Hidden Gems at CHI

Being ignored by others hurts. We write our favorite research papers and publish them in top venues, but over the years, they have been completely ignored by the research community: these papers receive few or no citations. How could that be?! Sometimes, it may not be the fault of the papers themselves; it’s just that they got buried somewhere in the mountains of published papers that grow higher every year.

In this project, I want to identify the less-attended papers at CHI¹. I attempt to provide a lens that helps people explore and appreciate these overlooked papers and perhaps follow up on them to bring these papers back to life.

Method

I have collected the CHI proceedings² (year, title, doi, authors) between 1982 and 2025 from the ACM Digital Library³, together with their citation numbers. I compute a hidden index for each paper. The hidden index is calculated using the paper's citation count minus the mean citation count of all papers from the same year, estimated through log-normal distributions⁴. The idea is that publications are cited differently each year, so we want to calibrate that to compare all papers. In general, a lower hidden index means that the work is more hidden. I then sort the papers based on the hidden index (low to high), and export the dataset in a .csv file.

How to Use?

You can simply download the .csv file and start exploring the dataset. You can take a serendipity-based approach where you browse through the paper titles line by line. For example, I have discovered the paper "Object manipulation in virtual environments: relative size matters" (with a hidden index of -122 from year 1999), which suggests that matching cursor size to target size can improve performance, an interesting idea worth exploring. You can also search for the least-attended paper of a famous researcher in the field and make fun of them. Moreover, you can take a systematic approach by analyzing the word clouds of papers with a hidden index smaller than -100.

Come on, treasure hunters, use your imagination to find the hidden gems at CHI.

Figure 1: A comparison of the average citation counts estimated using log-normal vs. normal distributions

Footnotes

1. ACM CHI is the flagship conference in the field of Human-Computer Interaction. ↩

2. I have tried to filter out the extended abstracts, but they may still be present in the dataset (they are a bit difficult to distinguish in earlier years). ↩

3. https://dl.acm.org/conference/chi/proceedings ↩

4. My statistical analysis shows that the citation counts of papers published in a given year are more likely to follow a log-normal distribution than a normal distribution. ↩