classrooms, conferences and team building. Based on the classic Oxford-style debate, where the winner isn’t the side with the most votes. It’s the side that changes the most minds.
- Read the card as a statement and everyone votes: for, against or undecided. Write down the numbers.
- Toss a coin to pick sides and debate: two minutes each, then one rebuttal per team.
- Vote again. The side that gained the most votes wins. If “for” goes from 4 votes to 9, that’s +5.
- Keep talking. Ask AI, compare its answer with the room’s, and keep the conversation going. The debate is where the conversation starts, not where it ends.
*Note: the better debaters and speakers may not have the more correct answer. Think for yourself and arrive at your best conclusion, then discuss it with others before turning to AI. And remember, this is meant to be open dialogue: educational, and fun.
Built on what scientists know about how AI thinks.
Every card is designed around a documented way that AI systems and people differ, or sometimes agree. These are the main research findings the questions draw on.
Hallucination
Language models are rewarded for guessing rather than saying “I don’t know,” so they can state made-up facts confidently. Kalai et al., 2025, Why Language Models Hallucinate.
Sycophancy
AI assistants tend to agree with the user’s view, even when it is wrong. Sharma et al., 2023, Towards Understanding Sycophancy in Language Models.
Common versus rare patterns
Models do better on familiar tasks and worse on rare ones, such as counting letters or unusual spellings. McCoy et al., PNAS 2024, Embers of Autoregression.
Whose view of the world?
AI answers most resemble people from Western, Educated, Industrialized, Rich and Democratic (WEIRD) societies. Atari et al., 2023, Which Humans?
Moral dilemmas
AI models show stronger biases than people in moral choices, including a pull toward doing nothing. Cheung, Maier & Lieder, PNAS 2025, Large language models amplify human biases in moral decision-making.
Human thinking traps in AI
Classic psychology tests show AI can share human biases like framing and anchoring. Binz & Schulz, PNAS 2023, Using cognitive psychology to understand GPT-3.
Children and smart speakers
Children tend to overestimate what voice assistants can do. University of Edinburgh, 2024.
Trusting machines
People can over-trust automation, or stop trusting an algorithm after a single mistake. Parasuraman & Manzey, 2010, automation bias; Dietvorst et al., 2015, algorithm aversion.
Behavioural economics
Cards on money and choices draw on loss aversion and framing (Kahneman & Tversky, 1979), defaults (Johnson & Goldstein, 2003), the decoy effect (Huber et al., 1982), fairness in the ultimatum game, and sunk cost.
Measurement and Six Sigma
Cards on measuring things use Goodhart’s law (“when a measure becomes a target, it stops being a good measure”), the 5 Whys and root-cause thinking, and the difference between normal variation and real defects.
Specification gaming
AI can meet the letter of a goal while missing its point. Krakovna et al., DeepMind, 2020, Specification gaming: the flip side of AI ingenuity.
AI literacy for every age, at every table.
The cards work for a 6-year-old meeting voice assistants, a grandparent meeting chatbots, and a team meeting AI at work. Ideally at the same table.
Families
Screen-light game nights that turn AI into a conversation, not a mystery.
Classrooms
Discussion starters for digital literacy, ethics, media literacy and critical thinking.
Libraries
Ready-made AI literacy programs for children, teens, families and seniors.
Conferences & team building
A fun, kid-level icebreaker for conferences, offsites and team-building events. No tech knowledge needed.
From an AI education company, founded in 2015.
Humanity Cards and Questions for Humans are created by Shingai Manjengwa, data scientist and founder of Fireside Analytics, a Toronto-based AI and data science education company.
Fireside Analytics courses and programs have reached more than 500,000 learners. The company built Canada’s first Ministry-inspected high school data science course and teaches AI literacy to students, families, educators, public servants and executives.
The card games bring that work to the kitchen table, the classroom and the conference room: plain language, real research, and human judgment first.
Frequently asked questions
What are Humanity Cards and Questions for Humans?
Two conversation card games from Fireside Analytics that build AI literacy. Players think about a question first, debate it, then ask any AI tool and compare the AI’s answer with their own thinking. They are made for families and classrooms, and work just as well at conferences and team-building events.
What is the difference between the two decks?
Humanity Cards are dilemmas about what we should do: rules, fairness, honesty, privacy and care. Questions for Humans are about how we see and know things: describing, counting, estimating, culture and thinking traps, where AI often reads the world differently from people.
What age are the games for?
Ages 6 and up, played with an adult. The cards are written at a kid’s level, with no profanity, so children, grandparents and colleagues can all debate the same question.
Do I need a particular AI tool?
No. Use whatever AI you have at home, such as Siri, Alexa, Google Gemini, ChatGPT, Claude, Microsoft Copilot or DeepSeek. Different tools give different answers, and that difference is part of the lesson.
Is there a right answer on each card?
No. The cards are designed for discussion, not scoring. Sometimes the AI agrees with you, sometimes it does not, and the conversation about why is the learning.
When will the card games be available?
Coming soon. The decks are in family playtesting. Join the waitlist to hear first.
Will there be a French edition?
Yes. A French edition is in development for families, French-language schools and French immersion classrooms.
Can we use the cards for team building, a conference or a company offsite?
Yes. Debate mode turns the decks into a team-building activity: teams toss a coin to debate for or against a card, then compare with what AI says. It works as an icebreaker for conferences, HR and learning events, and offsites. Ask about event orders.
Can schools and libraries order in bulk?
Yes. Fireside Analytics will supply classroom and library sets in batches. Register your interest.
Be the first to play.
Bientôt en français : des jeux de cartes pour développer la littératie en IA et l’esprit critique, pour les familles, les écoles et les bibliothèques.
Fireside Analytics Inc. · Toronto, Canada · Since 2015