What is a Mellon Forum?
Mellon Senior Forums at Branford College seek to foster in our Seniors the sense of belonging to the community of scholars. Organized in the form of dinner meetings to which both Seniors and their advisers are invited, they provide some of our most talented students the opportunity to present the results of their independent research projects. The Forums also offer the possibility of especially close interaction between the Head of the College, the Dean, and Branford Seniors in comfortable and intimate surroundings. Branford holds a number of Senior Forums in the Spring semester. After dinner is served, students deliver research papers, usually fifteen to twenty minutes in length, to an audience of fellow Seniors. Attempts are made to cover a wide range of topics in various branches of the Humanities, the Social Sciences, and the Natural Sciences. Each paper is immediately followed by a lively and intense question and answer period which permits the Senior to clarify (and perhaps modify) his/her critical stance.
Branford’s Mellon Forum coordinators are HOC De La Cruz, Dean Galindo, and resident fellow Steve Blum. Each senior participating in the Forums are matched with one of our Graduate Affiliates who mentor them. An info session is held in the Fall semester for those seniors interested in participating.
What is the Mellon Forum schedule for Spring 2025?
January 14th
Tetsu Kurumisawa
Show and Tell – Mixed Goal Inference Using Controller Input and Verbal Utterances for Shared Autonomy
Shared control supports people in teleoperating robots during tasks that require precision. They assist the teleop- eration based on the inferred goals of the teleoperator. We suggest that traditional shared control systems, which infer operator goals solely from controller inputs using Maximum Entropy Inverse Optimal Control, are limited in their effectiveness 1) due to their assumption of optimality in human action and 2) in complex scenarios with multiple action options clustered densely. In this paper, we explore how to enhance the goal inference by integrating the operator’s verbalization of their thought processes alongside traditional controller inputs. We present a novel application of large language models (LLMs) to ground verbal utterances and infer user goals effectively. Through an N=45 user study, we evaluate the effectiveness of this approach in a human-robot collaboration setting in- volving a partially-observable puzzle. We find that the novel system provides faster convergence time and higher accuracy for intention inference. However, we also realize limitations in the system, as these improvements in inferences do not result in a significant improvement in human performance in robot teleoperation and participants do not prefer the novel system over the classical system. This study demonstrates the limitations of existing frameworks of shared autonomy, explores the strengths and weaknesses of using language for user intent inference, and provides suggestions for future work.
Rome Thorstenson
Forecasting Deforestation in India with Deep Learning for the GREEN Meghalaya PES Program
The government of Meghalaya is enacting a Payment for Ecosystem Services (PES) program to reduce deforestation by paying landowners not to deforest. Our work uses machine learning models and extends the work of Ball et al. [1] to forecast the risk of deforestation year-over-year and multi-year-ahead across the state of Meghalaya, India at a 30x30 meter resolution to improve the targeting and efficacy of the program. This has great potential to improve additionality (net gain) from PES because targeting enrollment directs more resources toward those with the greatest likelihood of deforestation, rather than, say, those who would not have deforested anyway. In this setting, 3D CNN architectures with region-specific features can achieve AUC 0.87 on next-year-ahead deforestation risk predictions. We consider various architectures and reaffirm the effectiveness of 3D CNNs for this use case, prioritizing additional labeled data over more extensive historical context in training. Furthermore, this work finds limited benefits of extensive feature engineering—a positive sign for deforestation prediction in regions with less accessible data. The best three-year-ahead model achieved 0.83 AUC; for five-year-ahead (extrapolated due to limited data), an AUC of 0.81 was calculated. The code is available at https://github.com/Rome-1/Meghalaya-PES.
January 21st
Jose’ Marin-Lee
Toward Independence: Participatory Democracy in Texas from 1821 to 1836
This paper explores the evolution of participatory democracy in Texas during its transition from a Mexican territory to an independent republic. By analyzing constitutional frameworks, local governance practices, and political organizing from 1821 to 1836, it argues that Texan colonists developed a hybrid governance model that integrated participatory methods within a representative system.
Key aspects of this paper include an examination of Texas’s governance under the Mexican Constitution of 1824 and the Constitution of Coahuila y Texas, where settlers utilized town councils and public meetings to navigate barriers posed by cultural, geographic, and political challenges. The paper also delves into the role of popular assemblies, petitions, and local committees in addressing grievances, culminating in representative conventions and the push for statehood. By examining these dynamics, the paper opens a broader discussion on the adaptability of democratic systems, offering insights into how hybrid models can be implemented in diverse political contexts.
Anabel Moore
The Anatomical Theater as a Collection
Anatomical theaters in seventeenth-century Europe, such as that of Leiden University in the Netherlands, were not only sites of dissection but also sites of display. In particular, the first and most famous image of the theater, created by draftsman Jan Cornelis Woudanus (1570-1615) and later engraved by Willem Isaacsz van Swanenburg (1580-1612) in 1609, shows the space filled with human skeletons, allegories to the dead, and living spectators engrossed in conversation. In this artwork, the emphasis is less on the insights gained from live dissection and more on teaching, learning about, and viewing every aspect of the postmortem human body. This presentation focuses on the space, specimens, and spectators of the anatomical theater, and the unique ways in which Woudanus’s print presents humanistic themes of medicine while establishing a dialogue with the contemporary imagery and political ideology of the Low Countries. Further, Woudanus shines light on the origins of medical practice as a discipline distinct from other humanistic and scientific endeavors, one that is dependent on a fundamental understanding of the embodied human experience. I argue that Woudanus’s print represents the anatomical theater as a museum, wherein the public could engage with the body in a way that was not possible in other spaces. Finally, I link salient messages from Woudanus’s print to Yale’s very own anatomical collection, the Cushing Center at the Yale School of Medicine, and present a framework for understanding the significance of anatomical collections, both past and present.
Ana Rodrigues
Mediating the Luso-Atlantic via the Panoramic: An Analysis of Russell and Purrington’s Whaling Panorama
During the nineteenth century, few industries made more of a global impact than the whaling industry did. Given the fact that killing whales is now illegal in most countries and homes are now electrified, it might be difficult for a modern audience to imagine that whaling led to such an immense globalization across the world’s oceans. Fortunately, there exists an art object that can help modern audiences understand how whaling economies reached every corner of the globe: Benjamin Russell and Caleb Purrington’s 1848 moving panorama painting, The Grand Panorama of a Whaling Voyage ’Round the World.
The Whaling Panorama is painted on four segments of cotton sheeting which measure 1,275-foot long and 8-foot-tall and were originally wound and displayed on two large wooden spools. This apparatus, placed behind a frame, allowed the painting horizontal movement. As the painting rolled past its audience on these spools like a film, it displayed the sights of a semi-imagined multi-year whaling voyage that circumnavigated the globe. Its content encompasses the “battle scenes” of whalemen, but its most detailed depictions are of port cities. In the first two sections of the panorama, the artists paint the coastlines of the Portuguese Atlantic: the Azores Islands, Cape Verde, and Rio de Janeiro. My paper will linger on the details of each of these port cities— paying special attention to how they appear to be organized and portrayed in their architecture, landmarks, natural features, references to industry, harbor vessels, and other painted details.
January 28th
Ari Berke
Sanskrit in Esther
This thesis tracks the Sanskrit influence on the Book of Esther, particularly through the lens of the Bhagavad Gita. It focuses on Esther’s first chapter, which, this paper argues, contains seven Sanskrit words embedded within the Hebrew text. These words form a Sanskrit “sub-story” within the larger narrative, revealing new layers of meaning in Esther 1, and in Esther as a whole. This original Indian influence on Esther is then traced throughout the story of Esther and in the closely related holiday of Purim.
Betty Kubovy-Weiss
Give Them Bread and Circuses: A Phenomenology of Fat Clowns
Philosophers have long sought to understand why we find things funny. While a comprehensive theory of humor is not really feasible, the incongruity theory of comedy suggests that jokes are funny when they succeed in subverting our expectations. My thesis applies the incongruity theory to Maurice Merleau-Ponty’s body schema. The body schema suggests that the self is the body, and that the “body” is always in flux based on how we can engage with the world through touch. By overlaying these philosophical frameworks, the paper seeks to understand why we find fatness funny, using the fat clown as a case study to highlight the incongruities of physicality.
Emma Polinsky
Stopping Disconnections: The Catalan Anti-Energy Poverty Movement as a Model for Autonomous Regions
Energy poverty–the lack of essential, affordable, reliable, and safe energy services–is a growing problem worldwide. In Spain, the problem has manifested in disconnections–residents being cut off from power due to inability to pay. The Alliance Against Energy Poverty (APE) was created in 2014 to address the growing concerns of residents unable to afford basic energy services. Catalan Law 24/2015 was created as a result of their efforts, which requires energy companies to notify social services prior to disconnections. The law also creates a class of “energy vulnerable households,” which may not be cut off under any circumstances. This paper aims to analyze the work of APE in both its work to establish the Law 24/2015 and later to improve the implementation and enforcement of the law. I examine APE as a node–a site within an organizing governing structure where knowledge, capacity and resources are mobilized to manage a course of events. APE’s work in Catalonia and broader national impacts present a potential model for use of the nodal governance framework in the context of other subnational, autonomous regions that are wrestling with the issue of energy poverty.
February 4th
Leyli Granmayeh
Brenda Kim
Matthew Merritt
February 11th
Raymond Fedrick
Simona Hausleitner
Zach Moynihan
February 18th
Thomas Lowe
Miette Maoulidi
Atticus Margulis-Ohnuma
February 25th
David Acquaah-Mensah
Elizabeth Greenberg
Lauren Hartz
Halyn McKenzie
March 4th
Caue Riberio Pascarelli Lopes
Anika Seth
Selihom Yosief
March 25th
Odessa Goldberg
Ushuu Namarra
Iana Phipps
Kate Reynolds
April 1st
Olivia Bell
Gianna Campillo
Theo Curtis
Elyse Thomas
April 8th
Nyche Andrew
Nikhe Braimah
Sophia Chmelar
Kalei Memmer
April 15th
Caleb Hwang
Hannah Lothian
Kasey Maguire
Nicole Massoumi
April 22nd
Carl Bager
Gavin Guerette
Nora Hylton
Liana Schmitter-Emerson