Attention Junction, what’s your function? To analyze the views of two Wikipedia pages, identify spans of public interest, find overlaps, and turn them into Google / Google News searches. All while being free to use and free of ads. Let me show you how it works.

The Basic Idea

I have written before about my vision of Wikipedia page view data as “fossilized attention.” By analyzing a Wikipedia article’s views over time, you can identify days of high public interest, which probably mean something newsworthy/noteworthy happened. I have used this idea with Wikipedia Seismograph and the Gossip Machine portion of MiniGladys, but those identify individual days.

I wanted to see if I could make something that identified spans of public interest in a Wikipedia page, as sustained interest is even more indicative than a single day of spiked page views. And once you can identify a span, why not analyze two pages at a time to find where the spans overlap? I found that this was a fascinating way to explore a big story with a lot of different players — by analyzing different pairs of people, you can find different time spans and different story angles around an event.

One of the reasons I made Attention Junction was because I wanted a way to measure public interest in the Jeffrey Epstein files that wasn’t a mainstream media outlet saying “Everybody cares!” and two days later announcing “Everybody stopped caring!” (Spoiler: based on Wikipedia page views, my conclusion is the public has NOT stopped caring.) But I don’t want to make you read about that for an whole article so let’s do Katy Perry and Jeff Bezos.

Using Attention Junction: Katy Perry and Jeff Bezos

Katy Perry and Jeff Bezos usually don’t have much in common unless you count the fact that they’ve probably lost more money in their sofa cushions than I will make in my entire life. That’s probably why Katy Perry was one of the celebrities who took a quick trip to the edge of the atmosphere and back last April.

To start using Attention Junction, enter two Wikipedia topics. In this case, Jeff Bezos and Katy Perry. Secondly, choose the time span to analyze — anywhere from last month to an all-time analysis. I try to keep the span I’m monitoring on the short side — last year and the last 2 years work well. I only use the “Last Month” for breaking news/ recent events, and the 5 years / All Time options if I’m trying to get all the tea on two topics. Even then I’ll use the “Last Year” and “Last 2 Years” option first.

Once you select a time range, the analysis runs automatically and shows you time spans where both topics had high levels of public interest and where they overlapped. It also gives you a few statistics about how much of entire search span was overlapped, and the average page views for both topics. Scroll down a little bit more and you’ll see the actual charts for each overlapping span. Let’s look at one for Perry and Bezos.

This chart shows each topic, the time span covering the arc of public interest, and a timeline indicating the level of public interest that day. “Peak” indicates a z-score of 1 relative to an average of page views. High is a z-score of .75, Medium is .50, Low is .25, and minimal is .10. Spans of interest always start with Peak interest and decay over time; an algorithm determines how long each state continues before the span terminates. Hitting another audience interest peak in the middle of the span (as Katy Perry does here) resets the span to continue.

In this case Katy Perry had a 21-day streak of public interest, though the last week’s decays pretty quickly. Bezos, on the other hand, had a 9-day streak of interest that terminated more solidly, probably because the entire Internet wasn’t taking the mick out of him. The two topics overlap for nine days — April 14 to April 22.

Now of course two topics being popular at the same time does not necessarily mean that there’s an overlap of public interest. That’s why there are four buttons underneath the chart. The first two let you to do searches (for the overlapped time span) of Katy Perry and Jeff Bezos on Google. The third button lets you search both people for the overlapped time span on Google, and the last one lets you do a time span search on Google News. In this case, let’s take a look at the Google News search:

As you might expect, the date-bounded results are all about the Blue Origin launch and the aftermath. See how focused they are? If you look at the 2nd public interest these two people shared over the last year, you’ll see the results are about something completely different — Jeff Bezos’ wedding.

A More Historical Example: Covid-19 and remote work

The Blue Origin flight happened this year, but Attention Junction is good for historical research too, going back to 2017. As an example, let’s try an all-time search for the topics “Covid-19” and “remote work.”

Much to nobody’s surprise there are plenty of times that those two topics overlapped. Let’s take a look at one of them:

This is one of the earliest overlap spikes. See how the Covid-19 spike starts on February 25 and the remote work spike starts on March 4? It’s fascinating to me to do a time-bounded search of these two topics because suddenly everybody was trying to figure out remote work at once.

As AI generates more and more slop, I feel more and more compelled to explore using the persistent metadata of time and place to build focused search spaces. Discovering the time element of public interest and turning it into a date-based Google search means that after-the-fact slop generators are going to have a lot harder time breaking into those search results. And that’s just fine with me.