Schedule of Events | Symposia

Symposium Session 6 - Uncertainty Resolution across Learning, Memory, and Decision-making

Symposium Session 6: Monday, March 31, 2025, 10:00 am – 12:00 pm EDT, Independence Ballroom

Chair: Vishnu Murty¹; ¹University of Oregon
Presenters: Aaron M Bornstein, Ian C Ballard, Chelsea Helion, Vishnu P Murty

The ability to adaptively navigate the world is challenged by individual’s having incomplete models of how the world works. This type of uncertainty is pervasive across a wide range of behaviors, deciding where to invest our money, who to trust, or what strategy to use to win a board game. Thus, there is a constant need for individuals to resolve uncertainty, making it a growing topic of interest across multiple domains of cognitive neuroscience. In this symposium, we will bring together researchers from decision-making, episodic memory, and social/affective science to understand mechanisms of uncertainty resolution and its consequences on adaptive behavior. Our first talk from Aaron Bornstein explores how individuals dynamically shift their foraging behavior across the lifespan as a function of environmental uncertainty. Our second talk from Ian Ballard synthesizes data across humans and non-human primates to arbitrate whether the striatum’s role in uncertainty resolution reflects reward feedback or goal updating. Our third talk from Chelsea Helion explores uncertainty resolution in the context of social behavior and how it is modulated by internal (affect) and external (relational schemas) factors. Our final talk explores how resolving uncertainty in real time has downstream influences on long-term memory. With this symposium, we will showcase a variety of behavioral orientations and techniques that converge on answering a foundational question of how individuals resolve uncertainty in service of adaptive behavior.

Presentations

Sequential decisions adapt to uncertainty within an environment and across the lifespan.

Aaron M Bornstein¹; ¹University of California: Irvine

Humans and animals are often maligned as being bad ("suboptimal") at making decisions, especially decisions under uncertainty. But is this allegation justified? In this talk, I will present recent findings in the domain of patch foraging. Foraging requires individuals to compare a local option to the distribution of alternatives across the environment. Foragers, across a range of species, have been observed to systematically deviate from exogenous notions of optimality by “overharvesting”—staying too long in a patch. We developed a new computational model that explains the appearance of overharvesting as a by-product of two mechanisms: 1) statistically rational learning about the distribution of alternatives and 2) decisions that adapt to the uncertainty of these distributions–in particular, by using an estimate of local uncertainty to adjust how much to weight potential future rewards. Across four experiments (total N=525), we test this model using a variant of a serial stay-leave task and find that human foragers' behavior is consistent with both mechanisms. Our findings suggest that overharvesting, rather than reflecting a deviation from optimal decision-making, is instead a consequence of optimal learning, and rational adaptation to uncertainty about our learning. Importantly, this finding is replicated in distinct populations, each with experiential and biological factors that should alter their estimates of uncertainty. We show that individuals’ decisions rationally adapt to the predictability of their environment, the breadth of their experiences, and the capacity of their neural circuitry to represent and transmit information with high fidelity.

Cortico-striatal circuits for goal updating

Ian C Ballard¹; ¹University of California: Riverside

Reward feedback leads to learning but can also signal the need to update behavior. Learning and behavioral updating are interrelated but could rely on distinct neural circuits. However, a challenge for disentangling these constructs has been that reward prediction error (RPE), a learning signal, is perfectly confounded with behavioral updating in many tasks. As a result, neural signals that correlate with RPE could reflect behavioral updating. We designed a task to disentangle whether feedback signals in the striatum reflect RPE or behavioral updating. We accomplish this by varying the amount of information conveyed by negative feedback in a 2-arm bandit task so that behavioral updating is linked with positive and RPEs on different trials. We found that the BOLD response in the dorsal striatum reflected goal updating but not RPE. Specifically, the dorsal striatum responded more strongly to losing than winning money when losing money led to behavioral updating. Additionally, dorsal striatal feedback responses strongly predicted upcoming behavioral change. In the ventral striatum, we found a mixture of RPE and updating responses, suggesting that the ventral striatum translates learning into behavioral change. We found a markedly similar pattern in neurophysiological recordings from nonhuman primates: dorsal striatal neurons signaled upcoming behavioral change, whereas separate ventral striatal neurons tracked RPE and behavioral updating. Our results suggest that a large portion of the striatal feedback response reflects behavioral change rather than RPE and, therefore, can be used to probe altered behavioral updating in neurological and psychiatric diseases.

Identifying between- and within-person factors that moderate information processing in social uncertainty

Chelsea Helion¹; ¹Temple University

Navigating the social world requires processing information under uncertainty (e.g., Is that person trustworthy? Did she mean to hurt my feelings?). Across two studies, we examined two factors that influence information processing in the context of social uncertainty: 1) relational schemas, and 2) individual differences in sensitivity to uncertainty. To do so, we leveraged multimodal narrative stimuli (i.e., film clips) in which an individual had been accused of a crime. Participants were instructed to watch the film(s) while continuously rating their certainty that the accused individual was guilty. In the first study, participants (n = 43), adopted a relevant relational perspective (i.e., friend of the accused or victim, a detective). We found that perspective moderated information processing, such that individuals in the friend of the victim condition showed greater decision impulsivity than the other conditions. Using a convolutional neural network (EmoNet), we found that scenes associated with higher levels of emotional entropy (i.e., uncertainty in the emotional category elicited by the visual stimuli) were also associated with greater schema reliance. In our second study, using intersubject representational similarity analysis, we found that individuals higher in intolerance of uncertainty showed increased time-course similarity in circuitry associated with somatosensory integration (i.e., insula) and that individuals higher in state anxiety showed increased time course similarity in regions associated with social cognition (i.e., TPJ, ATL). Taken together, these studies identify two potentially important contributors to processing social uncertainty, which may inform complex social judgment and decision-making.

Uncertainty resolution during hypothesis testing dynamically alters episodic memory

Vishnu P Murty¹; ¹University of Oregon

Uncertainty signals that the world around us has changed, and individuals need to update their mental maps, a process facilitated by hippocampal-dependent memories. The relationship between uncertainty and memory is not a one-to-one relationship (i.e., more uncertainty, better memory). Rather, research in meta-cognition during re-study decision show that motivation to learn is greatest during moderate levels of uncertainty. However, research has yet to explore how dynamic changes in uncertainty resolution influences memory when individuals are building and modifying priors during goal-directed search. In this talk, we will highlight two studies that utilize a hypothesis testing paradigm, in which participants figure out rules to unlock treasure chests to reveal trial-unique memoranda. In Study 1, we show an inverted U-shape relationship between decision uncertainty and memory, such that memory was best at moderate levels of uncertainty during hypothesis testing. In Study 2, we replicate findings the associations between decision uncertainty and memory and show that uncertainty-related memory enhancements are related to an integration of (1) markers of encoding success in the hippocampus and (2) markers of uncertainty in the striatum and ventromedial prefrontal cortex. Together, these findings provide greater insight into how hypothesis testing primes mesolimbic-hippocampal interactions in service of storing features of the environment related to uncertainty resolution.