This summer, more than 80 students and faculty kept campus abuzz with the hum of hands-on field and lab research. The Seaver Summer Research Community’s work covered topics in biology, health and human sciences, mechanical engineering, computer science and more disciplines across the college. With such a diversity of research projects to choose from, we spoke to some of the students about their unique summer experience. Today, we spotlight Vivek Dhingra ‘25, a rising senior studying computer science under the mentorship of Andrew Forney, associate professor of computer science.
Describe your research in a way that a high school student would understand it?
There is a need to investigate the effect of changing certain factors that contribute to a specific environment. For example, how would students react in a high school setting if teachers had a requirement to provide support to any student receiving a grade below a C. These sorts of questions are difficult to answer because we’d need to set up an experiment within the environment we’d like to investigate. But what if we could simulate that environment? The simulation would include agents (AI that mimic humans in an environment) that interact with each other and a simulated environment fit with the variables we’d like to change. Now, we’d like to make these agents act as similarly to us as possible. In our current stage of the research, we are aiming to create more realistic agents by implementing causal networks that are supported by reinforcement learning. What this means is that each agent has a sort of naive understanding of how its world works, and the actions the agent chooses to carry out will then have some sort of feedback as specified by the simulation parameters, and the agent will adjust its understanding of the world to later choose actions that result in more positive feedback. For example, an agent may decide to increase how much time it sleeps as to increase the positive feedback given from a “health” node at the cost of getting less positive feedback from a “social” node.
Why would someone outside your field be interested in your project?
After the project has been fleshed out in the context of creating a simulation framework fit with these “smarter” agents, researchers can use it to mimic any environment as long as they have substantial data. It can be used within the context of courtrooms, corporate offices, government. Practically any environment where there are multiple people, as these are multi-agent simulated environments.
What are some of the lessons students learn while working on research?
Since I’m working closely with my professor who has been in the field for many years, I learn practical skills such as better code practices, but also the opportunity to understand his thought process during dense research with multiple stages like the one we’re currently tackling.
How does doing research complement coursework?
Coursework in CS at LMU usually involves projects that we tackle over a few weeks or longer if it’s a bigger project. In that, we write code to adhere to the algorithm we’re implementing to serve some purpose within the context of the project. If we’re trying to build an app that compresses data for the user, we might implement Huffman encoding and decoding. Research differs in the fact that the stuff we’re writing might be based on algorithmic paradigms we know, but the eventual outcome is almost completely unknown. We know how it’ll pan out in the context of a single sample by drawing it out, but who knows what happens after a thousand? I think that’s the biggest difference — being able to code without knowing exactly what’s coming up and not being able to query online sources to achieve exactly what your research is aiming for feels a little confusing, but also really satisfying.
What advice would you give to someone on the fence about doing research?
Whether you’re doing grad school or not, researching something you’re interested in can help immensely in gaining a deeper understanding and appreciation for the field.
