Author ORCID Identifier

https://orcid.org/0009-0001-9720-7264

Date of Award

Summer 7-31-2026

Document Type

Thesis (Ph.D.)

Department or Program

Computer Science

First Advisor

Elizabeth L. Murnane

Second Advisor

Michael A. Casey

Third Advisor

Nikhil Singh

Abstract

Healthcare communication depends not only on what information is available, but on how it is interpreted, communicated, and acted upon across different roles. AI can support these processes by summarizing, transforming, and organizing information, but it can also change what becomes visible, who is expected to act, and where responsibility resides.

This dissertation examines how human-centered AI can support communication and understanding across healthcare relationships through three independently developed systems. MusicalPT supports self--body communication in physical rehabilitation through adaptive musical feedback. CareInsights supports communication among professional caregivers in long-term dementia care by preserving and surfacing person-specific context across shifts. Family In The Loop supports staff--family communication through shared care information presented in role-specific views.

A retrospective synthesis of these projects identifies a recurring challenge: technical performance alone does not ensure that a system is appropriate to the relationships it affects. This dissertation describes this challenge through \emph{relational alignment}, which concerns whether a system's objectives, design, automation, system properties, and evaluation are appropriate to the roles, responsibilities, capacities, and vulnerabilities of the affected relational configuration.

From this synthesis, the dissertation develops Align, an organizing framework structured around five connected decision areas: communication objectives, design principles, automation and human control, required system properties, and evaluation. Rather than prescribing a standardized development process, Align provides a structure for making these decisions explicit and examining how they relate across clinical, HCI, and AI perspectives. The methods used to investigate and address these areas can vary by project, discipline, and context. A worked example of AI-assisted hospital discharge communication illustrates one way the framework can be applied, while two ongoing projects, Rescripting Therapy and Infant Care, illustrate how its considerations can inform earlier stages of design.

This dissertation makes two primary contributions: three implemented and evaluated healthcare AI systems across distinct relational configurations, and the Align framework for reasoning about the design and evaluation of AI-mediated healthcare communication. Together, they argue that healthcare AI should be assessed not only by what it can compute, but by how it shapes information, authority, responsibility, benefit, and burden across relationships of care.

Download

Share

COinS