Decision-making Flashcards
(28 cards)
Decision-making definition
The process of evaluating options to arrive at a specific choice or course of action.
7 elements of perceptual decision making
- Pre-existing knowledge.
- Motivation.
- Goals.
- Sensory encoding.
- Decision-variables.
- Decision-rules.
- Consequence evaluation.
Clinician’s dilemma
There is no optimal place for the response criterion (minimise misses vs. minimise false alarms).
What does d’ depend on?
- The separation.
- The spread.
Decision-variable in SDT
Often the ratio of likelihoods: l(e) = P(e|h1)/p(e|h2).
Decision-variable in sequential analysis
Sum of logarithms of likelihood ratios associated with each piece of evidence.
2 neural correlates of perceptual decision-making
- Sensory evidence: primary somatosensory neurons.
- Comparison between f1 and f2: ventral premotor neurons.
Race model neurons
Ventral intraparietal neurons. Firing seems to increase until a threshold is reached and a decision is made.
Expected value neurons (non-PFC)
Lateral intraparietal (LIP) neurons in macaques encodes both magnitude and probability. When magnitude varied, LIP neuron firing correlated with magnitude. When probability varied, LIP firing correlated with probability.
Expected value neurons (PFC)
Kennerley et al. (2009) found probability, payoff, and cost were encoded in macaque ACC, OFC, and LPFC neurons. ACC neurons encoded multiplexed representations of the three different decision variables.
Subjective value encoding
Macaque OFC neurons encode subjective value of options via an absolute value scale. OFC neurons show a U-shaped value of activation (firing rate correlates with subjective value of the chosen option).
Padoa-Schioppa and Asaad (2006) 3 types of OFC neurons
- Chosen value. Encodes value of chosen option (U-shape).
- Offer value. Encodes value of only one of the options.
- Taste. Encodes which option is being chosen, rather than the value.
Padoa-Schioppa and Assad (2008) OFC three juices
The values of the juices appeared to be encoded on an absolute value scale using a common currency.
3 impairments after OFC lesions
- Instrumental extinction. (The value of the action is changed as reinforcer is removed. OFC animals still persist with task).
- Object reversal learning. (The value of the options reversed. OFC animals still persistently choose previously rewarded object).
- Reinforcer devaluation. (Do not choose the non-devalued reward despite being fed to satiety).
OFC for adapting behaviour (otherwise show rigidity and slow to adapt to new reward contingency) and keeping track of changing value of rewards.
Rudebeck et al. (2006) cost-benefit decision-making rats T-maze
OFC: how long rats waited. ACC: how much effort rats put in.
Ellsberg paradox
Most people prefer the risky deck (certain that there is a 50/50 split) over the ambiguous deck. It suggests that you don’t wish to bet on red from the ambiguous deck but also don’t with to bet on blue.
Hsu et al. (2005) participants picked a colour from ambiguous or risky deck
- Amygdala and OFC: greater activity for ambiguity than risk (mostly prior to decision). Integrate incomplete information?
- Striatum: greater activity for risk than ambiguity (mostly after decision). Correlated with EV - anticipate potential rewards?
O’Neill and Schultz (2010) value and risk - macaques could choose between non-risky or risky option
Macaque prefer risky options (as risk increases, tendency to choose risky option increases). OFC neurons encodes the degree of risk.
De Martino et al. (2006) framing effect
Framing effect influenced people’s propensity to gamble. Participants least susceptible to framing effect (more rational) had heightened OFC, vmFC, and amygdala activity.
Thorndike’s spread of effect
Rewards affect not only the connection that produced them, but temporally adjacent connections too.
Walton et al. (2010) credit assignment
LOFC lesions impair credit assignment (associating outcomes with correct choice). The closer A was rewarded to B, the more likely monkey is to choose B again even though it was A that led to the reward.
Impairment not seen in controls or mOFC lesions.
Noonan et al. (2010) value comparison
mOFC lesions impaired macaque’s value comparison. Choices were impaired when close value comparison was required (V2 was close to best option V1).
Kolling et al. (2012) exploratory decision-making human fMRI
vmPFC/mOFC: more active during exploitation.
ACC: more active during foraging. Encoded search value.
Daw et al. (2006) four-armed bandit task human fMRI
Frontopolar activity: correlated with exploratory behaviour (testing the other options).
mOFC activity: correlated with pay-out.
vmPFC activity: correlated with probability of chosen reward.
Participants sampled other options to confirm which machine currently had the highest payoff.
