Everyone loves a good map
Several weeks ago, I shared new data about the red state / blue state AI divide, where people in red states are more negative about AI than those in blue states. The case study analyzed LinkedIn posts from 100 AI influencers: people who have 50K+ followers and are posting about AI at least 2x/month.
This week, I augmented that data with posts from an additional 250 AI influencers. I wanted to go beyond sentiment to see what people in all 50 states are actually talking about when they discuss AI.
To be honest, I was expecting I could share a fun map where everyone would be like awww, that's my home state! Like, see how no one talks AI and food more than people in Louisiana? Look at how New Yorkers talk AI and finance! That kind of thing.
I did get a cool map, but it was not the map I was expecting.
Using AI to study AI, that's me
To start with, I had to cluster everyone's posts by topic. That's right, I had to use legit technical skills for this one. I dusted off my old topic modeler, tossed in everyone's posts, and found the following categories:
- Politics, policy, and government: Posts about AI and voting, democracy, legal policy, and elections. As we approach the US presidential election, these topics are showing up more and more often.
- Privacy, bias, and ethics: I expected to find that privacy and ethics would be different topic categories, but nope. They co-occur so often that all my programmatic topic clustering grouped them together. Machines 1, Kieran's intuition 0.
- AI in specific industries: Posts discussing how AI is applied to specific industries: AI in mining, AI in agriculture, AI in art, and so on. I grouped these all together for this analysis.
- Education and schools. Among industry-specific topics, AI in education was in so many posts that it made sense to mark it as its own category. So I did.
- General: Posts about topics like machine learning, ChatGPT, data, or technology. These are broad AI topics and co-occur in LinkedIn posts with many of the other topics above.
- Other stuff. There's a long tail of other topics, but the long tail doesn't show up as often as you might think. 89% of the posts I analyzed don't include any topics outside the categories above.
You can see how often topics show up across the whole data set below. The totals add up to more than 100% because an individual post often includes more than one topic. For instance, a single post might be about AI, education, and kids' privacy. Another post might focus on AI and election bias.
From there, I set out to find, for each state, which topics show up more than expected. In other words, using the average frequencies above, I looked for the topic that was the most statistically overrepresented in posts from people in the state. I wanted to build a map showing the most overrepresented topic in every state.
Cultural geography is amazing, y'all
In my red state / blue state article about people's overall AI sentiment, I found differences based on region, so I was betting on finding some regional differences here too. But the regional clustering that I found stunned me. The map below shows, for each state, the topic that is most overrepresented in their social posts about AI.
Look at all those regional clusters of colors! This data shows that even in the digital world, what people focus on is highly regionalized. Who's doubling down on privacy, bias, and ethics? New England and the West Coast. Who's focused in on AI and education? The mid-Atlantic and the Great Lakes. Who's most worried about AI and politics? Lots of the South.
The clustering in the map reminded me of Colin Woodard's American Nations framework. Woodard argues that rather than having one universal “American” culture, we have 11 distinct regional cultures that are often at odds with each other . Regional cultural differences show up in politics, values, language, religion -- and apparently in people's focus within AI.
What do you think?
Thanks for reading!
Kieran
Catch up on nerd processor case studies | Subscribe | nerdprocessor.com