T18, Awareness, agency and control (Alfonso Rénart) Flashcards
what is a decision?
A choice you make about something after thinking about several possibilities
What are the two branches in in decision theory?
- Normative decision theory
2. Descriptive decision theory
How is uncertainty solved?
- evaluations of the outcomes of each course of action
- History of choices in similar situations
- chance
what is goal-directed behavior?
Instrumental conditioning
- stimulis
- action
- need a model - outcome
What is habitual behaviour=
Instrumental conditioning
- Stimulus
- need extensive experience - action
- outcome
what is pavlovian behavioue?
Classic conditioning
- Stimulus
- several stimulus creates a generalizes reflex
- only a small set of pre-determined actions reflexes - Action
- Outcome
what is the law of effect?
- Thorndike
- explains habitual behavior
- If an action is followed by a reward ut is more likely to happen again
- Computationally simple
—> act automatically in response to a certain stimulus if it has proven to be good in the past - inflexible, slow
—> the output has to change repeatedly to change the association
What is latent learning?
- Tolman
- goal-directed behavior
- what we usually think of as a decision making
-computatonally complex
—> “if I do this, this will happen” - flexible, fast
—> action can change as your model is updated
Name some facts about expected utility theory!
- Decisions are optimal
—> Normative - Places certain constraints on the form of the utility function
- Agens use the laws of probability
Name some facts about prospect theory!
- Decisions not optimal
—> describe how people choose not how they SHOULD choose
—> Descriptive
What is reinforcement learning theory?
Reinforcement Learning Theory shows that in order to learn to predict reward, one needs a teaching signal driven by errors in prediction:
Which Brain areas involved in value encoding/learning, and action selection driven by reward?
1. Prefrontal cortex —> value representation 2. Basal Ganglie —> action selection 3. Dopamine neurons —> reward prediction error
Selection of action?
- Habitual
2. Goal- directed
why is prediction errors useful?
to learn the value of actions
- are encoded by the activity of certain neurons in the brain
How do people typically make choices?
People don’t typically make choices in a normative fashion
what is perceptual decision-making
In perceptual decision making the relevant source of uncertainty Is not in the relative value of the different outcomes, but in the Perceptual identification of the different sensory stimuli
High threshold theory?
CENTRAL IDEA
- Sensory threshold
SIGNALS
signals less than the threshold don’t evoke perception
What is signal-detection theory?
Signal-Detection-Theory provides a framework for studying static sensory decisions.
It is able to tell apart effects due to uncertainty in sensory representations from those due to decision strategies
What is sequential sampling?
Sequential sampling problems deal with making decisions when sensory evidence arrives in time as a stream.
It exposes the speed- accuracy trade-off as a generic feature of these type of choices (although there can be SATs even if evidence is one-shot).
What is drift to bound models?
Drift to bound models are widely used to describe sequential sampling decision problems and they embody naturally the SAT.
They have both a normative interpretation (SPRT) as well as a descriptive one.
What is the ILD-discrimination task?
The ILD-discrimination task allows us to infer the sensory uncertainty
that the rats experience from measurements of their behaviour.
