Seminar Series | Graduate Student Colloquiums

Presenter: Tiantong Chen

Advisor: Dr. Allen Yi

Committee Members: Dr. Allen Yi, Dr. Jose Castro and Dr. Weinan Xu 

Presentation title: Design and Fabrication of Diffractive Optical Structures and Different Aspect-ratio Structure Arrays on Glass Surfaces By Compression Molding

Brief description: This presentation will introduce using compression molding process, which is a high volume and cost effective fabrication method to replicate various optical surface structures onto different glass surfaces, including visible and infrared glasses. In addition, experiments on replication of high filling rate and moderate or high aspect-ratio surface structures, which is challenging in compression molding, are employed in this research program. 

Presenter: Dane Morey

Advisor: Dr. Mike Rayo

Committee Members: Dr. Michael Rayo, Dr. David Woods, Dr. Samantha Krening

Title: This isn’t what it looks like, I was framed! Or, how algorithm observability impacts framing effects.

Abstract:

Despite advances in artificial intelligence (AI) and machine learning (ML) that have expanded the range of tasks machines can perform, the inherent brittleness of machines acting alone necessitates coordinating with people, or human-machine teaming (HMT). However, the potential for machines to induce or propagate brittle system performance is not eliminated by the presence of a human supervisor. Machines that make inferences and interject recommendations have a strong potential to influence or frame human decision-makers to interpret a situation in agreement with the machine, which, when misleading, can induce performance that is worse than if the human had been working unaided. It remains unclear what kinds of human-machine architectures can best balance the potential of machines to both improve and degrade team performance. This research investigates the degree to which the presence and absence of both machine observability and explicit machine inferences increases or decreases the risks of automation-induced framing effects in an experiment asking students of at least junior standing in the OSU College of Nursing to interpret the risk of patient decompensation while using a patient vitals display in different configurations. This study will be the first in the reported literature to investigate the degree to which a human-machine teaming architecture can simultaneously convey a machine’s inference and the appropriateness of that inference in a staged-world study with a large sample of domain practitioners, real ML algorithms, real data, and relatively naturalistic tasks.