1. Introduction

Why is it important or useful to build models of people’s cognitive strategies and representations?

$e^{i \pi} = -1$

(1) By testing our understanding of psychological phenomena in computer simulations, we are forced to be more precise by using the formal language of mathematics. This also helps to identify gaps in explanations. (Where the model fails, etc.)

(2) Computational modeling permits the transfer of insights about human intelligence to the creation of artificial intelligence (AI) and vice versa.

(3) Cognitive modeling of empirical phenomena is a way to infer the psychological mechanisms underlying complex behaviors, which is critical to predicting human behavior in novel situations, or designing interventions when behavior is compromised by disease.

The problem: inferring cognitive mechanisms from limited data is an ill-posed and underconstrained problem because any behavior could be generated by an infinite number of mechanisms. This is why cognitive scientists must have strong inductive biases to infer cognitive mechanisms from limited data.

Theoretical frameworks provide researchers guidance in the search of plausible hypotheses

Cognitive architectures (e.g., ACT-R, connectionism, and computational neuroscience) constrain the modeler’s hypothesis space based on previous findings about the nature, capacities, and limits of the mind’s cognitive architecture.

Assumes that evolution and learning have optimally adapted the human mind to the structure of its environment(Anderson 1990). Supported by empirical findings in perception, statistical learning, and motor control.

Through evolution and learning, the mind is adapted to the search for something beyond ourselves? God?

As psychologists, we are trying to understand a system far more intelligent than anything we have ever created ourselves; it is possible that the ingenious design and sophistication of the mind’s cognitive mechanisms are beyond our creative imagination.

However, the inductive biases of our cognitive models can sometimes be inaccurate. For example, human judgment and decision-making systematically violate the axioms of rational modeling frameworks such as expected utility theory, logic, probability theory. THey also do not specify the underlying cognitive and neural mechanisms that we want to understand.

Cognitive architectures (bottom up) vs rationality (top down) : complementary approaches

Brain represents the world as a trade-off between accuracy and metabolic cost.

2. A brief history of rationality

3. Resource rational analysis

4. Modeling capacity limits to explain cognitive biases: case studies in decision-making

lucy lai

1. Introduction

2. A brief history of rationality

3. Resource rational analysis

4. Modeling capacity limits to explain cognitive biases: case studies in decision-making

4.1 Costly information acquisition and limited attention

4.2 Computational complexity and limited computational resources

4.3 Resource-rational heuristics

4.4 Habits

5. Revisiting classic questions of cognitive psychology

5.1 Reverse-engineering cognitive mechanisms and mental representations

5.2 Cognitive architectures and capacity limits

5.3 Connecting psychology to AI and neuroscience

6. Challenges of resource-rational analysis

7. Conclusion