Memory Representations and Functional Brain Networks Supporting Adaptive Decision-Making Across Tasks and Age

Abstract

Everyday decision making, whether it be as inconsequential as deciding what to buy at a grocery store, to large financial decisions that have lasting consequences, share a common characteristic: they are all supported by adaptive decision making that enable people to draw on past experiences, simulate possible future scenarios, and weigh outcomes that are not immediately available or feasible in the present. Crucially, these functions depend on the use of episodic memory, which rely on the content of what is encoded and represented, how it interacts with different memory systems (e.g. semantic memory), and how different brain networks coordinate and communicate with one another. This relationship is especially important in the context of aging, as semantic knowledge is often organized around how things are categorized. Aging alters the balance between episodic specificity and semantic support, along with the content and format of memory representations and the network dynamics that recruit them. Together, these changes may systematically shape how value-based choice is formed and used across the lifespan.Across the dissertation, I argue that memory supports adaptive, future-oriented behavior depending on what is being encoded, how that content is organized, and how large-scale brain systems coordinate to support decision making and memory representations. This framework links three levels: (1) representational content, where successful episodic encoding can require either high similarity or item-specific distinctiveness depending on semantic class; (2) memory-guided valuation, where episodic simulation can shift intertemporal choices toward the present or the future, and how this may differ across age and (3) network dynamics, where integration of memory network systems differ across the lifespan and may be beneficial for cognitive processing. The first piece of this framework focuses on memory representation. Prior work often treats semantic support as uniform for episodic memory. However, semantic knowledge is structured, and different semantic classes may impose different encoding demands, sometimes favoring representational overlap (similarity) and other times differentiation (distinctiveness). In Chapter 2, I test this idea by examining subsequent memory effects in neural pattern similarity across semantic classes. I show that the relationship between within-class pattern similarity and later memory differs by semantic class: broadly, nonliving object classes show a memory advantage for higher within-class similarity, whereas living classes show that lower within-class similarity can be beneficial in select regions, such as the retrosplenial cortex. These findings suggest that successful encoding depends on a balance between similarity and item-specific distinctiveness, and they support the dissertation’s broader claim that memory’s usefulness depends on the form of representation that is constructed. The second piece examines how memory representations shape value-based choice, particularly when decisions involve balancing between immediate and delayed outcomes. Real-world decisions depend not only on information available in the moment, but also on memory-based simulations of future outcomes and affect that help maintain long-term goals. In Chapter 3, I test how memory-related processes relate to temporal discounting and “choose now” behavior using a temporal discounting paradigm with episodic tagging manipulations. Tagging had a small effect on intertemporal choice in younger adults, but not in middle-aged or older adults. Although there were no significant whole-brain age differences in the tagging effect, older adults showed differences in valuation and salience networks. This chapter asks when and for whom engagement of memory-related systems shifts preferences toward immediate versus delayed rewards, and whether memory engagement supports future-oriented choice or, under some conditions, promotes more impulsive decisions. The third piece investigates how memory-based decision processes are supported by network-level functional connectivity during memory-guided decision making. In Chapter 4, I test how functional brain networks reconfigure as cognition shifts from perception to memory. Using modularity analysis and task-based functional connectivity, I identify memory-relevant networks engaged during multi-attribute consumer choice. I find that older adults show the greatest integration in the global memory network, but the least integration in task-based memory networks; importantly, greater integration in the task-based memory network is associated with faster processing speed in older adults. This approach tests whether age groups differ in the extent to which memory processing depends on integrated network organization and whether individual differences in this organization relate to cognition. Overall, these chapters provide an integrated account of memory-based cognition, linking representational specificity, decision behavior, and network functional connectivity, while showing how aging may shift the systems that support adaptive choice.