Very broadly, I look at how slowly things change over time. Sometimes those methods involve computers.

This page includes independent research as well as my work with academic institutions. Keeping in the spirit of good science communication, every study outlined also contains a blurb in simple English about what the project is and why it's worth being excited about. 

Natural and Physical Sciences


Oberlin College, Depts. of Geology and Biology

A 3D-printed CT scan of the trilobite C. tesselatus, showing the fringe's general inability to let fluid pass through.

A 3D-printed CT scan of the trilobite C. tesselatus, showing the fringe's general inability to let fluid pass through.

The morphometric uniqueness of the trinucleid family has led to a considerable amount of attention in paleontology literature. In particular, their distinctive pitted cranidial fringe has been the subject of considerable debate for over a century. Although anatomically well understood, their function remains unknown. The vast majority of papers on the subject offer speculation on their usage, yet none have attempted to model their hypotheses with experimental methods. This paper attempts to shed light on the validity of the most contentious of these theories by physically modeling a trinucleid fringe and testing its resilience in several situations.

This project consists of various tests on 3D-printed CT scans of C. tesselatus, and aims to disprove a variety of claims made about the function of the pits. In particular, we will be examining its use as a filter feeding sieve, structural support, and a circulation mechanism. 

Non-linear ontogenetic shape change in Cryptolithus tesselatus (Trilobita) using 3-D geometric morphometrics: With Dr. Melanie Hopkins

American Museum of Natural History, Dept. of Paleontology

tl;dr: A common technique used by paleontologists is geometric morphometrics. If you take multiple snapshots of an animal as it grows, you can quantify how its shape changes by looking at how the shapes in the photographs change. The issue with 2-D morphometrics, however, is it loses all nuance in the third dimension. In this study, we used CT scans of a particular trilobite, Cryptolithus tesselatus, and noted how its shape changed during growth in three-dimensional space. From this nuance, we are able to make new inferences on the trilobite's paleoecology.

A decrease in the rate of cephalic shape change late in ontogeny has been documented for several species of trilobites, possibly associated with the cessation of segment release into the thorax. Qualitative descriptions of the ontogeny of Cryptolithus tesselatus Green, 1832, suggest that shape change in the cephalon was strongly influenced by the progressive accommodation of large funnel-shaped perforations (“fringe-pits”) over several molts. The number and arrangement of fringe-pits was established early in ontogeny, however, before thoracic segment release was completed. Due to the unusual and highly convex shape of the cephalon, we use three-dimensional (3D) geometric morphometrics to quantify shape change in this species and determine if there is a rate shift, and at what point in development this shift occurred. Three-dimensional morphometrics was made possible by extracting fixed and semi-landmarks from surface reconstructions of C. tesselatus rendered from CT scans of silicified specimens. Results show that the cephalon continued to change shape into adulthood, but that a threshold model with a rate shift associated with the cessation of new fringe-pits is best supported. 2D landmarks taken from the dorsal view fail to capture the dramatic change in convexity of the cephalon during development, but model comparison results are consistent with those based on the 3D landmark dataset, allowing comparison of this aspect of ontogenetic change with other species. Based on these comparisons, it appears that 1) trajectories are often better characterized by threshold models than simple linear regression models; 2) the timing of shifts may not be phylogenetically conserved.

Topographic Implications for Tornado Climatology

Oberlin College, Dept. of Geology

tl;dr: We've all been told the U.S. has an infamous "tornado alley" that stretches uniformly through the midwest. It's a lie. Kinda. While the majority of tornadoes have an overwhelming bias towards the midwest, there are five "clusters" that have unusually high tornado densities. What do those regions have in common? As it turns out, the five American tornado clusters all occupy a very narrow range of elevations and have roughly the same annual precipitation. By calculating zonal statistics via ArcGIS, we can isolate and quantify ideal tornado-forming conditions. By defining tornado density as a function of topography and climate, we can actually generate a map of tornado occurrence throughout the world with incredible precision.

Based on large-scale data analysis, the United States lacks a singular “alley” of storms and instead contains a patchwork of high density tornado clusters (Frates 2012). Each of these clusters has a statistically significant outlier in regards to tornado frequency— the most dense of these outliers has over one thousand times the national average. While these clusters have been previously described, little is known about why specific North American locations have such incredible tornadic potential. An understanding of these conditions could lead to better accuracy in storm prediction and increased safety for highly prone regions.

Data from the NOAA Severe Weather Project is used to calculate the pathways of all reported tornadoes from 1950 though 2015. Storms are visualized in ArcGIS by algorithmically graphing a line between each storm’s two terminal points (Figure II). A U.S. map is then generated that colors areas based on tornado likelihood (Figure III). Virtually all areas with high storm density, including the tornado clusters, are located in the midwest and Great Plains. This is due to the regions being disproportionately flat, balanced in moisture levels, and bordered by two major mountain ranges to the east and west (Brooks 2004). These conditions generate high cyclostrophic balance— an atmospheric state that gives rise to the vast majority of tornadoes (Frye and Mote 2010). 

The five largest U.S. tornado clusters all share narrow topographic and climactic ranges (Figures VII and VIII). These factors attain a balance of optimal storm cohesion, updraft, and low atmospheric density. While several other factors that contribute to tornado climatology have been described, such as soil moisture (Frye and Mote 2010), decade-long weather oscillations (Enfield et al. 2001) and global climate change (Diffenbaugh et al. 2008), this study suggests that storms can be predicted with significant acuity as a function of local topography and precipitation. This is especially significant as both of these statistical measures are easily collected in virtually all regions of the world— even those that do not have official geographic databases. 

With this model, a map of all regions on Earth that satisfy the elevation and climate ranges is generated (Figure IX). Not only does this model bear striking resemblance to maps of tornado distribution (Brooks 2004), but also suggests at the possible presence of tornado clusters in underreported areas. These areas, such as Southeast Africa and Patagonia, have few official storm reporting services as well as low population densities. This model suggests the presence of never before described tornado alleys, that very possibly exist despite the lack of reported incidents.

Climatic Fingerprints of Channel Sinuosity: with Dr. Matt Jungers and Willa Rowan

Oberlin College, Dept. of Geology

tl;dr: Are rivers near the equator "bendier" than rivers elsewhere? Our study used a computer program to crunch global databases to find the average "bendiness" of rivers in particular climates. We believe that we can predict where in the world a river is just by its bendiness.

Our research seeks to establish a correlation between a climate and its average degree of river sinuosity. Two computational models were used to find a metric of average sinuosity for all rivers of any given latitude on planet Earth, as latitude can be used as a proxy for climate. Our results, however, displayed little to no correlation between latitude and average sinuosity. We believe this is not completely due to any errors in analytical technique, but from a high degree of error within our already low-resolution dataset. Though a link between channel sinuosity and climate cannot be established from these results, it does not follow that such a trend does not exist.

A Geometry of Musical Counterpoint: WITH DR. Katherine Jones-smith AND Simon gilbert

Oberlin College, Dept. of Physics

tl;dr: What will pop music sound like in 50 years? We built a computer program to listen to pieces of music and create a unique ID for them based on counterpoint (a musical term for the relationship between two independent lines of music). In the future, we plan on having the computer look at hundreds of thousands of compositions and see if there are any surprising trends in counterpoint over the course of music history.

Using computational models, we created a program to visualize the harmony and counterpoint of a composition. We were able to generate a "fingerprint" for each piece of music, treating individual lines of music as physical projectiles. The project seeks to use this geometric model of composition as a way to identify trends in music as a function of time-- to notice how counterpoint has changed as a practice throughout history, and to theoretically project the trajectory of counterpoint in the future.

Computer science/Human Computer Interaction

The Alex Project: With Dr. Justine Cassell and Samantha Finkelstein

Carnegie Mellon University, The ArticuLab, Dept. of Human-Computer Interaction

tl;dr: Can we design computer programs to intuitively teach students linguistic dialects? American students who speak African American Vernacular English (AAVE) are strongly and very unfairly oppressed, especially in academia. The challenge is to teach kids how and when they should switch between an AAVE dialect and Mainstream American English (MAE) without claiming that one dialect is superior to the other in any way. The ArticuLab created Alex, a virtual human that is capable of speaking both AAVE and MAE. We worked on fine-tuning Alex so that students would begin to understand when and how the dialect switch occurs.

The Alex Project is a response to the oppressive stigmatization of non-standard dialects. The ArticuLab’s research primarily concerns AAVE-fluent children, and seeks to find a way to teach code switching without damaging the child's cultural identity or self-efficacy. A tool for helping guide this dialect switch is Alex, a genderless virtual human that can be programmed to have conversations in both dialects. My research looked at ways of designing a child-appropriate interface that would stress the nuances between the two dialects and when a code switch would be appropriate. While based in HCI research, my work also looked heavily at design theory, sociolinguistics, and child psychology. Some of our principle research questions were:

  • Do students learn differently with a virtual peer who speaks like they do?
  • How can we design educational technologies to be more culturally congruent?

My final presentation concerned how particular audiovisual changes in the design of the interface affected the dialect of the student Alex was talking with. 

Methods of Research in HCI, With Dr. Cecilia Aragon and Katie Kuksenok

University of Washington, Dept. of Human-Centered Design and Engineering

The computer-supported cooperative work (CSCW) community is one that bridges research and practice, and has within the last decade become particularly interested in studying social media content, both to understand how certain human organization phenomena play out at scale, and to build a better understanding of the relationship between emergent cultural practice and technical affordance of social media platforms and services.

Timelines of Controversy: Argumentative Types on Wikipedia Talk Pages, With Dr. Cecilia Aragon and Katie Kuksenok

University of Washington, Dept. of Human-Centered Design and Engineering

Every Wikipedia entry has an associated “talk page,” where editors can debate whether particular bits of information should be included and their reasons behind those decisions. However, not all talk pages are created equal— certain pages are more argument prone than others, and particular argumentative strategies are used grouped around certain topics. This project asked questions like:

  • Which topics are inherently controversial? 
  • How and when does controversy arise?
  • Can we use the predictive capabilities of these trends in order to improve the efficiency and factual quality of Wikipedia?


Lab Researcher: with Dr. David Hurst Thomas, Dr. Matt Napolitano, and Dr. Franco Fortuna

American Museum of Natural History, North American Archaeology Lab

The North American Archaeology Department of the American Museum of Natural History offers Lab Researcher Internship positions for undergraduates, recent graduates, and graduate students. The North American Archaeology Lab handles, stores, and analyzes a wide variety of artifacts from southeastern North America. Lab researchers have had the opportunity to work with faunal remains, lithics, Native American and European ceramics, Spanish colonial material culture, and a large number of other material culture types.

Analysis and Construction of a Samnite Villa: With Dr. Susan Kane

Sangro Valley Project, Oxford University and Oberlin College

The Sangro Valley Project completed its second and final year of excavations in San Giovanni di Tornareccio during the 2013 season. Work focused on three areas of the villa complex: the drain in SG 2000 was reopened to explore its relationship with Structure C and its associated run-off zone to the south; SG 3000 was expanded to determine the extent, function and phasing of Structure D; and exploratory trenches SG 5000 and SG 6000 were placed in the environs of the villa complex. The 2013 discovery of a lime kiln is proof that the villa complex was recycled, explaining why only the lower foundations of the walls of the complex remain and the overall lack of material evidence at the site.