Understanding human perception to understand machine perception.

Topical subject: Mythos! Download “Mythos and the Perception Game” here

The Perception Game and Mathematics. Download here.

The Perception Model is an interactive exercise that uses simple picture cards to demonstrate a powerful idea: humans and AI both make sense of the world based on the information they have been exposed to.

Game Overview

The Perception Model is a simple but powerful game that demonstrates how humans – and AI systems – interpret the world based on the information available to them. By using picture cards, the game reveals how perception, pattern-making and meaning-making differ between individuals and between humans and machines.

What the Game Demonstrates

Every person interprets an image through the lens of their lived experience, education and belief system. A child, with limited exposure to the world, will naturally make simpler, more literal connections. Adults, with richer life experience, will see deeper patterns.

AI systems behave in exactly the same way.

An AI model learns only from the data it is trained on. If the dataset is narrow or incomplete, the model’s interpretations will also be narrow, simplistic, or incorrect — much like a young child encountering something for the first time.

For example, if an AI model has never seen a house, it may interpret the first house image as:

• a triangular prism on a cube
• a box with flaps
• or any other shape it has seen before

This is how biases, misinterpretation and hallucinations arise.

As training data becomes richer – text, images, audio, video – AI models begin to learn more like humans do: by absorbing patterns from the world around them. But even then, they can still produce confident but incorrect answers, because they do not “understand” — they only predict.

Why This Game Matters

When your team plays this game, you:

• Reveal how each person’s mind works
• Understand how assumptions shape interpretation
• See how different people make different connections from the same input
• Experience first-hand how AI models generate outputs from patterns
• Understand why AI can hallucinate
• Appreciate why model explainability is essential in responsible AI
• Build social capital in your team

This game mirrors how foundation models and generative AI work: they learn from existing content to produce new content — sometimes brilliantly, sometimes incorrectly.

The Benefit of Playing

• Builds shared understanding
• Strengthens communication
• Surfaces hidden assumptions
• Demonstrates the limits of both human and machine perception
• Shows why diverse teams — and diverse datasets — matter
• Helps teams grasp AI risks in a safe, engaging way

It is a practical, memorable introduction to how AI learns, why it sometimes fails, and why human oversight remains essential.

Playing The Perception Model is important because it helps teams understand how differently people think, interpret information, and make decisions. When colleagues talk through what they see – and why they see it – they reveal their assumptions, lived experiences, and mental shortcuts. This builds psychological safety, strengthens communication, and helps teams appreciate the diversity of thought that drives good decision-making. It also mirrors how AI models learn: different inputs create different outputs. By experiencing this together, teams learn how to talk openly, challenge respectfully and understand each other’s perspectives – essential skills for any organisation working with AI or navigating complex, fast-moving environments.

To play, contact hello@SOFIAA.com