The Ditchley Lecture, as discussed in the previous post, has become something of an obsession as it likely determines my working future (and perhaps the future of Government policy, and hence the country).
The word "risk" is used nine times, but only once in the sense of something to be understood and avoided. And there it's financial risk which, we are told -- because of technology -- is now "more effectively hedged". (If you believe that...)
The word "uncertain" is not used at all.
And herein lies one fatal flaw, which I was grasping towards in my previous post. That the problems we face are problems of uncertainty and risk. A failure to properly understand financial risk caused the financial crisis. A failure to understand pandemic risk caused the current one -- if we had understood the risk we wouldn't have done stupid things. A failure to recognise that climate change is inherently uncertain -- and the uncertainty can lead to very bad, low probability outcomes -- and use a risk management framework to address it, will lead to consequences that far exceed covid-19.
(To be fair, there's a gesture towards this issue in "we need to understand that complex, adaptive systems demand respectful attention, not glib assertions of mastery" but that runs entirely contrary to the rest of the lecture, and the actions of the Government)
The second fatal flaw is that the speech is peppered with references to the public, citizens and society. Yet they are absolutely passive actors. Government will develop the solutions, the public will be lifted up by them. "To help the Forgotten Man" you do not, it seems, need to talk to the Forgotten Man. The Politburo decides what he wants and instructs the machine to give it to him.
It's a remarkably Stalinist approach, as has already been pointed out. There is nothing Conservative in this speech whatsoever.
We will move elements of the Party Apparatus to Teeside. We will purge the the Apparatus of Mensheviks -- the Party has already been purged of Gaukists, of course. But moving to Teeside is not moving closer to the public. The Party will decide what is good for you, and the Apparatus will deliver it.
You are no longer Forgotten. But no one is listening to you. You exist to justify our policies, and to vote for us. That is all.
Saturday, 4 July 2020
Wednesday, 1 July 2020
A distraction
we interrupt the regular service to provide you with a message about contemporary (UK) politics
For obvious reasons, this year's Ditchley Annual Lecture, given by The Chancellor of the Duchy of Lancaster, is worthy of both comment and analysis.
Superficially it is hard to disagree with the general thrust of the argument, that Government should better reward risk taking, that the Civil Service should have more diversity of thought, that centralising decision-making in London means that the cultural outlook of the majority decision-makers does not reflect the wider cultural outlook of the country, and that the Government is piss-poor at measuring success, and should do better.
But anyone with even a passing interest in UK politics over the last 5 years will have noticed that our politics is now entirely tribal. We are just as split between individuals who identify as pro and anti Brexit as the US is between those who identify as pro and anti Trump. A significant number of former comrades find themselves stuck behind enemy lines - more on the right than the left, where there are plenty of Conservatives aghast at what the current leaders of their nations are doing on trade, security, fiscal policy and common decency yet are unable to align themselves with the Tories/Dems - but on the left, too, though these types are perhaps more used to being out of step with the majority of 'their' tribe. Pro-protectionist anti-immigrant leftists exist, and have existed forever, but are very much in minority, and have been for years.
The lecture is aimed solely at one tribe, and blames their woes solely on a foe with all the realism of Mantel's Thomas Cromwell -- i.e. there are grains of truth, in here, rooted in documented historical fact, but the overall concoction is entirely fictional -- or perhaps some mythical version of Clive of India as a Hero of the Empire. If I were to ask you what went wrong in 2008, and has failed to go right since, I doubt you'd point the finger of blame at the Civil Service. Maybe you would. Maybe I'm too close. But I suspect the voices that are now claiming the Civil Service is ideologically attached to Europe would, 30 years ago, have claimed that the Civil Service was too slow to adopt to the benefits offered by globalisation. Where politicians lead, Civil Servants follow. The single market was a political creation and one which diminished the powers of Whitehall. It wasn't dreamed up by Mandarins nor do I imagine they were terribly keen on it in the early days, given that power flowed away from them. And we all know that austerity was a political, not technocratic choice. Officials seldom advocate policies which will see their Departmental budgets decimated, or as was the case for several Departments, far worse than decimated.
But I'm going on too long. Many many columns will be written about the impending re-organisation of Whitehall. Here are some others.
But I would like to point out what I think is a flaw at the heart of the argument that we need . It's about Maths and Science. As a physical scientist, I completely reject the assertion that decision-making should be "less reliant on those with social science qualifications and more welcoming to those with physical science and mathematical qualifications".
As a citizen rather than a Civil Servant I've seen policies come and go. Most policies, in my observation, fail or a best offer extremely poor value for money compared to readily available alternatives. Some crash and burn spectacularly, like privatization of probation services, or the Green Deal. Most chunter along, eating money and not doing much active harm or good. Occasionally we hit a great one (in my field, both the boiler regulations, which drove down UK gas use, or Contracts-for-Difference which, in the wind sector at least, have driven renewable electricity prices down beyond our wildest dreams.
But yeah. Policies usually suck. And they usually suck because the real numbers bear no resemblance to the forecast numbers.
So, more numerate people = better policy outcomes?
I'm afraid it's not that simple. There are a lot of numerate people in Government. The Government Economics Service runs the show, so far as policy appraisal (that is, assessment before the fact) goes. They know their stuff. And the Government spends a lot on building complex, often (in my view) over-complex, models to discover the truth.
And there you have it. If your model of the world is wrong, then no amount of mathematics can help you. Let me put to one side any beliefs I have about any given Government's moral model of the world. Let us just assume something morally neutral, like the way people make decisions to buy gas boilers. If your mental model of that "simple" decision is wrong (and, it turns out, it isn't a simple decision at all) no amount of numeracy can dig you out of the hole in which you find yourself. Indeed more complex models (by which we really mean equations, hidden from view, nestling in code) makes it worse, because you cannot easily see (or even difficultly see) where you went wrong, so many twists and turns have you taken.
Big data and machine learning (let us not even pretend that AI has any short or medium term role in policymaking) are a deeper concern. In finance these tools are already used. They make a small number of people a lot of money. They don't, obviously, know anything about the immediate health or outlooks for any given stock, beyond what values humans have already applied themselves, through the buying and selling which provides the data the machines then 'feed' on.
Nothing scares me more than the idea of a data scientist who knows nothing abut decarbonisation using a learning algorithm to find the 'answer' to a complex policy problem. This isn't a joke. If you 'analyse' the data without understanding what real world forces give rise to the data, you can kill lots of people.
If data can be used to provide insights which then allow humans, and most particularly social scientists (I exclude economists from this category), to investigate patterns and find out probably reasons for them, then more power to the data monkeys. But if we believe that the Civil Service is too reliant on Social Sciences, we will fail to generate policy that has a positive impact on citizens, because citizens are rooted in society, and society follows social science 'rules', not the rules of physics.
It took me 9 paragraphs to say that policies depend on a robust understanding of the behaviour of citizens, and that only by properly equipping the Civil Service with Social Scientists - not merely economists - can we ever hope for better policy. That numbers alone mislead without human insight. A lack of understanding of Natural Science, or Mathematics, or Social Science, will lead to choices which, with hindsight, look very foolish.
To illustrate my point with a field that I know something about. Civil Servants, we are told, need to know "how to interrogate climate modelling". Well, the Civil Servants who work on climate already possess this knowledge. Indeed, it is an area where the Civil Service already has deep expertise, and better, can draw on the deep expertise of the Met Office, and others. But that isn't the real issue. The big question is what "interrogate climate modelling" means. It should not mean that Civil Servants should be looking at climate models to better forecast when Hebden Bridge will next flood. Because climate models cannot tell you when Hebden Bridge will next flood. They can't tell you what the probability of Hebden Bridge flooding is. They can't even tell you what the probability of the level of rainfall that last caused Hebden Bridge to flood falling over the next 5 years is. They can tell you what the probability is of that rainfall level occurring according to the model but (and very few people seem to recognise this) the model is not reality.
What can climate models tell us? Well, if we have a range of climate models all of which can resolve rainfall at the level required to tell us something about Hebden Bridge and floods, we could look at the range of outputs from those models, and perhaps look at how well each model performs historically (a model which doesn't produce weather over the last decade that in any way resembles the actual weather over the last decade is probably a poor guide to future weather). It doesn't give us a numerical probability of Hebden Bridge flooding, because none of the models are reality and the 'average' of all the models certainly isn't a model of reality but it gives us the best possible scientific insight as to whether Hebden Bridge is more likely to flood over the next decade than it did in the last. Which is useful policy-relevant information.
But at this point you can see that it is easy to get lost in the numbers, and begin to believe that the numbers tell you some 'truth'. The do not. They numbers are shadows on the walls of a cave.
I'm not for a moment suggesting that climate modelling is a wasted endeavour or that we should not take notice of the outputs of climate models. It is precisely because climate models tell us that changing the concentrations of GHGs (and particulates) in the atmosphere changes the weather around us that we should pay attention to them, particularly as climate models are significantly more skillful (I don't care what the dictionary says, "skilful" looks awful and both spellings are valid) than anyone who tells you that GHGs don't cause climate change. But our response to climate models, particularly as civil servants, should not be to interrogate the data. It should be to look at the data, and decide what actions are a proportionate response to the evidence available.
What I am talking about here is risk management, at national (and, of course, global) level. The fact that the future is unknowable does not prevent us from making decisions because we can predict, with some greater or lesser certainty, what is likely to happen. I am very uncomfortable with putting numbers on these things, as it gives a false sense of precision. Rather I would suggest we believe that an outcome which all models predict in most circumstances is more likely than one which only some models predict, which is more likely still than an outcome which only one model predicts, and infrequently. (But of course, the low probability event can't be ruled out).
So if "interrogate climate modelling" means understand the strengths and limitations of climate models and use that to develop a risk framework that focuses the available resources for climate change adaptation where we predict they will do most good, and prevents stupid policy (like building in areas which are predicted to be increasingly prone to catastrophic flooding) then I am all for it. But if "interrogate climate modelling" means drilling down into the data to find deeper meaning, then I fear some fundamental misunderstanding of what a climate model is and isn't, has occurred.
I hope by now, if you have made it this far, you recognise that a framework to respond to climate change relies on social scientists just as much as it does on physical scientists (and that mathematicians, while they have an essential role in developing and verifying models, are of limited use when it comes to applying them in the real world).
We live (and continue to live, we hope) in a world of people. If you don't understand people, you can't build effective policy.
For obvious reasons, this year's Ditchley Annual Lecture, given by The Chancellor of the Duchy of Lancaster, is worthy of both comment and analysis.
Superficially it is hard to disagree with the general thrust of the argument, that Government should better reward risk taking, that the Civil Service should have more diversity of thought, that centralising decision-making in London means that the cultural outlook of the majority decision-makers does not reflect the wider cultural outlook of the country, and that the Government is piss-poor at measuring success, and should do better.
But anyone with even a passing interest in UK politics over the last 5 years will have noticed that our politics is now entirely tribal. We are just as split between individuals who identify as pro and anti Brexit as the US is between those who identify as pro and anti Trump. A significant number of former comrades find themselves stuck behind enemy lines - more on the right than the left, where there are plenty of Conservatives aghast at what the current leaders of their nations are doing on trade, security, fiscal policy and common decency yet are unable to align themselves with the Tories/Dems - but on the left, too, though these types are perhaps more used to being out of step with the majority of 'their' tribe. Pro-protectionist anti-immigrant leftists exist, and have existed forever, but are very much in minority, and have been for years.
The lecture is aimed solely at one tribe, and blames their woes solely on a foe with all the realism of Mantel's Thomas Cromwell -- i.e. there are grains of truth, in here, rooted in documented historical fact, but the overall concoction is entirely fictional -- or perhaps some mythical version of Clive of India as a Hero of the Empire. If I were to ask you what went wrong in 2008, and has failed to go right since, I doubt you'd point the finger of blame at the Civil Service. Maybe you would. Maybe I'm too close. But I suspect the voices that are now claiming the Civil Service is ideologically attached to Europe would, 30 years ago, have claimed that the Civil Service was too slow to adopt to the benefits offered by globalisation. Where politicians lead, Civil Servants follow. The single market was a political creation and one which diminished the powers of Whitehall. It wasn't dreamed up by Mandarins nor do I imagine they were terribly keen on it in the early days, given that power flowed away from them. And we all know that austerity was a political, not technocratic choice. Officials seldom advocate policies which will see their Departmental budgets decimated, or as was the case for several Departments, far worse than decimated.
But I'm going on too long. Many many columns will be written about the impending re-organisation of Whitehall. Here are some others.
But I would like to point out what I think is a flaw at the heart of the argument that we need . It's about Maths and Science. As a physical scientist, I completely reject the assertion that decision-making should be "less reliant on those with social science qualifications and more welcoming to those with physical science and mathematical qualifications".
As a citizen rather than a Civil Servant I've seen policies come and go. Most policies, in my observation, fail or a best offer extremely poor value for money compared to readily available alternatives. Some crash and burn spectacularly, like privatization of probation services, or the Green Deal. Most chunter along, eating money and not doing much active harm or good. Occasionally we hit a great one (in my field, both the boiler regulations, which drove down UK gas use, or Contracts-for-Difference which, in the wind sector at least, have driven renewable electricity prices down beyond our wildest dreams.
But yeah. Policies usually suck. And they usually suck because the real numbers bear no resemblance to the forecast numbers.
So, more numerate people = better policy outcomes?
I'm afraid it's not that simple. There are a lot of numerate people in Government. The Government Economics Service runs the show, so far as policy appraisal (that is, assessment before the fact) goes. They know their stuff. And the Government spends a lot on building complex, often (in my view) over-complex, models to discover the truth.
And there you have it. If your model of the world is wrong, then no amount of mathematics can help you. Let me put to one side any beliefs I have about any given Government's moral model of the world. Let us just assume something morally neutral, like the way people make decisions to buy gas boilers. If your mental model of that "simple" decision is wrong (and, it turns out, it isn't a simple decision at all) no amount of numeracy can dig you out of the hole in which you find yourself. Indeed more complex models (by which we really mean equations, hidden from view, nestling in code) makes it worse, because you cannot easily see (or even difficultly see) where you went wrong, so many twists and turns have you taken.
Big data and machine learning (let us not even pretend that AI has any short or medium term role in policymaking) are a deeper concern. In finance these tools are already used. They make a small number of people a lot of money. They don't, obviously, know anything about the immediate health or outlooks for any given stock, beyond what values humans have already applied themselves, through the buying and selling which provides the data the machines then 'feed' on.
Nothing scares me more than the idea of a data scientist who knows nothing abut decarbonisation using a learning algorithm to find the 'answer' to a complex policy problem. This isn't a joke. If you 'analyse' the data without understanding what real world forces give rise to the data, you can kill lots of people.
If data can be used to provide insights which then allow humans, and most particularly social scientists (I exclude economists from this category), to investigate patterns and find out probably reasons for them, then more power to the data monkeys. But if we believe that the Civil Service is too reliant on Social Sciences, we will fail to generate policy that has a positive impact on citizens, because citizens are rooted in society, and society follows social science 'rules', not the rules of physics.
It took me 9 paragraphs to say that policies depend on a robust understanding of the behaviour of citizens, and that only by properly equipping the Civil Service with Social Scientists - not merely economists - can we ever hope for better policy. That numbers alone mislead without human insight. A lack of understanding of Natural Science, or Mathematics, or Social Science, will lead to choices which, with hindsight, look very foolish.
To illustrate my point with a field that I know something about. Civil Servants, we are told, need to know "how to interrogate climate modelling". Well, the Civil Servants who work on climate already possess this knowledge. Indeed, it is an area where the Civil Service already has deep expertise, and better, can draw on the deep expertise of the Met Office, and others. But that isn't the real issue. The big question is what "interrogate climate modelling" means. It should not mean that Civil Servants should be looking at climate models to better forecast when Hebden Bridge will next flood. Because climate models cannot tell you when Hebden Bridge will next flood. They can't tell you what the probability of Hebden Bridge flooding is. They can't even tell you what the probability of the level of rainfall that last caused Hebden Bridge to flood falling over the next 5 years is. They can tell you what the probability is of that rainfall level occurring according to the model but (and very few people seem to recognise this) the model is not reality.
What can climate models tell us? Well, if we have a range of climate models all of which can resolve rainfall at the level required to tell us something about Hebden Bridge and floods, we could look at the range of outputs from those models, and perhaps look at how well each model performs historically (a model which doesn't produce weather over the last decade that in any way resembles the actual weather over the last decade is probably a poor guide to future weather). It doesn't give us a numerical probability of Hebden Bridge flooding, because none of the models are reality and the 'average' of all the models certainly isn't a model of reality but it gives us the best possible scientific insight as to whether Hebden Bridge is more likely to flood over the next decade than it did in the last. Which is useful policy-relevant information.
But at this point you can see that it is easy to get lost in the numbers, and begin to believe that the numbers tell you some 'truth'. The do not. They numbers are shadows on the walls of a cave.
I'm not for a moment suggesting that climate modelling is a wasted endeavour or that we should not take notice of the outputs of climate models. It is precisely because climate models tell us that changing the concentrations of GHGs (and particulates) in the atmosphere changes the weather around us that we should pay attention to them, particularly as climate models are significantly more skillful (I don't care what the dictionary says, "skilful" looks awful and both spellings are valid) than anyone who tells you that GHGs don't cause climate change. But our response to climate models, particularly as civil servants, should not be to interrogate the data. It should be to look at the data, and decide what actions are a proportionate response to the evidence available.
What I am talking about here is risk management, at national (and, of course, global) level. The fact that the future is unknowable does not prevent us from making decisions because we can predict, with some greater or lesser certainty, what is likely to happen. I am very uncomfortable with putting numbers on these things, as it gives a false sense of precision. Rather I would suggest we believe that an outcome which all models predict in most circumstances is more likely than one which only some models predict, which is more likely still than an outcome which only one model predicts, and infrequently. (But of course, the low probability event can't be ruled out).
So if "interrogate climate modelling" means understand the strengths and limitations of climate models and use that to develop a risk framework that focuses the available resources for climate change adaptation where we predict they will do most good, and prevents stupid policy (like building in areas which are predicted to be increasingly prone to catastrophic flooding) then I am all for it. But if "interrogate climate modelling" means drilling down into the data to find deeper meaning, then I fear some fundamental misunderstanding of what a climate model is and isn't, has occurred.
I hope by now, if you have made it this far, you recognise that a framework to respond to climate change relies on social scientists just as much as it does on physical scientists (and that mathematicians, while they have an essential role in developing and verifying models, are of limited use when it comes to applying them in the real world).
We live (and continue to live, we hope) in a world of people. If you don't understand people, you can't build effective policy.
Subscribe to:
Posts (Atom)
