EEG Controlled Robotics; When Gut Feelings Go Down the Drain: The St. Francis Dam Disaster and the Perils of Skipping Reason

In both my technical capstone and my STS research, I explore the complex relationship between intuition and reason in engineering decision-making. My capstone project focuses on using reinforcement learning algorithms to train a robotic arm to respond to EEG brain signals, creating a system that learns from experience and corrects its behavior over time. My team used machine learning to systematically correct or improve on raw, intuitive responses or EEG signals. This process is particularly dissimilar from my STS research project, which analyzed what led to the collapse of the St. Francis Dam in 1928. Through the Social Intuitionist Model of Moral Judgment I argued that the tragedy was mainly attributed to the intuitive, overconfident decision-making of William Mulholland, the head engineer on the project, and his team. I show the outcomes when human intuition is not checked by reasoning or systemic learning. While my capstone project looks forward, designing a system to learn and adapt, the other revisits the past, examining the danger of relying too heavily on instinct without the balance of systemic engineering review. The technical capstone can be seen as an attempt to embody the lesson of STS research.

My technical project worked to further the research in the field of prosthetic control through an EEG by reinforcement learning. We first captured and filtered EEG signals associated with intentional mental commands (e.g., muscle tensing vs. relaxing) using an OpenBCI system and saved them as training files. Then we used them to train a reinforcement learning model to classify and respond to these EEG patterns. Following this process, we designed and constructed a robotic arm that could grasp and ungrasp its fist. Once completed, we implemented real-time EEG-to-actuation processing using the OpenBCI system and Raspberry Pi. At the conclusion of the project we demonstrated the systems functionality through live trials with pre-collected and real-time data.

In my STS research paper, I utilized Jonathan Haidt’s Social Intuitionist Model of Moral Judgement to analyze the St. Francis Dam collapse. This framework claims that moral decisions are made through intuition. Rational reasoning is what comes afterwards in which a person searches for arguments that will support the already-made moral judgment. I demonstrate that the collapse was a result of intuition-driven decision-making overriding rational analysis because William Mulholland and his team relied on personal expertise, gut instinct, cognitive biases, and social pressures rather than systematic engineering review.

Overall, having the opportunity to work on both of these projects in the same academic year has provided a greater understanding of the role in decision-making in engineering. While developing reinforcement learning algorithms to train a robotic arm, I became increasingly aware of how important it is to design systems that can learn, adapt, and correct themselves. At the same time, studying the St. Francis Dam collapse prompted me to be aware of how easily overconfidence and unexamined instincts can lead to catastrophic failures. Throughout my engineering career, it is my intention to remain aware of all the factors that influence engineering decisions.

School of Engineering and Applied Science

Bachelor of Science in Mechanical Engineering

Technical Advisor: Sarah Sun

STS Advisor: Benjamin Laugelli

Technical Team Members: Hailey Boyd, Joshua Rivas-Zelaya, Cayla Celis