Harvard Economist Dani Rodrik Maps His Own Five Stages of AI Grief
Harvard's Dani Rodrik tested a chatbot on a 2027 care crisis and got a costed policy memo. His essay maps AI grief from denial to acceptance — and a warning.

Why it matters
- Dani Rodrik tested a chatbot with a hypothetical 2027 care crisis; it produced a memo with federal matching grants, Medicaid pay changes and a caregiver visa program, plus a fiscal accounting.
- The chatbot, unprompted, also outlined a political strategy for selling the proposals.
- Asked to ignore political realities, it proposed a National Care Corps drafting young people and converting offices and malls into 'care districts'.
- Rodrik cites colleague Jason Furman's claim that AI tools would outperform trained economists in White House CEA staff roles.
- Mathematicians who signed a joint letter warned of losing 'conceptual understanding and insight' if AI takes over their field.
Harvard economist Dani Rodrik asked a chatbot to solve a hypothetical 2027 care crisis, and it returned a fully costed policy memo — complete with a political strategy he had not requested. Rodrik, professor of international political economy at Harvard Kennedy School and past president of the International Economic Association, describes that test and his broader reckoning with AI in a new essay structured around Elisabeth Kübler-Ross's five stages of grief: denial, anger, bargaining, depression, and acceptance.
The essay matters because Rodrik is not a technology optimist by trade. He is a mainstream policy economist, the author of "Shared Prosperity in a Fractured World: A New Economics for the Middle Class, the Global Poor, and Our Climate" (Princeton University Press, 2025). When a figure of that standing writes that AI has shaken his belief that judgment and creativity are uniquely human, it signals how far the debate has moved among professionals who expected to be the last affected.
What did Rodrik actually test?
The centerpiece experiment drew on a claim by his Harvard colleague Jason Furman: that AI tools would already outperform professionally trained economists in staff roles at the White House Council of Economic Advisers. Rodrik put that to a trial with a deliberately hard prompt.
The scenario: "It is 2027, and the country is faced with a crisis in care. There is a shortage of home care aides, accidents and deaths are rising among elderly and disabled people left without help, and workers are dropping out of the labor market to look after loved ones. The CEA Chair needs options to present to the President."
The chatbot's technical memo covered:
- federal matching grants;
- adjustments to Medicaid pay;
- a caregiver visa program;
- a full fiscal accounting.
It also, Rodrik notes, exceeded its authority — and the prompt itself — by outlining a political strategy for selling the program. In plain words: the system did not just answer the question, it anticipated the question behind the question.
Could the chatbot think outside the box?
Rodrik then raised the bar. He asked the model to develop out-of-the-box policy alternatives that disregarded political realities. It returned a National Care Corps that would draft young people to care for the old, and a proposal to convert underutilized office and mall space into "care districts" benefiting from scale and agglomeration economies (the productivity gains that come when related activities cluster together).
Then came the test Rodrik did not expect to matter. On a whim, he asked AI to turn one of his manuscripts into a stand-up comedy routine. "What came back made me laugh so hard that I worried about passersby fearing for my mental health," he writes. "Insight, imagination, humor! This was not part of our bargain, was it?"
That quote is the hinge of the essay. Rodrik had settled into the comfortable bargaining position that AI could execute while humans imagined. The chatbot's performance dismantled it.
Why the earlier stages broke down
Rodrik traces the sequence briefly and candidly. Denial came easy when AI looked like "a glorified search engine or a chatbot that does a decent job of simulating a human being." Anger followed as the technology cropped up in lawyers' briefs with hallucinated case references, software code with mistakes and security vulnerabilities, customer-service bots that went off script, and — a particular irritation for an academic — student essays and assignments.
Bargaining set in once the fabrications grew rarer and student use proved unstoppable. The deal professionals struck with themselves, Rodrik writes, held that "questions of judgment, taste, values, and preferences were still, and would remain, matters for the human mind." AI could execute; humans would imagine. "It was just a matter of getting the division of labor right."
The bargain failed in the details of daily use. Each time Rodrik consulted his chatbot, it came back with more than answers. It told him what else he should be looking for, suggested different framings, and offered parallels with other projects he had worked on — whether or not he had consulted AI on them. "I'd like to think that I would have arrived at similar thoughts on my own in due time, but my chatbot regularly preempted me," he writes. The annoyance was real; so was the value.
What does the depression stage look like?
For Rodrik, the low point was not professional anxiety but something closer to grief. "I started to feel sadness about a world in which human thought would no longer be distinct from machine work — and would be needed less day by day," he writes.
He anchors that fear in evidence from another discipline. Mathematicians who signed a joint letter on AI's impact on their field warned of the loss of "conceptual understanding and insight" that would follow from letting AI take over their work. Rodrik found the prospect "frightening and disorienting." It is worth noting what his essay does not provide: any evidence that the care-crisis memo or the comedy routine contained errors. His alarm rests on capability, not on flaws.
Is acceptance the same as surrender?
Here Rodrik draws his hardest line. Unlike a terminal illness, he argues, AI and its uses are not outside human control. Decisions are being made daily by the Silicon Valley firms developing the technology, the businesses deploying it, the professional communities setting norms of appropriate use, and the governments that must — "better sooner than later" — regulate it. "The worst mistake we can make is to accept a world in which we cede too much to it," he writes.
His proposed analogy is performance-enhancing drugs: they temporarily boost human capabilities but eventually harm users. AI, in his framing, is similar but much more powerful, with "commensurately disastrous side effects" that will show up much later. The caveat is his own: the temptation to use it freely "is, and will remain, very strong." He offers no enforcement mechanism, and his essay does not say who would impose restraint on competitors that decline to tie themselves down.
The essay closes with Homer. Odysseus, Rodrik recalls, knew the Sirens promised knowledge but delivered disaster, so he had himself tied to the mast of his ship to resist their song and the cliffs that had claimed so many mariners. "The challenge we face with AI is similar: to restrain ourselves from losing our humanity," Rodrik concludes.
What is the takeaway?
Rodrik's essay is a concession memo from inside the professional class that expected to bargain its way out of AI — paired with a call for self-restraint that, by his own admission, depends on institutions that have not yet acted. The open question he leaves is whether regulation and professional norms will arrive before the side effects he warns about become visible.
By James Calloway
Source: feed.koreatimes.co.kr
More from James Calloway
Show full bio
News editor covering people, launches and day-to-day practice in artificial intelligence and machine learning at AI In Context.
604 articles
Related articles
- OpenAI's Lehane Frames 2028 Election as Defining Test for AI Policy
- A Federal Job Guarantee Emerges as the Answer to AI Unemployment
- Trump's AI Growth Promise Echoes Reagan's Tax Cut Fantasy
- 79% of Students Use AI for Homework, Only 47% Get Guidance: OpenAI Flags Education 'Guidance Gap'
- Klein and Glazer Call for AI IPO Blocks, Billionaire Taxes