Magazine

Vote for me, I know nothing

Image caption Should politicians don a metaphorical blindfold before making policy?

In the public imagination, knowledge is associated with wisdom. But in his regular column, Michael Blastland asks if ignorance is the new clever.

Would you vote for a politician whose pitch was ignorance? Here's why it might not be such a daft idea.

First, let's make it sound even dafter. Because after the politician admits to knowing nothing, her next step is to put on a blindfold.

This is how it works.

She wants to run the country. She needs some policies. Since she has no idea which ones are right, she experiments - little experiments first because they're easier. She takes a policy and tries it.

Because she has no idea of its effect, she needs one group of people to try the policy on and one group to leave alone.

Knowing nothing, she mustn't influence the result. Hence the blindfold. Ideally, she assigns a random bunch of people to each group without knowing which will face the policy. When it's all over, but before she knows which group was which, she decides who did best. Then she finds out if the policy worked. If it worked, she adopts it.

Hurrah, she has a policy.

Bit of a faff, isn't it? Better to vote for someone who knows exactly which policy is best and get on with it.

Except that in almost all areas of social research, the method of our mythical know-nothing politician, the method of the ignorant who say they don't know what works, is taking over. What she has done is an outline of what's known as arandomised controlled trial(RCT).

RCTs are best known in medicine. Now some people talk of a randomisation revolution in government.

It can sound odd in abstract. And I've stretched the point. So here's a true story that has been making waves in the past week or two, to show its use.

Dr Rachel Glennerster used to work in government.At a conference recentlyshe said: "It's quite scary when you start looking at things how little rigorous evidence there is behind a lot of conventional wisdoms."

She used the example of micro credit, the policy of making small loans available to women in poor areas. Millions of dollars, pounds and rupees were going into micro financewith strong claims about the benefits, she said.

Imagine a politician on the right who isn't ignorant, but knows the answers - naturally, business is what makes an economy grow and we must encourage entrepreneurs. Of course this will work.

Or on the left - women's empowerment is key to raising standards because women are committed to improving health care, education and the welfare of the family. Of course this will work.

And isn't it obvious? Women in a village who take up micro credit do better than women who don't.

This is true. But those women who take out micro credit are very different from those who don't. They are not a random selection - "as demonstrated," Glennerster says "by the fact that in a conservative society they stepped up and said I want to do something new." That is, those who took out a loan "were more pushy." The conclusion "women who were more entrepreneurial did better," was hardly surprising.

They needed a new experiment to randomly compare entrepreneurial women who did and didn't receive micro credit, rather than entrepreneurial women who received micro credit with un-entrepreneurial women who didn't. When they did this, there were "no reductions in poverty, no impact on women's decision-making, education, no more expenditure on health - so a lot of the things that are being claimed… we did not find." Micro credit was useful, but far from the miracle once believed.

The ignorant researcher will try a policy of micro credit, but will also experiment separately with others, and will make it a matter of principle to say that he doesn't know which works. And he will use the blindfold when he can to try to apply the policies randomly, rather than applying them to selective groups.

Often, it isn't anywhere near as simple as that. RCTs can take time and money. Good randomisation is harder than putting on a blindfold. Good ignorance takes brains. Measuring the results is often hard. There are some policy problems - global climate change is an example - where this kind of experiment is tricky. You'd need two planet earths. And some might defend all those things that inspire people to try to improve life according to particular values they already believe in, like ideology.

But there could be a bigger space for ignorance. Those who say they know, often don't. Those who know only that they're ignorant might set about learning in a rigorous way. Learning to be ignorant has been one of the most important discoveries of statistics, sometimes called the science of evidence and, contrary to its reputation, a discipline obsessed with how what we think we know can be wrong.

Is that a lesson politics can ever learn?

Of course, if the premise of all this - that ignorance is the new clever - is itself very knowing, does that make the whole idea stupid? Or is that just trying to be clever?

Related Internet links

The BBC is not responsible for the content of external Internet sites