The Neural Representation of DecisionMaking under Uncertainty Scott Huettel Psychology & Neuroscience Duke University @ Decision Neuroscience Scott Huettel, Duke University “…there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns -- the ones we don't know we don't know.” - Rumsfeld (2003, in press) @ Decision Neuroscience Scott Huettel, Duke University “To preserve the distinction… between the measurable uncertainty and an unmeasurable one we may use the term ‘risk’ to designate the former and the term ‘uncertainty’ for the latter. … It is this [type of] uncertainty which has been neglected in economic theory, and which we propose to put in its rightful place.” F. H. Knight (1921) Risk, Uncertainty, & Profit @ Decision Neuroscience Scott Huettel, Duke University Outline of the Talk • Uncertainty defined • Case study: Ambiguity – Contributions of lateral prefrontal, parietal cortex? – Contributions of orbitofrontal cortex, amygdala? • Discussion: What should neuroscience seek? @ Decision Neuroscience Scott Huettel, Duke University “Information occurs only if there exists some a priori uncertainty, and the amount of information is determined by the amount of the uncertainty – or, more exactly, it is determined by the amount by which the uncertainty has been reduced.” Wendell “Tex” Garner @ Decision Neuroscience - Garner (1962). Uncertainty and Structure as Psychological Concepts, p. 3. Scott Huettel, Duke University @ Decision Neuroscience Scott Huettel, Duke University Lateral PFC Sequence Length LPFC BOLD Signal Change 0.100% 7 6 5 4 3 2 1 0.075% 0.050% 0.025% 0.000% 0 4 8 12 16 -0.025% LPFC Huettel, Mack, & McCarthy (2002) Nature Neuroscience @ Decision Neuroscience Scott Huettel, Duke University Ubiquitous Risk Signals Insula Posterior Cingulate Preuschoff et al. (2007) J Neurosci Parietal Cortex McCoy & Platt (2005) Nat Neuro Prefrontal Cortex Huettel et al. (2005) J Neurosci @ Decision Neuroscience Scott Huettel, Duke University From Glimcher & Rustichini (2004) Science. @ Decision Neuroscience Scott Huettel, Duke University Risk vs. Ambiguity Trial Types Risky - Certain Risky - Risky Ambiguous - Certain Ambiguous - Risky Trial Structure Decision (RT) Expectation (4.5-6s) Outcome (2s) Huettel et al. (2006) Neuron @ Decision Neuroscience Scott Huettel, Duke University L 20 BOLD Parameters (a.u.) aINS Signal change (%) LPFC LPFC 0.12 AC AR RC RR AC AR RC RR AC AR RC RR pPAR aINS 0.16 20 pPAR 0.2 Time since trial onset (s) @ Decision Neuroscience 20 Trial Type Scott Huettel, Duke University Risk: Expected Utility Ambiguity: α – MaxMin Expected Utility Ambiguity Averse Risk Averse Ambiguity Averse Risk Preferring Ambiguity Preferring Risk Averse Ambiguity Preferring Risk Preferring Subjects (n=13) @ Decision Neuroscience Scott Huettel, Duke University Ambiguity… in Lateral Prefrontal Cortex? (0) Less More (3) Correlation with Economic Preference More (1) (RC + RR) - (AC + AR) (AC + AR) – (RC + RR) (0) Less Risk Preference aINS LPFC Ambiguity Preference 1.0 α - Ambiguity 0.8 0.6 0.4 0.2 0.0 LPFC pIFS pPAR β - Risk pPAR Resampling Analysis Ambiguity preference (1-α) @ Decision Neuroscience Risk preference (β) Scott Huettel, Duke University Ambiguity… in Orbitofrontal Cortex? Hsu et al. (2005) Science @ Decision Neuroscience Scott Huettel, Duke University Evidence from Cognitive Neuroscience • Lateral prefrontal cortex – Important for establishing rules for behavior – Implicated in reasoning, response selection • Orbitofrontal cortex – Important for learning about (aversive) stimuli – Implicated in inhibition of behavior @ Decision Neuroscience Scott Huettel, Duke University Interim Take-Home Message Concepts from decision science are unlikely to reflect unitary psychological constructs nor single neural modules @ Decision Neuroscience Scott Huettel, Duke University Conditioning of Risk and Ambiguity Here, ambiguity reflects the expected revelation of information; i.e., a potentially known unknown. Bach et al. (2009) J Neurosci @ Decision Neuroscience Scott Huettel, Duke University Stimuli involving ambiguity evoked greater activation in dlPFC and PPC than those involving risk or ignorance. Bach et al. (2009) J Neurosci @ Decision Neuroscience Scott Huettel, Duke University In a task similar to an implicit Wisconsin Card Sorting Task (i.e., learning rules without immediate feedback)… pIFS …stimuli that eliminate potential rules evokes activation in lateral PFC (the posterior inferior frontal sulcus)… …proportional to the # eliminated rules. Huettel & Misiurek (2004) @ Decision Neuroscience Scott Huettel, Duke University Ventral PFC Damage Increases Risk Seeking Clark et al. (2008) Brain Cf. Goel et al. (2007) Cerebral Cortex, who argue for potential laterality effects in reasoning: RH impairs under incomplete information @ Decision Neuroscience Scott Huettel, Duke University Reward / Value Kringelbach (2005) Nat. Rev. Neurosci Hsu et al. (2005) Science Plassmann et al. (2007) J. Neurosci. @ Decision Neuroscience Scott Huettel, Duke University • Ambiguity effects in orbital PFC – Tasks (3): Gambles, knowledge, opponent Emotional aversion signal:preceded pushes behavior – Activation decision (slow) away from–ambiguous/risky options?activation Aversion led to increased – Subject sample: very ambiguity averse Hsu et al. (2005) Science • Ambiguity effects in lateral PFC – Task (1): Gambles Cognitive signal: supports creation of decision – Activation coincident with decision (fast) scenario under ambiguity? – Aversion led to decreased activation – Subject sample: ambiguity neutral LPFC Huettel et al. (2006) Neuron @ Decision Neuroscience Scott Huettel, Duke University A Lesson from Psychology… for Neuroeconomics? “A concept that is synonymous with a single operation is nothing more than a restatement of an experimental result. But a concept that arises as a consequence of converging operations has a reality that is independent of any single experimental observation. Wendell “Tex” Garner “But before we can get convergence, we must introduce variation in our experimental procedures. … Stabilizing on a few techniques… [is] utterly self-defeating… because it completely drops the critical part from critical realism.” The Processing of Information and Structure p. 187 @ Decision Neuroscience Scott Huettel, Duke University Acknowledgments Collaborators on these Projects • Gregory McCarthy • Michael Platt Laboratory Members • McKell Carter • Chris Coutlee • John Clithero • Debra Henninger • O. Mullette-Gillman • Brandi Newell • Allison Scott • David Smith • Adrienne Taren • Vinod Venkatraman • Richard Yaxley @ Decision Neuroscience neuroeconomics.duke.edu Support • NIMH, NINDS, NIA • Duke Institute for Brain Sciences Scott Huettel, Duke University