Lecture on Mars Computational Theory, Color Vision, and Face Perception

Overview of Today's Lecture

• Mars Computational Theory Level of Analysis
• Case Study: Color Vision
• Intro to Methods in Cognitive Neuroscience
• Case Study: Face Perception

Aim of Cognitive Science

• Central Question: How does the brain give rise to the mind?
• Framework: Mind is considered a set of computations that extract representations.
  • Percepts (e.g., seeing motion, color)
  • Thoughts (e.g., questions, internal dialogue)
• Long-Term Goal: Write code to simulate everything that minds do.

David Marr’s Three Levels of Analysis

1. Computational Theory: What is computed and why?
2. Algorithm and Representation: What are the inputs/outputs and how is it computed?
3. Implementation (Hardware): How is the algorithm physically realized?

Understanding Vision and Color Vision

• Visual Input to Output Pathway: Light hits the retina → Processing → Perception of the scene
• Visual Motion: Information extracted (e.g., presence of motion, person, direction, health)

Ill-Posed Problems

• Many perceptual and cognitive processes are ill-posed (more unknowns than knowns).
• Example in Vision: Infer the actual shape from the image on the retina (inverse optics).
• Example in Language Acquisition: Ambiguity in word meanings (e.g., Gavagai problem from philosophy).

The Case of Color Vision

Importance of Color

• Functional Uses: Identify fruit, objects, people, infer health, detect traffic lights, etc.
• Example: Finding berries easier with color vision.

Computational Challenges

• Reflectance vs. Luminance: Infer object color despite varying light conditions.
• Ill-Posed Nature: Many possible interpretations for the same visual stimulus.

Sources of Additional Information

• Common Sense Reasoning: Understand what color helps with by experiencing perception without color (e.g., demo with grayscale images).
• Algorithm/Representation Level: Write 'code' to solve the vision problem using cues and assumptions.

Psychophysics in Vision Science

• Definition: Study of the connections between stimuli and perceptual responses (e.g., visual illusion demos).
• Example Demo: Identifying car colors under different lighting conditions to illustrate how our brain compensates for the color of illumination.

Functional MRI (fMRI)

Basics of fMRI

• fMRI vs. MRI: Functional MRI measures brain activity by detecting changes in blood flow (BOLD signal).
• Signal Pathway: Increased neural activity → Increased blood flow → Changes in oxygenated vs. deoxygenated blood detected by MRI.

Advantages and Caveats

• Advantages: High spatial resolution, non-invasive method for studying brain activity.
• Limitations: Indirect signal, poor temporal resolution, can only compare conditions.

Case Study: Face Perception

Importance of Face Perception

• Essential for social interaction, identifying individuals, reading emotions, etc.
• Phenomenon of Prosopagnosia: Difficulty or inability to recognize faces.
• Variability in Face Recognition Ability: Normal distribution; some people (super recognizers) are extremely good at it.

Understanding Face Recognition

• Structural Problem: Inputs (images of faces) → Outputs (identification of individuals).
• Challenges: High variability in facial images (templates vs. abstract representations).

Behavioral Studies and Experiments

• Template Matching Hypothesis: Implies difficulty in recognizing new faces without multiple templates.
• Abstract Representation Hypothesis: Suggests using invariant features across images.
• Experimental Evidence: People are less accurate in matching faces without prior familiarity.

Functional MRI Studies in Face Perception

• Objective: Determine if face recognition is distinct from object recognition at the neural level.
• Example Experiment: Differential brain activity in response to faces vs. objects.
• Findings: Consistent activation in specific brain regions (e.g., FFA - Fusiform Face Area).
• Controls: Testing different conditions (e.g., faces vs. hands) to ensure specificity.

Importance of Multiple Levels of Analysis

• Integration Required: Computational theory, psychophysics, and neuroimaging all contribute uniquely.
• Ongoing Research: Further experiments and validation needed to confirm findings.