🧠 What is a Memory Model?

Models of memory = 🤖 for the mind?

A memory model is like a little machine that:

  • Takes in sequences of inputs
  • Stores representations
  • Produces memory behaviors

🎯 It lets us simulate what minds (and brains) do!

What's the input to a model? 🧩

Usually:

  • A list or sequence of experiences
  • Could be:
    • 📝 Words
    • 🧠 Concepts
    • 🌍 Sensory events
    • 🕰️ Time-varying contexts

What's the output from a model? 🎬

Behaviors we can measure:

  • 🗣️ Free recall, recognition, etc.
  • ⌛ Timing, errors, response curves
  • 🔄 How memory changes with new input

🤔 What kinds of models will we see?

From the course outline:

  • 🧠 Hopfield nets (attractor memory)
  • 🎯 Search & recall processes
  • 🧩 Contextual encoding (e.g. retrieved context model)
  • 🧪 Laplace-transform–based systems
  • 🧬 Biological circuits for memory
  • 🤖 Modern deep nets (LSTMs, Transformers)

📚 We’ll build, analyze, and compare them!

🔍 How do we evaluate models?

  • ❓ Does it fit the data? (qualitatively or quantitatively?)
  • 🧪 Does it predict new behaviors?
  • 🧠 Does it teach us something about cognition or the brain?

🧱 All models are approximations

“All models are wrong, but some are useful.”
— George E. P. Box (1976)

We’re not trying to recreate a brain; we’re building simplified systems to understand memory better.

🎯 Goals when using models

  • Break down complex behavior into understandable pieces
  • Generate testable predictions
  • Build bridges between psychology, neuroscience, and machine learning

🛠️ Memory models are tools — let’s learn how to use them.