Decision Making

A commitment to proposition or selection of action, resulting in overt act of choosing, based on evidence, prior knowledge, and beliefs.

A final commitment, decision following a deliberation process preceding commitment

A decision we can control and influence

• 1: Task & motivation
• 2: Generate hypothesis 
• 3: Beliefs & prior knowledge 
• 4: Sensory input 
• 5: Useful forms of evidence
• 6: Generate decision variable 
• 7: Apply decision rule
• 8: Motor output

• A stimulus must be discriminated from a distractor. The harder the task, the harder it is to do this
• Through giving these tasks to monkeys we can gain more insight into how decision making works

• Monkeys trained to shift gaze to target in VS task, initially visually responsive neurons in the frontal eye field respond indiscriminately to both targets and distractors.
• Neural activity THEN evolves to signal location of target BEFORE eye movement to target. Neurons in FEF even discriminate target from distractors with NO eye movements, consistent w/ idea of advance decision making
• SO sensory input --> decision process --> motor output

The number of action potentials that fire (their strength)

• Lateral intraparietal Area
• Found within the intraparietal sulcus 
• Controls eye movement, and is possibly associated with working memory
• Activity builds up slowly, and is continued after cessation of a visual stimulus
• Activity predicts the time and direction of the subsequent eye movement, even when the sensory information is ambiguous
• Evidence that LIP neurons are not simply motor neurons reflecting a decision already made. Choice-dependent differences in activity emerge early in response, and the rate of response build-up depends on stimulus strength (ie. strength of evidence)

• Monkeys trained to report net direction of motion in a load of moving dots.
• Proportion of dots moving in same direction (ambiguity) varied by experimenter.
• As fraction of dots moving in random directions INCREASES, proportion of correct responses DECREASES.
• Relationship between behavioural choices and activity of individual neurons in MT visual area, even when motion is weak or absent.
• Suggests perceptual decisions can be based on weakly correlated activity of small population of MT neurons.
• Decision of where to move eyes = sensory input --> LIP --> motor input.

• Not simple sensory neurons: Activity builds slowly and continues after stimulus offset. Activity in them predicts time and direction of eye movement even when sensory info is ambiguous.
• Not simple motor neurons which reflect already made choice:  Choice-dependent differences in activity emerge early, and rate of response build up depends on strength of evidence.

• MT responds to motion, LIP collects evidence from MT and calculates difference in input from neurons in MT.
• As difference grows, evidence accumulates.
• This predicts speed-accuracy trade-off: Reduced threshold, requires less info to make decision (risky and quick). This is biased because it is close to threshold already (risky and quick)
• Still uncertain if decision involves competition between separate reflexes or evolving state of neural networks.

• 1: Is the object moving left or right?
• 2: Left or Right hypothesis generated
• 3: Beliefs & Prior knowledge of movement 
• 4: Neural activity in middle temporal area
• 5: MT preference (left or right)
• 6: Buildup of LIP activity 
• 7: LIP activity exceeds either L or R criterion
• 8: Eyes are moved left or right

• Monkeys fixate on visual stimulus and make decision on which direction targets moved.
• Get large reward for correct answer.
• In 1/2 trials, there was spot monkey could look at to get a small reward if they weren't sure. --> less risky decision, lower pay off.
• Measured CONFIDENCE in choice.
• Activity of neurons in LIP predicted whether monkey would take risky/safe choice.
• Rats had to smell odour and decide which of 2 other odours were most similar. Activity recorded from orbitofrontal cortex (goal-directed decision making) matched prediction of confidence, with rats sure on correct trials and unsure on error trials.