I had to brew it and let it sit in my head. Left it there with patience, like how you would brew cold-brew coffee.

Or I can say I let it simmer, on low heat until it finally boiled and things clicked. ****How I found my thesis topic is a long story. Personally, it’s long. A lot of things happened during that time. But I’m wrapping the story into a shorter one. It’s been six months since I started doing my thesis experiments (an additional 1~2 months prior to formulate it). Now, since I’m near the end, I’m writing this blog post to recall the “journey” and be ready to wrap it up.

Let’s start in early 2019.

It was my second semester in MS, and I was taking CS 298, which is a required pre-thesis course. I had to start thinking of possible domains, applications, and ideas for my thesis topic. I naturally gravitate towards things that involve humans, so I was really considering problems related to the brain. The thing is, that area is not our lab’s expertise. I had to ask for help.

So I did. I reached out to Paul Regonia, who was once my instructor during undergrad. I know he also worked with brain stuff, and is now pursuing that for his Ph. D. in Japan, so I asked if he has any advice he can give me. Should I push through with it? How hard will it be to work on something our lab don’t have much experience in? But when he answered, he gave me so much more: an internship with their lab in Nara Institute of Science and Technology (NAIST). Well, he didn’t technically give me the internship, he just referred me to his sensei, and I still had to do the formalities and the papers from there on. I did. I told them I was hoping for a sandwich program. I wanted to work with them for my thesis dissertation. By the end of that semester, I was already reading up on imitation learning (a subarea of reinforcement learning), eye-tracking and programmers’ eye movements. (It’s no Neuroscience, but it’s still human behavior and I’m okay with that.) By the end of that semester, I had a ticket and a student visa to stay in Japan from August to October.

Obligatory photo.

However, things didn’t really go according to plan.

Recruiting programmers to have their eye movements recorded wasn’t easy, and we didn’t have enough data to work on more experiments. I wasn’t able to reach the Imitation Learning part of their project. I was worried because I couldn’t wait: I had a deadline to finish my graduate studies (scholarship).

But at that time, thanks to the copious amounts of time I had, plus the great research environment at NAIST, I had already read a large deal of literature about eye-tracking. The more I scoured and traced research papers through Nature, ScienceDirect, and Google Scholar, the more I became fascinated by eye movements. I had moved on from reading about eye tracking in programming to eye tracking for Cognitive Science. How certain behaviors spontaneously correlate with certain human qualities. A jittery eye movement can mean I’m doing some hard thinking (e.g. solving a math problem)? Eye movements can be used as biometrics? Eye movements can reveal the images we’re visualizing in our heads? Cool. Cool cool cool.

A bigger picture of the literature was forming in my head, and I had this stuck to the wall beside my desk at NAIST’s (amazing) Guesthouse:

Photo taken August 21.

This photo was taken August 21, just some three weeks into my internship. A couple more weeks passed by and I did some other analysis on the data that we already had**. On October 1, I left NAIST and went off to travel and see more of Japan. I shrugged everything about research off my mind. It was a blissful week. But these ideas were always in the back of my head.** I was realizing that it could be a thesis topic. Finally, I came home to the Philippines with that question: *Is a vector representation of eye-tracking data possible? (*That, and a complete set of home brewing equipment: a Hario scale, grinder, kettle, and carafe; and a Kalita 102. Yes, I was going to **brew.)**

When I got back, I took one more week off research to settle in at home again.

The question remained in my head, I was aware it was there, I was aware about the wrongness of the question: it wasn’t a proper one. Is a vector representation of eye-tracking data possible? Of course, it’s possible. The right question was how. The other questions were: why? how useful can it be? what is it good for? but I didn’t touch them. Let them sit there while I took my time to rest. Like leaving ground coffee in immersion with cold water , slowly extracting the coffee compounds…

When I got back (to research), I had the fervor to pour over (hehe) research papers again (and yes, I was brewing pour-over coffees at home everyday). This time, with a more purposeful approach: to answer the better questions sticking out from my mind. Every morning was a whirlwind of papers and references. I was killing the Open Link in New Tab button (and whispered some appreciation for Safari’s amazing memory management and to the Notion app for being a superpowered note-taking app). Every morning was like that, until the references circled one another, or until I told myself to stop breadth-searching and start depth-searching (or: stop exploring and start exploiting? 🤭)

A portion of the RRL table I had put up on Notion. 81 papers, plus more that I read before setting up this table. And this was just the first half of my RRL.

It took me a few more weeks before I could say I was ready to present my thesis topic proposal to my adviser. It was hard consolidating so much information from different areas of literature: eye movements themselves, their applications, computational methods for eye movements, and possible methods for my thesis. When I was at last able to convince myself of the soundness of what I was wanting to do (create a vector representation), I then had to narrow in on how exactly I was going to do that (graphs? point processes? images?), what specifications of data I needed, and where I can get those.

Then, on November 12, I finally sent an e-mail to my adviser.

On November 18, I made the drive from Laguna to UP and came home with an approved thesis topic.

It’s done brewing.