Lightnews — Scholar-powered news

Arpit Narechania @arpitnarechania.bsky.social · 1d

Thrilled to share Exploropleth - our new #GIS #dataviz tool.

💡 @Mapmakers, you can simultaneously explore & compare 16+ data binning methods on a choropleth map, directly in your browser!

🎓 @Instructors, you can use the tool to teach how choropleth maps can mislead people!

exploropleth.github.io

Screenshot of a new, open source, web-based, geospatial visualization tool, Exploropleth, that lets users interact with a catalog of established data binning methods, and subsequently compare, customize, and export custom maps. The screenshot shows small multiples of 16 choropleth maps each showing Life Expectancy data for US counties, but using distinct data binning methods (e.g., natural breaks, standard deviation, pretty breaks, and so on).

7 10

Arpit Narechania @arpitnarechania.bsky.social · 1d

Thrilled to share Exploropleth - our new #GIS #dataviz tool.

💡 @Mapmakers, you can simultaneously explore & compare 16+ data binning methods on a choropleth map, directly in your browser!

🎓 @Instructors, you can use the tool to teach how choropleth maps can mislead people!

exploropleth.github.io

clio @cl10.bsky.social · 2d

Hi data people, check out Exploropleth! exploropleth.github.io/exploropleth/. It's free, open source software that shows how binning methods affect maps. There's also a teaching worksheet and a journal article in CaGIS: exploropleth.github.io
@arpitnarechania.bsky.social &
@alexendert.bsky.social

Different maps that show life expectancy by county in the U.S., where the data is divided using different data binning methods.

Box plot style graphics show data points for adult percent obesity by county in the U.S.

Arpit Narechania @arpitnarechania.bsky.social · Apr 3

🙏 Big thanks to my coauthors: Alex Endert from the Georgia Tech Visualization Lab & Atanu R Sinha from Adobe Research.

📄 Paper: arxiv.org/abs/2502.00682
📹 Video Overview: youtu.be/ASgLVAljtHk
🎙️ Recorded Talk: youtu.be/eh1WVYl25Dk

(6/6)

Screenshot of the first page of the research article showing its title, authors, abstract, and part of the introduction.

Arpit Narechania @arpitnarechania.bsky.social · Apr 3

🤔 Interestingly, participants in the 🤖AI condition reported both higher post-task benefit & regret.

This highlights the importance of understanding how different guidance sources impact user behavior (espec. AI), which can help design more effective guidance systems!

(5/6)

1

Arpit Narechania @arpitnarechania.bsky.social · Apr 3

🥁 The result? Guidance Source Matters!

Participants utilized guidance differently during analysis, including expressing varying levels of regret, even when its quality was constant!

E.g., those receiving guidance selected a lot more attributes than those without it!

(4/6)

1

Arpit Narechania @arpitnarechania.bsky.social · Apr 3

💡 Participants could request guidance from their assigned guidance source, in the form of attribute recommendations, one by one, up to 10 times.

We held the quality of guidance constant across all sources (7 relevant & 3 irrelevant recommendations, randomly sequenced)

(3/6)

1

Arpit Narechania @arpitnarechania.bsky.social · Apr 3

🎯 We conducted a study with 5 conditions:

1. 🤖 AI
2. 🧙 Human Expert
3. 🧑‍🤝‍🧑 Group of Analysts
4. 💡 Unattributed Guidance (no source attribution)
5. ❌ No Guidance (control)

We tasked participants to analyze a dataset, select relevant attributes & make #dataviz for a biz report.

(2/6)

1

Arpit Narechania @arpitnarechania.bsky.social · Apr 3

❓Do people’s perception & utilization of guidance change during analysis if they know its source, e.g., if it came from 🤖AI, 🧙Human Expert, or 🧑‍🤝‍🧑Group of Human Analysts?

✨Thrilled to share our new #iui2025 paper, presented last week at ACM IUI in Cagliari, Italy.

(1/6)

Teaser figure showing our study prototype for visual data preparation and analysis, with a Data Attributes View on the left (showing the list of data attributes of an uploaded dataset), a Data Records View in the center (showing summary statistics and individual records for an attribute), and a Guidance Panel to request guidance (in the form of data attribute recommendations to shortlist) on the right.

1 1

Arpit Narechania @arpitnarechania.bsky.social · Jan 19

Find more information about my work on my personal website at narechania.com

Find more information about the HKUST #CSE PhD program at cse.hkust.edu.hk/pg/admissions/

[2/2]

Arpit Narechania

Hi, I am Arpit Narechania. I am an Assistant Professor in the Computer Science and Engineering department at The Hong Kong University of Science and Technology (HKUST). I received my Ph.D. from Georgi...

narechania.com

1 1

Arpit Narechania @arpitnarechania.bsky.social · Jan 19

Are you looking for #PhD student positions in data visualization, visual analytics, #HCI research?

I am recruiting multiple #PhD students (fully funded) to join my group at The Hong Kong University of Science and Technology #HKUST in Fall 2025.

If interested, EMAIL me! Reposts appreciated 🙏 [1/2]

1 3 10