Abstract

Decision-making is a fundamental part of adaptive behavior and can be compromised in psychiatric illness. Humans and animals are continuously presented with objects and opportunities as options to decide between. One challenge is that these options may independently vary along multiple dimensions or attributes that we care about. This can include attributes of physical reward, such as amount, delay, or uncertainty. When these attributes are present, the possibility of an abstract, cognitive reward, advance information to resolve uncertainty, may also motivate our decisions. It is unclear how the brain makes these subjective decisions, trading off between these multiple attributes based on an individual’s subjective preferences. The theoretical framework of value-based decision-making suggests that neurons integrate preferences for multiple attributes into a scalar quantity, value, that can be compared to make decisions. The lateral habenula, thought to control neuromodulatory systems in the brain, including signaling negatively-signed reward prediction errors to oppose midbrain dopamine neurons, and the dorsal raphe nucleus, the principal forebrain source of the neuromodulator serotonin, are two ancient brain structures that are strongly implicated in decision-making and value processing. However, little is known about the computations signaled by neurons in these areas in the context of multi-attribute decision-making. Furthermore, the lateral habenula and dorsal raphe nucleus are strongly reciprocally connected, but little is known about computations that may be common to them. This thesis work addresses a few key questions. Firstly, we utilize a newly developed multi-attribute decision-making task paradigm to investigate the computational underpinnings of attribute integration into value across humans and monkeys. We show that humans and monkeys integrate attributes following conserved value computations. Then, we investigate whether and how neurons in the lateral habenula and dorsal raphe nucleus signal these value computations in decision-making. Lateral habenula neurons predominantly signal integrated value, including the value of advance information, and manipulating lateral habenula activity with electrical stimulation affects choice in a manner consistent with this integrated value signal causally contributing to online decisions. Dorsal raphe neurons predominantly integrate the value of physical reward attributes. Also, these neurons signal reward predictions errors in response to post-decision feedback. These results shed light on the computations underlying value integration of multiple attributes and identify shared computations in the lateral habenula and dorsal raphe nucleus. They also suggest that value-based decisions may be guided by online neural processes in this circuit. This will contribute to our understanding of how the brain makes decisions and potentially provide insight into the basis of decision-making deficits in psychiatric illness.

Committee Chair

Ilya Monosov

Committee Members

Dan Moran; Deanna Barch; Meaghan Creed; Todd Braver

Degree

Doctor of Philosophy (PhD)

Author's Department

Biomedical Engineering

Author's School

McKelvey School of Engineering

Document Type

Dissertation

Date of Award

4-29-2026

Language

English (en)

Download

Included in

Neurosciences Commons

Share

COinS