Behind the Scenes Q&As

Slurp on this: What Axios’ Soup Dumpling Index tells us about AI in the newsroom

The Big Mac Index has been around for years for people to intuitively compare and contrast prices at their favorite chain restaurants. The opening of Din Tai Fung in New York City inspired Felix Salmon, Sarah Grillo and Danielle Alberti from Axios to create a similar index using Din Tai Fung dumplings. They reported on their findings in the article “Soup Dumpling Index: How prices compare around the world.” However, the project came with its challenges. We spoke with Danielle Alberti, data visualization editor at Axios, about the obstacles they faced during data gathering and on the increasingly important role of AI in this process — a key question on everyone’s mind given the rise of AI technology.

The following interview has been edited for clarity and length. 

What drew you to this story? More specifically, why Din Tai Fung?

New York-based Felix Salmon brought Sarah Grillo and me onto the project — three food enthusiasts eager to explore this topic. In a food-centric city like New York, the opening of a Din Tai Fung is sure to bring its fair share of comments, critiques and attitudes. From big fans to the haters, everyone will have an opinion. Spoiler alert: the restaurant is already booked out. In a place where everyone seems to have their own favorite little hole-in-the-wall spot for soup dumplings, it’s fascinating to see how a high-profile chain like Din Tai Fung will be received.

How did you gather data? 

The team had to get creative as most Din Tai Fung locations do not list prices on their online menus. We started by gathering data was to identify 20 geographically representative locations using a map of where the restaurants are. We picked the most influential Din Tai Fung locations around the world.

From there, we had to rely on unconventional methods, such as combing through Yelp and Google reviews and zooming in on photos of menus. The use of menu photos added complexity, requiring us to check for the most recent updates due to price fluctuations. Some locations, such as Shanghai and Beijing, proved much harder to find prices. In the end, we resorted to calling restaurants directly to inquire about prices and portion sizes. 

DON’T MISS  How the Center for Collaborative Investigative Journalism is reporting on water access around the world

Online ordering platforms led to issues with discrepancies or inflated prices, seen in cases like the Philippines. We had cross check every data point to ensure that our findings were accurate and up-to-date.

Given your tough experience in collecting data, what do you think is the future of data collection? 

Artificial intelligence could automate aspects of data collection, such as scraping information or interpreting text from images. This will reduce some of the manual effort involved, however,  AI still requires significant human oversight, and there’s uncertainty about whether the time investment would be worth it. Automation speeds up the process, but truly adds value by freeing up humans to analyze the data and draw conclusions.

However, it is important to note that building and maintaining AI systems takes time and effort. We still need human effort to find images and generate prompts for AI to follow. I do still think that despite this effort, it could be beneficial to start incorporating AI into the work we do. It is most important, though, that we find the right balance between automation and human input. 

Axios has begun to implement some use of AI into their journalism workflows.

Is Axios implementing AI into your journalism work? If so, how? 

At the moment, Axios has begun to implement some use of AI into our journalism workflows. We use AI for very simple actions. For example, an internally developed charting tool to generate line charts, bar charts, and tables. However, more beneficially, we have started using AI to automate alt texts for charts – a task that reporters often overlook. Now, at Axios AI automatically generates the alt text, with reporters reviewing it for accuracy. We update the prompts and record them in a spreadsheet if adjustments are needed.

DON’T MISS  Six fascinating projects from the 2019 Computation + Journalism Symposium in Miami

This process streamlines tasks and improves the quality and accessibility of content while fostering a more efficient workplace. We at Axios also believe it represents an ethically responsible way to use AI. Although Axios has been experimenting with other AI applications, such as for SEO purposes or in ways similar to how other news reporting sites use AI, nothing else has been fully integrated yet.

Back to the story, how was the pricing measurement determined? Is there a correlation between the number of Din Tai Fung locations in a city and the price of the pork xiao long bao? 

We chose $10 as the measurement because we were aiming for consistency, given that serving sizes vary across locations. In the U.S., for example, a serving of pork xiao long bao typically includes 10 dumplings, whereas in Tokyo a serving only includes six dumplings. It was easier to calculate and work with unit prices. We decided to stick with the $10 benchmark to make it easier to convert currencies to US dollars. 

In my opinion, there is no direct correlation between the number of Din Tai Fung locations in a city and the price of dumplings. Instead, prices largely reflect the cost of living in each city. London ranks the most expensive location, known for its high living costs, followed by New York and Los Angeles. This mirrors the premise of the Big Mac Index, where pricier cities tend to have higher food prices across the board.

Leave a Reply

Your email address will not be published. Required fields are marked *

Get the latest from Storybench

Keep up with tutorials, behind-the-scenes interviews and more.