How Well Can ChatGPT Analyze a Complex Map, Part 1?

This second-to-last post in the current series on ChatGPT examines the chatbot’s ability to find and explain patterns found on a relatively complex map. ChatGPT cannot (yet?) make maps, but can it intelligently analyze them? To test its abilities, I asked it to write an analytic essay on the geographical patterns found on a map of the 2020 U.S. presidential election with counties shaded by the margin of victory (a map of this sort is posted below). It managed to capture the basic contours reasonably well, focusing on regional divisions and the urban-rural split. But it missed many of the more subtle patterns and failed to notice any exceptions to the general rules that it had inferred. I therefore began to refine my queries, asking for more details about specific patterns. I have been told that this “narrowing in” strategy is needed to get the most out of ChatGBT. I was not impressed, however, with the results. If anything, the chatbot lost coherence when given additional information and more pointed requests.

Although I ran a number of these experiments, I will discuss the results of only two, one today and another in the next and final post in this series. The query that I used for today’s post specified that ChatGPT should pay attention to exceptions found for the general patterns that it had identified.

The chatbot’s response to this prompt is reproduced below, along with my comments on its errors and points of confusion. My comments are [IN BRACKETS AND IN ALL CAPS].

ChatGPT Response:

The 2020 United States presidential election was a historic event that brought about a complex tapestry of geographical patterns when analyzing the margin of victory across all U.S. counties. While certain general trends [I ASKED ABOUT PATTERNS, NOT TRENDS] emerged from the map, there were notable exceptions that provided a deeper understanding of the diverse political landscape in the country.

General Patterns:

When examining the map shaded by the margin of victory, several broad geographical patterns became evident:

    1. Coastal Strongholds for Democrats: Counties along the East and West Coasts, from Maine to Florida and Washington to California, predominantly voted for the Democratic candidate, Joe Biden, with significant margins of victory [THIS IS ESSENTIALLY UNTRUE. WITH JUST A FEW EXCEPTIONS, BIDEN ENJOYED A “SIGNIFICANT MARGIN OF VICTORY” IN COASTAL COUNTIES ONLY IN CALIFORNIA. FROM SOUTHERN DELAWARE TO SOUTHERN FLORIDA, MORE COASTAL COUNTIES VOTED FOR TRUMP THAN BIDEN (SEE THE FIRST MAP BELOW). AND WHAT ABOUT THE GULF COAST? IS IT NOT “COASTAL”?]. This trend aligns with the historical tendency for urban and coastal areas to lean left. [ONLY IF ONE HAS A VERY SHALLOW DEFINITION OF “HISTORICAL.” AND WHY DOES THE TERM “URBAN’ APPEAR IN A SECTION FOCUSED ON THE COAST/INTERIOR DIVISION?]
    2. Inland Republican Strongholds: In contrast, many [MOST, ACTUALLY] inland counties in the Midwest and South showed robust support for the Republican incumbent, Donald Trump [THIS IS TRUE NOT JUST IN THE MIDWEST AND SOUTH]. These regions often featured rural or less densely populated areas, which have traditionally leaned conservative. [MANY INLAND, NON-METROPOLITAN COUNTIES IN THE SOUTH SUPPORTED BIDEN, OFTEN BY A WIDE MARGIN. ALL HAVE BLACK MAJORITIES. THE CHATBOT COMPLETELY MISSED THIS IMPORTANT EXPLANATORY FACTOR.]
    3. Urban-Rural Divide: The urban-rural divide was pronounced, with densely populated urban centers and their surrounding suburban counties favoring Biden. On the other hand, rural counties consistently supported Trump, reflecting deeper cultural and demographic divisions.

Notable Exceptions:

While these general patterns were dominant, there were several counties that stood out as exceptions, challenging the conventional wisdom [THIS IS NOT A MATTER OF “CONVENTIONAL VERSUS UNCONVENTIONAL WISDOM,” BUT RATHER ONE OF “FINE-LEVEL VERSUS GROSS-LEVEL ANALYSIS”] of their respective regions:

    1. Urban Counties in Red States: In typically conservative states, such as Texas and Georgia, urban counties like Harris County (Houston) and Fulton County (Atlanta) experienced a significant shift toward the Democratic Party [HARRIS COUNTY, TX DID NOT SHOW A “SIGNIFICANT” BLUE SHIFT IN THIS ELECTION, AS CAN BE SEEN ON THE MAP POSTED BELOW].  These counties, characterized by their growing and diverse populations, flipped or came very close to flipping from Republican to Democratic [THERE WAS NO “FLIP” OR “NEAR-FLIP” INVOLVED! HARRIS COUNTY HAS BEEN HABITUALLY SUPPORTING DEMOCRATIC PRESIDENTIAL CANDIDATES FOR SOME TIME.]
    2. Rural Counties in Blue States: Conversely, some traditionally Democratic states like Minnesota and Oregon saw rural counties trending toward the Republican candidate [RURAL COUNTIES IN OREGON HAVE BEEN MOSTLY “RED” FOR SOME TIME. ALSO, MY QUERY DID NOT ASK FOR AN ANALYSIS OF ELECTORAL CHANGE]. For instance, in Minnesota, several northern and western counties that had previously supported Democrats shifted toward Trump [AS CAN SEEN ON THE MAP BELOW, SEVERAL NORTHERN AND WESTERN MINNESOTA COUNTIES SAW A “BLUE SHIFT” FROM 2016 TO 2020].
    3. Florida’s I-4 Corridor: Florida, a perennial swing state [NOT ‘PERENNIAL; ONLY FROM 1992 TO AROUND 2020], demonstrated a unique geographical pattern [WHAT IS  “UNIQUE” ABOUT IT?]. While the state as a whole leaned slightly toward Trump, the I-4 corridor, which stretches from Tampa Bay to Orlando, remained a closely contested battleground. This region, known for its political diversity and mix of urban and suburban areas [RURAL AREAS AS WELL], defied the state’s overall trend [A LOT OF ATTENTION HAS BEEN GIVEN TO THE MIXED AND CHANGING VOTING PATTERNS OF THE I-4 CORRIDOR; SEE THE SECOND MAP BELOW. BUT AT THE COUNTY LEVEL, THIS REGION IS NOT REALLY AN EXCEPTION TO FLORIDA’S GENERAL ELECTORAL PATTERNS, AND IT SAW ONLY MINOR ELECTORAL SHIFTS FROM 2016 TO 2020; SEE THE THIRD MAP BELOW].
    4. The Rural-Urban Split in Swing States: In swing states like Wisconsin, Michigan, and Pennsylvania, the margins of victory in urban counties, such as Milwaukee, Detroit, and Philadelphia [THERE IS NO “DETROIT COUNTY”], were crucial in securing these states for Biden. Conversely, the majority of rural counties in these states remained staunchly Republican. The tension between urban centers and rural areas was starkly visible in these key battleground states.
    5. Native American Influence: In Arizona and New Mexico, Native American reservations and communities played a significant role in the election. Several counties with sizable Native American populations, such as Apache County in Arizona, favored Biden. This underscores the importance of demographic diversity and tribal influence in these southwestern states [YES, BUT WHAT ABOUT NATIVE AMERICAN COMMUNITIES IN OTHER STATES, PARTICUARLY MONTANA AND SOUTH DAKOTA?].
    6. Maine’s Split Decision: Maine stood out with its unique [NOT UNIQUE; NEBRASKA HAS THE SAME SYSTEM] allocation of electoral votes. While the state as a whole went for Biden, it was one of only two states (alongside Nebraska) [OK, BUT THIS STATEMENT CONTRADICTS  “UNIQUE’ IN THE PREVIOUS SENTENCE] to split its electoral votes by congressional district. Trump managed to secure one of Maine’s congressional districts, emphasizing the granularity of political preferences even within a single state. [“EVEN” WITHIN A SINGLE STATE? THERE IS TREMENDOUS “GRANULARITY” IN REGIONAL VOTING PATTERNS IN EVERY STATE. BESIDES THAT, MY QUERY WAS ABOUT VOTING PATTERNS AT THE COUNTY LEVEL, NOT ABOUT ELECTORAL VOTES]

In conclusion, the geographical patterns found in the 2020 U.S. presidential election, as indicated by the margin of victory across counties, painted a diverse and multifaceted picture of American politics. While overarching trends pointed to the urban-rural divide and regional tendencies, notable exceptions illuminated the complexity of the electorate. These exceptions often hinged on factors like changing demographics, local economic conditions, and unique historical and cultural influences [LOCAL ECONOMIC CONDITIONS WERE NOT MENTIONED IN THE ESSAY], illustrating the rich tapestry of American democracy. Understanding these exceptions is crucial for a comprehensive analysis of the political landscape in the United States.

How Well Can ChatGPT Analyze a Complex Map, Part 1? Read More »