Projects

Current

Research

Mind-Wandering Self-Reports

Mind-wandering can generally only be measured by self-reports. You can't directly measure it, so you have to rely on asking people whether their mind was wandering (eg: "were you thinking about something other than the task?"). But people consistently misrepresent their own mental states, both intentionally and unintentionally. For example, people tend to make up reasonable explanations for their decisions and behaviour (attribution bias). Thus, maybe people are more likely to report mind-wandering when they make mistakes, subconsciously reasoning: "I didn't perform as well as I should have, so I must have been distracted." To test this, we're using the Sustained Attention to Response Task to control whether people make errors, then following up the task with a mind-wandering report. We'll compare the "error" group to the "no error" group and see whether the former reported more mind-wandering than the latter.

Cognitive Biases in Nuclear Power Regulators

Currently in the literature review stage, this project aims to assess what cognitive biases might affect optimal decision-making in nuclear power plant regulators. Specifically, we're interested in small modular reactors (SMRs). Given the history of nuclear plants and their perceived danger, regulators tend to be extremely conservative about approving changes. However, in order to make nuclear power a viable clean energy solution, the regulations need to keep up with the technology innovations--lest nobody invest in that innovation. We still have to keep things safe, though. The goal of this project is to preemptively identify places where regulators might be unnecessarily conservative.

Bias Injection in LLMs

With the rise of GPT-4 and other LLMs, a concern may be the AI agents' biases. For example, we know that ChatGPT is programmed not to say certain things. Hopefully the folks making these LLMs have good intentions, but this project addresses the possibility of malicious agents interfering in LLM training. If an attacker injects inappropriate biases during training, and a national intelligence agent uses that LLM to summarize documents, the LLM may produce skewed results that affect the intelligence agent's decisions.

HRI Metrics Metric

As a member of GMU's brand-new Human Robot Interaction lab, I'm helping to develop a database where we compare current state-of-the-art metrics in HRI to a set of benchmarks to assess how well the metrics have been verified and communicated.

Personal

All The Poems

I am compiling an anthology of all the poems and quotations that I enjoy or have some connection to. This book was born out of a frustration with carrying around anthologies of specific authors’ poetry where the content I wanted to reference only made up ~30% of the book. So, I decided to make my own. It has lots of poems by my favorite authors, single poems by assorted other authors, poems by friends & family members (and by yours truly), and a selection of quotes and short prose.
I published and printed the first edition (a single copy) in the summer of 2021. The second edition is coming soon, now with better formatting, hundreds more pages of poems (~215 → ~515), more indices, and (hopefully) better binding. The indices are my current challenge; there are so many poems now that I can no longer do indexing by hand. My plan is to parse the giant Google Doc that is my anthology using its consistent formatting as a cue. I want to make a table: title, page number, author, first line, topic(s). From there, I can make an index for each of those columns (other than page number, of course).

Textiles

As of the beginning of 2022, I am challenging myself to not buy any clothes. I will only repair clothes I have, make new ones, and accept free ones. This came about after two events: in the Fall of 2021, my host mama in Uganda taught me how to make my own dress—and in the Spring of 2022, I threw out egregious quantities of clothing while moving. The former showed me that I have the ability to make reasonable, wearable clothes, and the latter reminded me how disgusting and wasteful the fast fashion industry is. I want to force myself to be thoughtful about the clothing I add to (and have in) my wardrobe, with the intention of avoiding waste, expressing myself creatively, becoming more self-sufficient, and being more aware of the materials that affect so much of life.
I also have several other projects in the works: lots of adjustments to not-quite perfect clothes, adding pockets to all my dresses, making a few long-sleeve shirts, and a top-secret crochet project (which I began in the summer of 2020, I believe, and which will probably take several more years).

Past

Research

Environmental Preferences

At the beginning of my senior year of college, I won a grant to do independent research in Environmental Psychology. I mostly just wanted to try my skills at self-driven research generally, but I selected a question from my question book (where I keep a list of questions I have about the world that could be answered). I was curious about whether people who grow up in certain environments are more or less likely to admire those types of environments later in life. I had noticed that I love mountainous, richly green terrains, and I wondered whether people who grew up in a flat desert (rather than the rich, green, mountainous land I was raised in) would love the mountains and greenery more than I do (because of the novelty) or less than I do (because they’d feel less “at home” in it than I do).
I selected photos of different terrain types on a 3D grid of water (low, med, high), tree-cover (low, med, high), and elevation change (low, med, high). So, low water, low tree-cover, and high elevation change might be a mountainous desert scene. High water, low tree-cover, and low elevation change might be a sandy beach scene. I ran an online survey where participants scored those photos according to comfort, beauty, and value (i.e., “How important is it that we preserve this landscape?”). They also reported their own childhood locations—the places where they were raised.
I collected a LOT of data, and then I was graduating and moving to East Africa and the data-analysis portion slowed down to a crawl. I’ve done a decent amount of it since then, but I haven’t polished the results. It turns out I may have collected a little too much data! Oh, also, that project was accepted to the 2021 Environmental Psychology conference as a poster, but regrettably, my university’s funding to fly me out there fell through.

Anti-Scam Dialogue Generation via Scripts

During my second REU at Carnegie Mellon University (in 2020), I studied dialogue generation for adversarial anti-scam use via natural language processing and a study of scammer behaviour.

Redundant Github Development

During my REU at Carnegie Mellon University in 2019, I helped develop a bot to identify redundant work in GitHub repository forks to decrease software engineers' mental workload.

University-Sponsored

Robotic Marching Band Platform

On a team of four, I helped build a remote-controlled moving platform for instruments and marching band members.
At my college, seniors in engineering spend the year on a small-team design project. Some projects have interdisciplinary teams, but typically the individuals within a team mostly do stuff related to their major. As a CS major, I thought most of our projects sounded boring compared to others’ options. Programming a Minecraft mod vs. building an ember generator to test fire behaviour? Easy choice. Anyway, I always loved to pack in as many non-major-field things to my degree as I could. So, I applied for lots of the mechanical engineering projects (which were generally the most exciting-sounding, and almost always the most hands-on).
This project, the marching band platform, was a multi-year project which typically alternated between mechanical and electrical engineering teams. The years overlap (Fall-Spring vs Spring-Fall), so a mechanical engineering team would begin while the electrical engineers were wrapping up. This term it was mechanical, but the faculty mentor for this project selected me because he figured I could work on the electrical side to improve cohesion between the teams.
Now, again, I am a computer science major. Computer science is not the same as electrical engineering. I’d done some soldering, some digital logic, some Arduino programming —but I knew next to nothing about voltage or circuit design. Mechanical engineers don't seem to realize the difference between CS & EE, so I faced a steep learning curve. I was also learning carpentry, milling, welding, and textiles. I dabbled in every aspect of the project and worked with people in Marching Band, Mechanical Engineering, Electrical Engineering, and Family & Consumer Sciences.

Hartung Grasses

In my senior year of college (2020-2021), I won a grant to replant some of the lawn on my college campus. I’ve always been disgusted by lawns—their monoculture-ness, blandness, and resource-sucking-ness. I had grand plans to replace all the lawns on campus with cheaper-to-maintain, beautiful moss lawns or flowery gardens. Unfortunately, I’d vastly underestimated the cost of lawn replacement—even if they would save the university money in the long run, it’s very hard to get them to put down the initial bill. Though I received thousands of dollars for this project, I had to settle for a fairly small section of lawn. We covered the old lawn with mulch to kill it off, then re-seeded it with native, more drought-resistant grasses. No soft moss lawn or bee-humming flowers for me after all!
I’d also thought of doing a collaborative project (universities *love* big collaborative projects) between the civil engineers and computer scientists to reduce the waste of sprinkler runoff. I walked around campus at night often, and I was regularly disgusted by the way sprinklers would be watering pavement for hours on end. So, I imagined up a competition with a small monetary award (you can get college students to happily work extremely hard for very little pay if you do a competition) where civil engineers would GIS-map the campus greenery, the places that needed watering (grass) and the places that didn’t (pavement). Then you’d send the results to the computer science students and have a competition where they use ML to optimize sprinkler placement according to that data. Unfortunately, I didn’t realize how bureaucratically lethargic universities are, nor how much it costs to redo an entire campus worth of irrigation.

Personal

Textiles

Almost all the clothes I have made have been part of “family sets”, where we make clothes for the whole family from one bolt of fabric so that it’s all matching, Sound of Music style. The first time I did this was in the Fall of 2021 with my Ugandan family; I made a dress for myself and the others made skirts, shirts, etc. for themselves. For Christmas 2021, I made another set for my “real” family: shorts for myself, a tie for my dad, a dress for my mom, a shirt for my brother, and various accessories for my sister. Then, for Christmas 2022, I made a set for my Portland host family. The girls made themselves a dress and a skirt, and I made a dress for the mom, a tie for the dad, a bandana for the dog, and a shirt for myself. It’s a fun way to represent togetherness and my adoption into these different families.
Some other assorted textiles projects: a tshirt quilt with many of my old sentimental tshirts and a comforter that was no longer especially white; upholstered pillows; matching set crocheted scarf/hat/mittens; crocheted stuffed animals; crocheted water bottle holders.

This Website

Creating this website has been a goal for a long time, though I was always too intimidated by the prospect to make much headway. My dad purchased the domain annikaesau.com many years ago, but I struggled to decide what to do with it, let alone *how* to do anything with it. In 2022-2023, I finally began to give it form; its purpose is to present information about my life, activities, and interests in an organized fashion. I’m mostly designing it with myself as the target audience, but it seems useful to publish it anyway, if only for the sake of accountability. This site is intentionally designed to be extremely plain text. I know it's not sleek; it's not meant to be. As much as I love beautiful things, I'm also inclined toward simplicity. I'm tired of over-designed (and addictive) technology, and I'm all for minimizing time on screens.

Bike Shelter

I built a bike shed in front of my house to protect my bike from wind and rain. It consists of three walls and a slanted roof, and it's about 6 feet long by 3 feet wide and 4-5 feet tall. This was my first large independent construction product! I purchased lumber for the first time :-) including a bunch of 2x4s which I brought home in my roommate's tiny sedan.