A Signaling Problem in the Age of Artificial Intelligence
Why rising graduate unemployment and artificial intelligence both point to the same structural shift
Recent data from the Federal Reserve Bank of New York show that unemployment among recent college graduates has risen above the overall unemployment rate — an unusual reversal that signals more than cyclical weakness.
And decades of data from the Federal Reserve Bank of St. Louis indicates the economy is moving toward objectives-driven work, where value is created by delivering outcomes under constraint — framing problems, making trade-offs, coordinating across functions, and defending decisions with incomplete information.
At the same time, artificial intelligence is automating routine cognitive tasks. That does not lower expectations. It raises them. When machines handle execution, the premium shifts to judgment.
This creates a signaling problem.
Because artificial intelligence can generate polished memos, models, or slide decks, final products are no longer reliable indicators of mastery. Understanding the process is especially critical: How was the problem framed? What assumptions were made? What trade-offs were considered? What are the limitations of the data, of the approach, of the conclusions?
Recent research from Matt Sigelman and the The Burning Glass Institute and aiEDU reinforces this pattern, showing that artificial intelligence is reshaping the cognitive demands within disciplines rather than eliminating them.
The labor market rewards judgment, not simply execution.
While it is critical for K12 and higher education to incorporate artificial intelligence skills into the education process, the more import focus is developing judgement and the ability to recognize and lean into limitations.