Gone are the nights spent stuck on code, just to realize one hour later that the villain was a missing semicolon and a misspelled function name.
My use of generative AI to work through similar issues exponentially increased my productivity and understanding, cutting down my working time by an estimated 30%.
Even with all the support school can offer, TAs and professors cannot be available for 24/7 help.
As a student who didn’t use AI in undergrad, it took me a while to learn that its use did not equate to cheating.
When used ethically and honestly, it means maximizing the time I spend working, getting timely answers to small questions, digging into deeper explanations of concepts, and ordering my work more efficiently.
Using AI to quiz me on class concepts, suggest more efficient problem-solving or algorithm creation, or even set up an outline for presentation slides, has left room for me to fully embrace the creative parts of my brain, no longer impeded by my current knowledge.
These are some crucial ways AI supported my journey and allowed me to focus my time.
Concepts and Foundational Knowledge
The summer involved learning a lot of foundational statistical knowledge, and the key was learning how all the concepts are connected. I used both ChatGPT and Copilot to test my knowledge and confirm my learning through questions like:
In R, what is the difference of aov() and lm() to test interactions of two categorical predictors with a continuous response?
Quiz me on the following topics from my statistics class with a mix of multiple-choice and short-answer questions.
I fit a Logistic regression model I ~ C, where C is a quantitative variable and I is a binary outcome. I did an LRT to compare this to a GAM model where the spline smoother was C, and it was I ~ 1 (python). The Likelihood ratio test had a significant p-value, which tells me that the variable C fails the assumption of linearity in the logit. Is this correct?
Project Breakdown, EDA, Troubleshooting
The summer project was very open-ended in a way that was super exciting to me, but there were many times that my eyes were bigger than my stomach in terms of what I wanted to test and create. This project coincided with learning R and Python, and I was new to both languages. With the use of Copilot and ChatGPT, I asked for help breaking down the EDA process with specific leading questions – I wanted AI to guide me through the process, then I would write the code structure I thought would work and ask AI to identify bugs or logic problems.
Having the structure of what needed to be done, helpful coding hints without complete reveal, and assistance with interpreting my findings, made my learning significantly more successful.
I was able to bring greater insights to my team, be more creative with delay propagation models and predictive models, troubleshoot and test many more ideas than I would have on my own, and overall spend more concerted effort on the parts of the project that would benefit me most.
Troubleshooting
Troubleshooting code is the fate of everyone in data science, which can be a fun puzzle or the longest part of your project. Before my use of AI, problems with syntax or errors in building functions would have meant excessive time spent online, waiting in a queue for TAs, or hours of work. Having AI at my fingertips to troubleshoot this, and wonderful TAs at school to discuss problems or concepts that AI can’t help with, has made learning coding languages incredibly less daunting and much more exciting. Some questions I have asked are:
I have the following traceback error from Python – can you lead me through how to interpret it and identify the problem in my code?
I have written the following code for XYZ purposes. It is successful at XYZ, but I want to write it as a function so I can loop through variables– how do I update the code to do so?
I set up an IDE in VSCode named ABC, but when I try to install the following package within that IDE I get an error – how can I find where the issue is?
This code feels clunky. Is there a better way to structure it or streamline the process?
The encouragement of my professors to utilize AI was a major contributor to my summer success of use. The IAA outlines clearly what level of generative AI is allowed for various assignments, teaches us how to use it ethically, and reminds us that it should be a tool in our toolkit.
Each week, I learn that I am only scratching the surface of AI use in analytics. I am more than able to recognize that there are many potential pitfalls to these tools, but ethically harnessing them to support my learning has been invaluable.
Pic: Summer Project Team Presentation
Columnist: Rodriguez Natalie