Amitai Shenhav
<p class="MsoNormal" >SHENHAV, Amitai, <b>Value Choice Mechanisms</b>, Princeton, Jan 23, 2015</p>
<p>
I am 5 minutes late. This a Stfd FriSem in Jordan Hall Psych in room 050.
Amitai is a post-doc at <b>Princeton</b>. <b>Amitai Shenhav</b>;
(this is a superb and important talk and the room is crammed to the gills: SRO)
His Abstract:
Choices abound in our daily life. Two largely separate bodies of research offer different insight into the process by which we engage with these choices.
Research on goal-directed, value-based decision-making describes how we weigh the costs and benefits of our options, whereas research on cognitive control describes <b>how we exert the necessary cognitive effort</b> to overcome prepotent biases that interfere with the execution our choice (e.g., flankers directing us toward the incorrect response).
I will discuss recent and ongoing studies aimed at examining questions that arise at the intersection of these two fields.
First, I will describe two studies that explore the <b>cognitive effort costs</b>we associate with the act of making a choice: one study examines the neural circuits that drive simultaneously positive and aversive experiences of being offered multiple good options (e.g., <b>great graduate schools to attend</b>); the other study explores the costs of considering alternatives to our default option (i.e., the choice we are prepotently biased to make in a given situation).
Second, I will describe a recent theoretical framework and ongoing modeling work that seeks to address how we weigh the costs and benefits of control itself (i.e., <b>how much cognitive effort is worth exerting</b>?).
Finally, if time allows I will briefly describe an ongoing set of projects aimed at exploring the dynamics that occur when<b> prepotent biases/impulses toward a tempting option </b>or stimulus dimension<b> interact with careful cost-benefit decision-making</b>
(e.g., <b>weighing the healthiness and tastiness of one’s food</b> options).</p>
<p>
he reports initially on <b>Shenhav and Buckner PNAS 2014</b>.</p>
<p>
S = subject given choice of high value vs low value products.</p>
<p>
most anxiety provoking are the dfficult choices, ie low value vs low and high vs high.</p>
<p>
high value product = iPOD and low value = $15 lamp eg.</p>
<p>
<b>conflict = value of outcome * intensity of the mental competition = choice anxiety</b>..</p>
<p>
postive affect = value of best options.</p>
<p>
see <b>festinger in 1957</b> eg.</p>
<p>
he discusses dual valuation circuits...</p>
<p>
see Haber and Knutson in 2010.</p>
<p>
he shows <b>Medial PFC</b></p>
<p>
the <b>dorsal areas of medial PFC are involved in response control and evaluation</b> vs</p>
<p>
<b>ventral regions which are involved with stim evaluation</b> ..
and same layout goes for the <b>striatum</b>..
dACC = dorsal Anterior Cingulate Cortex.</p>
<p>
that is, <b>ventral ACC and vental striaturm are also involved with positive. affect</b></p>
<p>
but <b>dorsal ACC and dorsal striatum are involved with choice anxiety
</b> and response evaluation...</p>
<p>
<b>dorsal ACC... involved in cost of resolving the conflict</b>.</p>
<p>
he seeks to evaluate the <b>role of dACC and conflict during the choice...</b></p>
<p>
<b>the task is stay or switch</b> ... <b>if S given choice of switching products</b>
once he is out of the scanner.</p>
<p>
high high might be <b>digital camera vs video camcorder</b>...</p>
<p>
see <b>shenhav and buckner PNAS 2014</b>.</p>
<p>
he shows that dorsal ACC is activated longer (abt 20 seconds) vs only 10 seconds
in dorsal ACC in those Subjects that decided to switch products.
(ie, they mull it over longer)</p>
<p>
(he shows a slide labeled dACC and conflict after the choice?)</p>
<p>
he also shows areas that are evoked in dACC in whch
there is overlap and some non overlap btwn
choice reversal areas (more posterior in dACC) and choice anxiety (more anterior/rostral)</p>
<p>
<b>part 1: the cost of having multiple good options</b></p>
<p>
a) why the dueling affective responses?
and now b) why (if there is conflict) are we still drawn to these difficult choices.</p>
<p>
is there some implicit decision process that is on-going in ventral ACC.</p>
<p>
but how abt just a pavlovian process that is more automatic.</p>
<p>
as in Grabenhorst and rolls in 2011.</p>
<p>
<b>ie is this value comparison (a voluntary process)</b>
or simply an <b>automatic, pavlivian response</b>.</p>
<p>
ie are these different valuation networks?</p>
<p>
study 2: he puts in a cue to indicate to S whether upcoming trial is low-low,
low-high or high-high.</p>
<p>
he shows that PCC = posterior Cingul. Cortex lights up at time of pre-choice cue.</p>
<p>
low = desk lamp vs water bottle</p>
<p>
low-high might be sodoku game vs iPOD shuffle</p>
<p>
high-high = digital camera vs camcorder.</p>
<p>
actual comparison might be taking place in <b>ventro-medial PFC</b>...</p>
<p>
blue regions on the medial surface include <b>PCC (which is part of DMN network)</b></p>
<p>
one system drives us to want as many options as possible ; the other system
drives the conflict involved in choosing among them.</p>
<p>
Q: how abt choosing among multiple bad choices.</p>
<p>
how abt conflict driven choice costs?</p>
<p>
Q abt opportunity costs...</p>
<p>
he addressed imagined losses...</p>
<p>
(he mentions the fact that we avoid losses more than we seek = gains)</p>
<p>
Herb C: refers to cog dissonance literature... we perseverate...
did I really make the right choice? we continue to mull it over and this
results in vacillation.</p>
<p>
people usually stay more than they switch...</p>
<p>
is it reducing cog dissonance or
is it that people greatly prefer the initial choice they made.</p>
<p>
*</p>
<p>
he continues on the topic of choice conflict and dACC.</p>
<p>
<b>perhaps dACC guides foraging decision</b>... ie do we choose now or
go to a new patch of forest to forage.</p>
<p>
(foraging in the modern era = going to a website other than Amazon to shop, eg)
or... in the political domain: do we examine a new challenger or just
vote for the incumbent politician...
there should be increasing difficulty as value of the challenger increases...</p>
<p>
Kolling... Rushworth in Science: Neural Mechs of Foraging...</p>
<p>
ia v influential paper.</p>
<p>
how valuable are other choices by comparison to the current choice...</p>
<p>
ie we evaluate something better vs current task...</p>
<p>
ie costs vs opportunities...</p>
<p>
(here he continues to try to explain what is going on in dACC)</p>
<p>
foraging task...</p>
<p>
this Kolling Science paper was in 2012.</p>
<p>
another expt of his:
in stage 1: S is offered 5 points vs 8 points</p>
<p>
but also is shown a foraging set:
ie, if instead of either of these you decide to forage: here are the possibilies
but note that foraging costs them one point</p>
<p>
say forage set is 5,8 12, 7, 2, 11</p>
<p>
then at stge 2 they see the probabilities of their actually getting those points...
(eg 50% prob of getting 8 points vs 80% of getting 5 points)</p>
<p>
so, S could forage indefinitely...</p>
<p>
but their bias is to move to forward to stage 2.</p>
<p>
he shows other studies that confirm his finding...</p>
<p>
original study showed</p>
<p>
<b>1) subjects are biased against foraging...</b></p>
<p>
but is the zero point really the best cut point where the cross over of the
decision to accept the choice or forage actually occurs?.</p>
<p>
ex:suppose incumbent and challenger both offer you a tax cut of 6.5%</p>
<p>
that could be the level of max difficulty in deciding between them but say that actually
(because of aversion to foraging)... the challenger might need to to offer a 10% tax cut for you to consider switching.</p>
<p>
is, this is something over and above the -1 cost of foraging...</p>
<p>
he considers a
drift diffusion model... a la Ratcliff and Krajbich in 2010.</p>
<p>
the result is that foraging actually costs more than just -1.</p>
<p>
dACC is better explained by the costs and demands of choosing vs the foraging value.</p>
<p>
best time to switch is during the cross over of value.</p>
<p>
the curse of opportunity...
eg considering a better hairstyle or new career...</p>
<p>
multiple neural mechs of decision making and their competition
under changing risk pressure...</p>
<p>
and see neural mechs of foraging... he shows 3 paper titles...</p>
<p>
</p>
<p>
on to <b>part 3: the value of control</b>..</p>
<p>
dACC is assoc with many processes in the literature: task diffic, errors, adaptation, choice effort, punishment, reward, motivation, action selection, surprise...
so, what is its actual role:</p>
<p>
consider a version of the Stroop task</p>
<p>
he shows a model of pathways involved</p>
<p>
dACC decides how much control is required to override saying RED
and it evokes lat PFC to recruit more effortful choice.</p>
<p>
but... is the demanded control worth exerting....</p>
<p>
the assessment of the payoffs are obv important...</p>
<p>
a normative acct of control evaluation</p>
<p>
he shows giant word GREEN (in red)</p>
<p>
in parallel you eval control costs...and how much control to exert...primary task</p>
<p>
he shows a model that bridges btwn cog control literature and decision and eval literature.</p>
<p>
EVC = expected value of control ...
looks at incr payoff related to incr control...</p>
<p>
ie depends on incentives for exerting control and difficulty of choice.</p>
<p>
general comments on other work...</p>
<p>
interactions btwn goals and temptation: a multi-level approach (ongoing work)</p>
<p>
eg so I eat the healthy food (the apple) or the junkfood, the cake: apple vs cake...</p>
<p>
(short time scale)
he shows an experiment that he did that was a</p>
<p>
stroop type task...
S must alternatively attend to either the MOTION of the random dots or the COLOR of the dots. (with Newsome style random dots) and is given varying rewards for success...</p>
<p>
longer term dynamics in everyday life: do I write a my paper versus checking my email...</p>
<p>
an expt he did:
primary task... type a stream of random numbers but need to skip (say) the twos.
but S id occas. interrupted by temptation to read his facebook news feed.</p>
<p>
collabs: randy buckner for part 1...</p>
<p>
part 2 thru 4...
matt botvinick and jon cohen and mark straccia...</p>
<p>
</p>