Image Source: securityboulevard.com

I recently concluded my remote internship with a research group at New Jersey Institute of Technology, headed by Prof. Christian Borcea and Prof. Chen Yi. Under their guidance, I worked on two papers which are currently under review. The first concerns a deep learning solution to help online publishers deal with ad-blocker users and the second is a comparative study of the two most common counter-measures to ad-blocker users. I wrote about the issues surrounding ad-blocking in another blog post.

Despite having no prior research experience, I was thrust into the midst of university-level research. And frankly, it was intimidating at the beginning. Everyone in the room (or rather, Zoom call) has a PhD or decades of professional experience. And I am a high school student. How can I contribute in such an environment?

I've learnt many valuable lessons from this 8-month long journey, many of which are applicable to business contexts and life in general. 

Play to your strengths, but don't neglect your weaknesses

Our group's studies revolved around data collected in collaboration with Forbes Media. Therefore, our experiments had two key elements — (1) data science, and (2) the intersection between economics and statistics, also known as econometrics. While I had data science internships in the past, my technical knowledge is not comparable to that of the PhD students and professors I was working with. And this was my weakness. My strength was my ability to digest related literature and consequently, come up with interesting research questions and hypotheses.

For example, we initially evaluated our deep learning strategy to deal with ad-blocker users by calculating how it affects revenue. Through my readings, I discovered the importance of reader engagement to publishers. If our strategy decreases engagement, a publisher's revenue will suffer in the long run since readers might switch to other news sources. Bearing this in mind, I proposed to evaluate our strategy using both revenue and engagement, which eventually contributed to the evaluation section of our paper. This showed me how, regardless of the level of technical knowledge one has, one can bring something unique to the table if he/she identifies her strengths, and capitalizes on them.

Yet weaknesses cannot be neglected. Deep weaknesses will drag one down regardless of how much he/she plays to her strengths. To contribute more constructively to our data experiments, I consistently read up on econometrics and other related statistical techniques. At the end of my internship, while I still had much to learn, I managed to conduct most of the data experiments in the second paper.

Communicate data with storytelling

Prior to my internship, I was under the perception that research was merely the act of systematically conducting experiments, and that research papers, as an extension, were akin to lab reports. But this couldn't be further from the truth. Data without a story is meaningless. Regardless of the field, all professional papers must contain a narrative or story, which minimally answer the following questions:

1. In what way are these results novel?
2. How do these results fit into the existing literature?
3. Are these results interesting?
4. How can these results be used for practical purposes?

Let's say I show that those under 30 are 50% more likely to use ad-blockers than those over 30.. And I end my “paper” here. This begs the question: So what? Why should anyone care that younger people are more likely to use ad-blockers? This statistic is only useful if it has managerial implications and connects to the existing ad blocking literature. 

Data without a story is meaningless.

If academic journals require highly technical research to have a coherent and interesting story, wouldn't this be even more important in a corporate setting? Being trained in data science and software engineering makes it easy for us to be caught up in technical details and lose sight of the big picture. We should always present technical results through storytelling, especially to non-technical professionals, rather than through pure statistics.

Success is built on failures

I spent a total of 5 and a half months on data experiments for the second paper. Yet the experiments included in the final paper took me at best 2 weeks. So where did the other 5 months go? Into failures. 

For example, I wanted to find out how a particular counter-measure to ad-blocking will affect the engagement of loyal and less loyal users differently. To answer my question, I spent a month studying and experimenting with cluster analysis, a data science technique. However, cluster analysis ended up being too unreliable and did not generate interesting results anyways. A month’s worth of work was abandoned.

Thomas Edison made 1,000 unsuccessful attempts at inventing the light bulb. When asked how it feels to fail 1,000 times, Edison elegantly replied: "I didn’t fail 1,000 times. The light bulb was an invention with 1,000 steps.". Edison could not have explained my experience better —  research is built on failed experiments. Yes, I did not end up using cluster analysis, but it nudged me towards the right statistical technique to answer my question.

"I didn’t fail 1,000 times. The light bulb was an invention with 1,000 steps." ~ Edison

As anyone can imagine, failed experiments are incredibly discouraging. But discovering what doesn't work only increases the chance of finding what works.

Conclusion

Working with highly-talented researchers is a humbling experience. Yet the feeling of contributing to academia as a high school student is rewarding and empowering. I will likely be returning in several months to continue working on ad-blocking research. In the future, I plan to apply what I've learnt not just in academic settings, but also translate this knowledge to life in general, because education is fundamentally about personal growth, not just getting straight As or on-paper experiences.

---

Special thanks to Sanjaay Babu for editing this article.