spectator.org
How Think Tanks Should Use AI … And How They Shouldn’t
“What is a think tank?” is a question I get asked all the time.
The job of a think tank is to generate the ideas that politicians implement in office. If politics were the movie industry, those we elect would be the stars on the big screen; think tanks help write the script.
“But don’t politicians have ideas of their own?” I hear you ask.
Of course they do (though having spent 12 years in the House of Commons before running a think tank, I’d suggest they have them less often than you might imagine). Even with a clear goal — cutting taxes, improving childcare, encouraging growth — elected leaders often need help with the detail of how to make it happen.
Free-market think tanks have been enormously influential. Think tanks — or at least the free-market version of them — were more or less invented by an Englishman called Antony Fisher. He founded, among other institutions, the Institute of Economic Affairs in Britain in 1955, the Fraser Institute in Vancouver in 1974, the Manhattan Institute for Policy Research in New York in 1977, and the Atlas Network in 1981.
Fisher was famously advised by Friedrich Hayek that, instead of seeking elected office in the House of Commons, he would have a greater impact if he shifted public opinion so that even the wrong people had to do the right thing.
It was think tanks — Fisher’s own IEA prominent among them — that made the case for privatization and helped Britain roll back the state (remember when Britain used to do that?). They drew up the blueprints for Reaganomics and Thatcherism, drove the sweeping welfare reforms of 1996, popularized school choice — now spreading across America — and powered the state-by-state push to abolish the income tax (Americans for Tax Reform, in particular).
Several decades on, the liberty movement has built a myriad of institutions — from giants like Heritage and AEI to state-based minnows such as MCPP in every U.S. state. The left has, of course, followed suit with its own think tanks (although progressive think tanks strike me as comparatively inert — perhaps because the left already owns so much of academia).
Given that ideas can change the world, I’ve never understood why some think-tankers feel the need to brand their organisations “do tanks” rather than think tanks. Done well, ideas possess such extraordinary agency that there is never any need to half-apologize for the suggestion that thinking alone is not enough. (If no one is reading the ideas in your research papers, it might be because the research papers are the problem, not the ideas.)
To be effective, a think tank, like a good journalist, should be friendly toward elected officials without becoming their friend. Never join the political club in the belief that doing so will buy you influence. Being friends with everyone might get you to the bill-signing ceremony; pushing hard is what ensures the bill being signed actually advances liberty.
Having spent over a decade in the House of Commons, I think it is rare for a politician to adopt the right policies because they see the light. More often it is because they feel the heat. A policy is taken up when it is presented in the right place at the right time and offers elected leaders a solution to a pressing problem.
Unless a think tank has a clear sense of mission, there is always a risk it begins to claim credit for things that would have happened anyway. Like a shaman performing a rain dance and then pretending to have produced the precipitation, it simply looks daft.
In a world where a handful of people with laptops can do extraordinary things, the size of your payroll is no longer a reliable proxy for output — if it ever was.
Alongside Fisher, few have done more to shape how think tanks work than Joe Overton. His great insight was that, at any given moment, there exists a range of politically acceptable policy options. Those we elect operate within that window of acceptable opinion — the Overton window — and think tanks should aim to shift it: making the once-unthinkable debatable, then sensible, and finally what everyone insists they always believed.
Over the past two decades, the digital revolution has made Joe Overton’s insights even more salient, as the established media has lost ground and think tanks have gained the means to reach people directly. Think tanks adapted to the advent of the internet, but that change will be as nothing compared with the transformation we are about to see with AI.
Consider, for a moment, policy research — a vital part of what a think tank does (in my view, a think tank that produces no original research is not thinking, but I digress). Next month, here at MCPP, we will publish our annual Fat Cat report — a list of the highest-paid public officials in Mississippi. AI’s significance has little to do with using it to help write the report (we write it; AI polishes and edits it). Nor does its real importance lie in turning what was once a dry PDF printout into an interactive web tool (if no one is reading your research-paper PDF, take note).
What truly changes is that, once a report like our annual Fat Cat has been published, AI allows us to transform it into a self-regenerating report that “lives” online. The 2027 version will almost literally update itself as new data becomes available.
Two decades ago, when most people listened to music, they used CDs. It wasn’t that there was a shortage of digital music plastered all over the internet by the early noughties. It was just that you needed the technical know-how to find it (or a willingness to break copyright law). Then along came Spotify and others, and that sea of digital music could suddenly be accessed by everyone.
That is what is happening in the policy world. There is an ocean of digital data placed on servers by public bodies every day, but until now you needed to be a spreadsheet obsessive to make use of it.
One of our jobs as think tanks is to build the policy equivalent of an MP3-player interface — like the OpenData tools we are creating at MCPP — so the public can finally see what has long been there but remained inaccessible.
Will AI transform think tanks? I’m certain of it. In a world where a handful of people with laptops can do extraordinary things, the size of your payroll is no longer a reliable proxy for output — if it ever was. Done competently, the bang for the donor’s buck will be extraordinary.
AI is going to allow think tanks to reaffirm their founders’ original vision. After decades of growth, it is only natural that some institutions may have drifted a little from that vision; the day-to-day pressures of running a large organization can take on a logic of their own. Perhaps if just occasionally the metrics some folk report have something of the old Soviet tractor production targets about them, now might be the time to change.
Freed from the need to grow simply for the sake of growth (or one might say the need to be seen to be growing for the sake of growth), we can stay sharply focused on what has always mattered most: generating the ideas that shift the Overton window and advance liberty.
The tools are here. The data is waiting. In a world where we can act almost as fast as we can think, the opportunities for idea-rich think tanks are without limit. The only constraint is the quality of our thinking — and for think tanks, that ought not to be much of a problem.
READ MORE from Douglas Carswell:
More Money Won’t Fix Our Schools. Our Data Proves It
Great Nations Aren’t Destroyed by Enemies. They’re Destroyed by Debt.
Y’all Street Is Eating Wall Street’s Lunch
Douglas Carswell is the President & CEO of the Mississippi Center for Public Policy.
New Posts (Home)
ChatBox
Popular Posts 
Game Zone
Top PodCasts
