Files and links (1)
thesis_revised_july_31_ronilo_r16.33 MB
Embargoed Access, Embargo ends: 08/28/2027
Dissertation
Extension, critique, and formalization of explanation in machine learning
University of Iowa
Doctor of Philosophy (PhD), University of Iowa
Summer 2025
DOI: 10.25820/etd.008135
Abstract
Oftentimes, it is not sufficient for a machine learning model to achieve good predictive performance; in cases where a model is used in high-stakes decision-making, it is preferable to use an interpretable model. Examples of such applications include medical diagnosis, automated allocation of healthcare benefits, and legal sentencing. In these contexts, the study of explainable machine learning (XML) models is important to our society. This thesis contributes to the study of XML in two main ways. The first part of this thesis focuses on increasing the problem-solving capacity of XML. The problems of constructing intrinsically interpretable deep neural network (DNN) architectures for reinforcement learning (RL) agents solving MDPs with visual or spatio-temporal state-spaces are addressed. Both solutions extend the idea of prototype-based deep neural networks (DNNs), which use case-based reasoning for interpretability.We show that by using CNN-encoded MDPstates as prototypes in a prototype-based model, we can construct RL agents that play Atari games well and use their learned prototypical states to explain their policies. I also show that by using sets of concepts about spatial statistics in a study area as prototypes in a prototype-based model, we can construct interpretable clones of human behavior policies.
The second part of this thesis begins with a study of the problematic usage of the popular post hoc explainers SHAP and LIME in business research. We argue that misinterpretation and misuse of SHAP and LIME can lead business researchers to be misled by SHAP and LIME explanations. Motivated by a desire to develop an ML explanation method that does not have the many issues SHAP and LIME do, we move on to propose a new ML explanation method that adapts philosopher of science Wesley Salmon’s Statistical Relevance (SR) model of explanation.
Details
- Title: Subtitle
- Extension, critique, and formalization of explanation in machine learning
- Creators
- Ronilo Ragodos
- Contributors
- Nick Street (Advisor)Tong Wang (Advisor)Qihang Lin (Committee Member)Patrick Fan (Committee Member)Palle Jorgensen (Committee Member)
- Resource Type
- Dissertation
- Degree Awarded
- Doctor of Philosophy (PhD), University of Iowa
- Degree in
- Business Administration
- Date degree season
- Summer 2025
- DOI
- 10.25820/etd.008135
- Publisher
- University of Iowa
- Number of pages
- xviii, 186 pages
- Copyright
- Copyright 2025 Ronilo Ragodos
- Language
- English
- Date submitted
- 07/18/2025
- Description illustrations
- illustrations (some color)
- Description bibliographic
- Includes bibliographical references (pages 134-149).
- Public Abstract (ETD)
In recent years, AI technology has seen a surge in popularity, but its success has been accompanied by fear and mistrust. Popular AI chatbots like ChatGPT may be viewed as untrustworthy because they sometimes say strange or factually incorrect things and what they say can never be fully explained. AI chatbot models are usually far too complex for a human to ever understand, and the programs and code that can give insight into how they work are hidden away from the public, guarded by corporate interests.
My research projects are reactions against untrustworthy AI systems and fall into two categories. The first category of my research relates to the development of AI models that are trustworthy because they explain their decisions, can be used for free, and because the code they run on can be viewed online by anyone. Unfortunately, I have found that even when we use so-called explainable AI, we may be misled by the AI's explanations. I explore this phenomenon in the second area of my research and offer new approaches to AI explanations. My goal is to find new ways for AI to explain things to us in such a way that the chances that we get misled are minimized.
- Academic Unit
- Tippie College of Business
- Record Identifier
- 9984948641502771
Metrics
13 Record Views