Topic 10: Decision Making Flashcards
What is decision making?
some behaviors can be considered deterministic (e.g. a reflex)
many cognitive behaviors can be thought of as involving non-deterministic decisions
What are the three broad kinds of decisions?
perceptual decisions
value-based decisions
foraging decisions
What were the methods of the Newsome et al. (1989) study on perceptual-based judgments?
tested global motion perception using a random dot motion (RDM) paradigm
monkeys trained to make a judgment about whether the direction of motion embodied in the stimuli presented was consistent with the direction the onscreen arrow was pointing by making an eye-movement towards one of two targets (“pref LED” or “null LED”)
coherence of motion of dots varied (e.g., 0%, 50%, etc.)
What were the results of the Newsome et al. (1989) study on perceptual-based judgments?
activity of neurons in the MT (middle temporal) region were found to be closely associated with the overall coherence of motion (e.g. more motion –> more activity, certain neurons turned to respond to motion in one particular direction, etc.)
they also applied microstimulation to artificially stimulate those neurons in an attempt to test whether they make causal contributions to the judgments being made
found that activating those MT neurons changed the judgments being made; but did microstimulation affect what the monkey was actually perceiving, or just what it was deciding? how would we know?
Can we treat eye movements as a way to operationalize a decision?
we understand there is a close link between eye-movements and action
can we treat eye-movements as a way to operationalize a decision?
eye movements are controlled by the frontal eye fields (FEF) and superior colliculus (SC)
the lateral intraparietal cortex (LIP) links the MT (middle temporal) region with the FEF and SC
What was the Salzman et al. (1990) study on the lateral intraparietal cortex?
used a version of the random dot motion (RDM) paradigm in which eye movements to the right or left target were made on the basis of the direction of movement
logic for cirtical comparisons (or “catch trials”): on stimuli displays that have a coherence of 0, an eye-movement must be made to either the target on the right or left of the screen (because it is a forced alternative choice, i.e. no response option for “no coherent motion”)
critically, the perceptual component in either case is the same but the decision being made of course differs
What is a motor field?
the concept of a motor field is similar to that of a receptive field, e.g. a neuron will become activated when making a saccade towards their motor field, i.e. a specific location in the visual field
a given neuron that is involved with initiating eye movements towards the upper right-hand corner of the monitor may start firing when that kind of saccade (eye movement) is starting to be planned (and/or executed)
a different neuron (that is also involved in controlling eye movements) may only start firing when an eye movement towards the upper left-hand corner of the monitor is planned/executed
What are catch trials?
we can measure the firing pattern of a give LIP neuron that has a motor field which overlaps with one of the two targets that an eye moment is made towards to input the monkey’s response (“left” or “right”, to indicate the direction of movement is judged to be to the left or right)
if LIP neurons show different patterns of activation on “catch trials” in which different responses were provided, that can’t be explained on the basis of perceptual processing alone (because the perceptual information on all of the catch trials is the same, i.e. 0% coherence)
therefore, any such differences should reflect contributions to the decision making process above and beyond those related to basic sensory processing
What did LIP neurons show in the Salzman et al. (1990) study?
a gradual increase in firing rate before a saccadic movement towards a target (for catch trials in which a saccade is made towards their motor field)
a suppressed level of activity until after the saccade has been completed (for catch trials in which a saccade is made away from their motor field)
the results on these catch trials suggest LIP neurons are somehow involved in the decision making and/or motor planning process(es), rather than a role that is limited to perceptual processing
a graded nature of this effect was also observed for trials with coherence >0%: greater coherence results in a steeper increase in the rate of firing
incentivizing faster responses (with larger rewards) when the coherence was higher also produced faster responses, which also correlated with changes in LIP activity
What are the LIP contribution in judging motion?
the way LIP neurons seem to be functioning could match a description in which the gradual accumulation of information within a system is used to guide a decision
we can relate this to the idea of a buildup of a familiarity signal that, in conjunction with recollection-based processes, can guide memory-based decisions
What was the method of the Donner et al. (2009) study on a RMD paradigm?
tested a RMD paradigm with humans that used a manual (rather than saccadic) response to indicate whether or not coherent motion was detected
they used MEG, and relied on the fact that two characteristic electrophysiological changes are known to be associated with limb movement: suppressed activity in the beta bad (12-36 Hz), and enhanced activity in the gamma band (64-100 Hz), both contralateral to the movement
by monitoring for changes in the power of these bands, attempted to infer when a movement was about to be initiated
What were the results of the Donner et al. (2009) study on a RMD paradigm?
found evidence for the expected changes i primary and premotor cortex (M1, PMd) 1.5 seconds prior to a button press
suggests some decision making mechanism is acting at that time/place
but again, could that simply be i response to a signal “farther upstream” that is ultimately more “responsible” for making the decision
What is a common currency?
comparisons between different kinds of options rely on this abstract measure of subjective value, a kind of “common currency” for choice
what would represent this “common currency” in neurological terms?
the term neuroeconomics is sometimes applied in these contexts
How can we compare perceptual to value-based judgments?
perceptual judgments are (at least supposed to be) objective
value-based judgments are inherently subjective
What were the methods of the Platt and Glimcher (1999) study on neuronal activity and expected value?
trained monkeys to fixate on the centre of the screen while target stimuli are presented in the periphery
cue indicates the monkeys are supposed to make an eye movement to one of two targets: one produced 0.15 mL of juice, the other 0.05 mL (3-1 ratio)
recorded activity of neurons in LIP
What were the results of the Platt and Glimcher (1999) study on neuronal activity and expected value?
as expected, monkeys tended to favor the high-reward target
however, they also sometimes chose the low-reward target (in line with the “matching law” i.e., they chose the high reward target 3 times to 1)
this allowed the researchers to compare LIP neuronal activity just prior to selecting lower and higher reward targets
they found that the activity in LIP neurons was closely related to the expected value (e.g. more robust firing for higher expected values)
What is the role of the orbito-medial PFC (omPFC) in common currency?
lesions in the omPFC impair decision making in various domains
the omPFC receives input from all modalities and limbic areas
neurons in the omPFC have also been shown to respond to flavors in a way that in modulated by satiety
What was the Plassmann et al. (2008) study on OFC activity?
OFC modulates it’s response in line with expectations
wine that is presented as more expensive more strongly activates the OFC
pleasantness rating also affected… did the elevated OFC activity “cause” the higher ratings, vice versa, or something else altogether? similar to the question of whether faster responses correlated with greater (LIP) activity are causally related
What is value transitivity?
the value of one item can be expressed in terms of a quantity of any other (e.g. if you really like beer A, and only sort of like beer B, you might happily trade two pints of beer B for only one pint of beer A)
a major assumption about how we go about making value-based judgments is that we have some way of encoding value in a way that demonstrates this property (value transitivity)… so what could be the “common currency” (i.e., neurologically relevant mechanism) that encodes that value?
What was the method of the Padoa-Schioppa and Assad (2008) monkey study on forced choice?
recorded neural activity in the OFC when monkeys were presented with a series of two-alternative forced-choices (2AFC), in which one of two possible rewards was chosen
rewards were always different types of juice, offered in various quantities (e.g., 1 “shot” of juice A vs. 3 “shots” of juice B)
What were the results of the Padoa-Schioppa and Assad (2008) monkey study on forced choice?
they found the monkeys seemed to have relatively stable preferences, and reliably accepted certain kinds of offers (e.g. less of their preferred juice instead of more of a lesser preferred juice) while rejecting others
these presumably reflected the differing values ascribed to each juice
What were the three different kinds of neurons in the OFC which responded to different properties in the Padoa-Schioppa and Assad (2008) monkey study?
one type of neuron was interpreted as representing the value of a specific juice
another type of neuron was interpreted as representing the value of the other juice
another type of neuron was interpreted as representing the value of the choice, that is, the difference in value between the two options
What was found regarding judging value in monkey studies?
Deaner et al. (2005) also observed relatively stable estimates associated with the value of various rewards, including juice and the opportunity to view images of other monkeys
as with Padoa-Schioppa and Assad (2008), this demonstrates transitivity and supports a particular view of how these judgments are made
also various lines of behavioral evidence for similar effects of transitivity for humans (e.g. “paying” a virtual currency to view images of attractive individuals)
What is the confirmation bias in neuroeconomics?
the general models we’ve been talking about could be argued to have been heavily influenced by economic theory, much of which may not be (strictly speaking) “accurate” (perhaps in general, though also maybe just as these things apply to individuals rather than groups, etc.)
does that bias the way we look for evidence and/or try to build arguments about how humans make decisions?
think back to our discussion about localization of function potentially being an oversimplification, which may also encourage confirmation bias
What are confounds related to experimental design in neuroeconomics?
could the dominant characterization of what’s happening in the brain during decision making reflect a confound related to the way these experiments are typically designed (i.e., only ever choosing between two options with a 2AFC, which vary on singular dimension, or factor)?
what about situations in which there is more complexity? e.g. when there are more than two options?, more than one relevant dimension?
What are the three dimensions of foraging decisions?
encounter value
search value
search cost
What is encounter value?
the value of exploiting whatever option/resource one has currently available to them
What is search value?
the potential value in exploring and stumbling upon a better option/resource than what is currently available
What is search cost?
the energetic cost (e.g. calories) of searching for more options
What is the role of the ventromedial prefrontal cortex (vmPFC) in foraging decisions?
vmPFC and/or OFC activation is typically interpreted as corresponding to perceived value, which is what most of the work we’ve discussed here emphasizes
in the context of foraging, however, those three critical dimensions (encounter value, search value, and search cost) don’t seem to be embodied in what’s happening within the vmPFC/OFC
What is the role of the dorsomedial frontal cortex (dmFC) in difficult decisions?
can often be predicted on the basis of how “difficult” a decision is (in various contexts, not just value learning)
although dmFC activation is often discussed as reflecting processing related to error, difficulty, or conflict “monitoring” in other contexts, it has also been proposed to represent (at least at times) the relative benefit of exploring in foraging situations
