Interactive Tools
These interactive demonstrations allow you to explore key computational cognitive models hands-on. Rather than just reading about how these models work, you can manipulate parameters, provide inputs, and observe how the models respond in real time. This kind of active exploration helps build intuition for the computational principles underlying cognition.
Each tool runs entirely in your browser—no installation or programming required. Simply interact with the controls and watch the models come to life.
Conway's Game of Life
The Game of Life is a cellular automaton devised by mathematician John Conway. Despite having extremely simple rules, it produces remarkably complex emergent behavior—patterns that move, replicate, and interact in surprising ways.
What you can explore:
- How simple local rules give rise to complex global patterns
- The concept of emergence in computational systems
- Different initial configurations and their long-term behaviors
- Classic patterns like gliders, oscillators, and still lifes
Relevance to cognitive science: The Game of Life illustrates how complex behavior can emerge from simple rules—a principle central to understanding how cognition might arise from neural computation.
Interactive Activation and Competition (IAC) Model
The IAC model, developed by McClelland and Rumelhart (1981), demonstrates how knowledge can be stored and retrieved in a connectionist network. The demo recreates the famous "Jets and Sharks" network, where information about gang members is encoded in connection weights between units.
What you can explore:
- How activating a person's name retrieves their properties
- How activating properties retrieves people who match
- Graceful degradation and pattern completion
- Competition between similar representations
- The dynamics of spreading activation
Relevance to cognitive science: This model shows how memory retrieval can work through constraint satisfaction rather than explicit lookup—activations spread and settle into states that reflect the structure of stored knowledge.
Neuron Sandbox
The Neuron Sandbox lets you experiment with a simple perceptron—the building block of neural networks. You can adjust weights and thresholds to see how a single neuron computes logical functions, and discover which functions are learnable and which are not.
What you can explore:
- How weights and thresholds determine a neuron's output
- Which logical functions (AND, OR, NOT, NAND) a single neuron can compute
- Why XOR cannot be computed by a single neuron (linear separability)
- The geometric interpretation of neural computation as drawing decision boundaries
Relevance to cognitive science: Understanding single neurons helps build intuition for how neural networks learn representations and why multi-layer networks are needed for complex computations.
Bayesian Number Game
This demo implements Tenenbaum's Bayesian model of concept learning using the "number game." Given examples of numbers that belong to a concept (like 16, 8, 64), the model infers what rule generated them and predicts which other numbers might belong.
What you can explore:
- How the size principle shapes generalization (smaller consistent hypotheses are preferred)
- The difference between size-based and uniform likelihoods
- How adding more examples sharpens the model's inferences
- Why "suspicious coincidences" (e.g., all examples being powers of 2) drive strong conclusions
- The full space of 151 hypotheses the model considers
Relevance to cognitive science: This model captures how humans generalize from sparse data—not too narrowly and not too broadly—through rational statistical inference over structured hypothesis spaces.
Using These Tools in Your Learning
Here are some suggestions for getting the most out of these demonstrations:
Make predictions first. Before changing a parameter, predict what will happen. Then test your prediction.
Explore edge cases. What happens with extreme values? Unusual inputs? Boundary conditions?
Connect to the readings. These tools complement the course readings. Try to relate what you observe to the concepts discussed in lecture and the textbook.
Compare models. Notice how different models embody different assumptions about cognition. The IAC model uses distributed representations; the Bayesian model uses explicit hypothesis spaces.
Think about limitations. What aspects of cognition does each model capture well? What does it miss?