Cognitive Modelling ARTICLES Flashcards
Diffusion Model of Decision Making & Value learning through reinforcement learning
What does the diffusion model aim to explain?
Two-choice decision-making by modeling evidence accumulation.
What are the two main criteria in the diffusion model?
Evidence thresholds for decision alternatives.
What determines the time to make a decision in the diffusion model?
The time to accumulate enough evidence to cross a threshold.
What is ‘drift rate’ in the diffusion model?
The rate at which evidence accumulates toward a decision.
How does stimulus difficulty affect drift rate?
More difficult stimuli lead to lower drift rates.
What does between-trial variability in the model account for?
Differences in starting points and drift rates across trials.
What does the term ‘non-decision time’ refer to?
Time spent on processes other than decision-making, such as encoding and response.
How does the model explain the speed-accuracy tradeoff?
By adjusting the distance between decision thresholds.
What is the significance of response time distributions in the model?
They provide a detailed fit to empirical data on decision speed and accuracy.
How do the model’s parameters vary with stimulus discriminability?
Higher discriminability increases drift rate and reduces decision time.
What role does noise play in the diffusion process?
It introduces variability in evidence accumulation.
What is the function of setting different thresholds in the model?
To prioritize speed or accuracy depending on task demands.
What experimental data supports the diffusion model?
Distributions of reaction times and error rates under varying conditions.
How does starting point variability influence decision outcomes?
It biases decisions toward one threshold over the other.
What is a key advantage of the diffusion model over simpler decision models?
Its ability to predict both choice accuracy and detailed response time patterns.
What historical models influenced the development of the diffusion model?
Sequential sampling and random walk models.
How are ‘fast errors’ explained by the model?
They occur when the starting point is closer to the incorrect threshold.
What is a ‘first passage time’ in the context of the model?
The time at which the evidence first reaches a decision boundary.
Why is the diffusion model particularly useful in cognitive neuroscience?
It links decision-making processes to neural mechanisms like firing rates.
What types of tasks are best suited for diffusion model analysis?
Speeded, two-choice decision-making tasks.
What is reinforcement learning?
A trial-and-error method to learn decision-making strategies that maximize rewards.
What are prediction errors in reinforcement learning?
Differences between expected and received rewards.
What role do dopamine neurons play in reinforcement learning?
They signal prediction errors.
What is the Rescorla-Wagner model?
A mathematical model of learning based on prediction error updates.
What is temporal difference learning?
An extension of reinforcement learning that predicts long-term rewards.
How do dopamine neurons respond to unexpected rewards?
With phasic increases in firing rate.
What is the significance of blocking in learning theory?
It shows that previously learned associations can prevent new learning.
How do dopamine neurons react to predicted rewards?
They show no phasic activity.
What is the role of the striatum in reinforcement learning?
It integrates reward signals from dopamine neurons.
What is the main idea behind the Bellman Equation?
It represents value as the sum of immediate reward and future rewards.
What is a major application of reinforcement learning in neuroscience?
Understanding decision-making and reward processing.
How does the prediction error change with increased learning?
It decreases as predictions become more accurate.
What is the learning rate in the Rescorla-Wagner model?
A parameter controlling the size of prediction updates.
How do drugs of abuse affect the dopamine system?
They enhance dopamine activity, reinforcing addictive behaviors.
What is second-order conditioning?
Learning associations between stimuli that predict other reward-predictive cues.
How does variability in dopamine neuron responses reflect learning?
It tracks changes in prediction errors over time.
What is the impact of stochastic rewards on learning?
They create variability in prediction error signals and slower learning curves.
What does a negative prediction error indicate?
That a reward was less than expected or omitted.
How are reinforcement learning theories tested experimentally?
Using tasks that vary reward probability and timing.
What is the relationship between motivation and dopamine?
Dopamine signals influence motivation to perform rewarded actions.
IN-CLASS MINI-QUIZ QUESTION 1:
The linear ballistic accumulator (LBA) and the diffusion decision model (DDM) belong to the same class of decision-making models. Describe a key feature or assumption shared by these 2 models.
IN-CLASS MINI-QUIZ QUESTION 2:
Describe a benefit of combining behavioral and neural data with cognitive modelling, as is done in model-based cognitive neuroscience.