May 8th 2019
A few years back, the President of AAAI (the American AI Association) and VP for Research at Microsoft (same guy) got a lot of press for figuring out who might have cancer from patterns of search queries. Some might have thought that a bit too big brother, including me. But I also knew this MD-PhD from Stanford days, in the late-80s, and knew that his motives were medical, to save lives, not just look for ad-targeting opportunities (ahem, other search engine companies).
I don’t remember at the time whether he also did geographical analytics, looking for higher rates in specific locales. My university colleagues recently got regional cancer reporting data and found counties that had statistical anomalies. One never knows whether that’s aggressive reporting or nearby toxic waste dump. Still, geographical identification of potential hot spots, for public health purposes, seems even lower hanging fruit than diagnosis-by-bing.
In fact, one of the better authors I met at an IEEE health informatics conference a couple of years ago was doing just that — using big data geo-visualization to give early warning of outbreaks. I see she’s now VP of analytics at Goldman Sachs in NYC. This is why talent is hard to keep on civic problems.
PeakMetrics can do the same thing by looking at media mentions. It’s got the same causal ambiguities as HIPAA de-identified reports, and the same indirect inference problems that search-query-to-cancer-diagnosis had. But it’s incredibly easy to do the query and pose the hypotheticals.
Who’s been talking about measles recently? As usual, it’s important to contrast the mentions-map with a control query, lest one be confused by the fact that media outlets are concentrated in particular locales. Cities.
So let’s contrast the measles mentions with John Hickenlooper mentions. I could expound on why that’s not a bad choice, but if the apparent arbitrariness is amusing, fine.
Here is the East Coast, measles on the left, Hickenlooper on the right. One can see measles mentioned from Iowa to North Carolina through Kentucky and Tennessee (specifically Nashville, Chattanooga, Knoxville, Brevard), and across Northern Indiana (South Bend, Fort Wayne) up to Detroit, in greater amounts. Strong mentions in the Boston-NYC-DC corridor, though that could just be amplification of the big media locations. One also sees contrasting interest in Dallas, Tulsa, Wichita, and Omaha, nearly a vertical line through the center of the country.
And here is the West Coast. Seattle and Arizona cities are lit, as are LA and San Francisco especially.
This could of course be an artifact of local media being vigilant rather than responsive, but the two tend to go together.
What’s really interesting is that we have a large mix of left politics, right politics, and swing-state politics here. Perhaps measles, and public health generally, are unifying or at least non-partisan worries.
Here is the PeakMetrics Cloud for /measles/, which shows greater left-media phrasing despite the mentions concentrated in some red states. The cities in red states do have a lot of blue and center-blue reporting after all.
The screen shots don’t show Florida, but the cities lit by measles mentions there are St. Petersburg, Jacksonville, Miami, and Orlando.