My immediate response to this blog post was a sense of recognition. And not just in the sense that it reads a lot like a Hacker News comment. Stripped of the rhetoric, hyperbole, and political bias much of what is in the post reminds me of a decade ago when the Coalition gained power, I was asked to set up the first Government Skunkworks, and the Government Digital Service was formed.
So some initial thoughts based on that 10 year view.
The section on data has a strong Hari Seldon/John Campbell feel to it, the main problem in the government world is not that data is not predictive but rather people do not want to believe those predictions. The current Australian fires are a case in point, all the data predicting them existed, it had been analysed and recommendations on how to mitigate the problem had been published. But successive national and state governments in Australia took no action.
Data cannot substitute for political will, especially when it comes to wicked problems like Climate Change.
Data is a powerful tool but like all tools it requires purpose, context and intent.
A classic example is LEO, the Longitudinal Educational Observatory (no, me neither). LEO is a Department for Education system which is a mashup of educational attainment (GCSE Woodwork) and earnings data from HMRC.
LEO allows you to see which University courses lead to the highest earnings, which A-Levels are linked to social mobility.
This all sounds interesting but there are two big catches which make LEO a policy paralyser rather than a policy enabler.
The first is the “Satnav Paradox”. Older readers may recall that when satnavs were first introduced one of their key sales points was that you would never be stuck in traffic ever again because the system could automatically reroute you. We all know what happened, everyone is rerouted the same way so at best nothing has improved and, in fact, added frictional cost of decision making has made things worse.
The same applies to LEO, touching the system changes the system and this time on a timescale measured not in seconds but in years or decades. If everyone knows that running a burger van in Broughty Ferry has the highest potential earnings then Broughty Ferry will be facing a tsunami of burgers at some point from 2025 onwards and the earnings from running a van will collapse.
The second problem with LEO is linked to the timescale point. When people talk about cybernetic government they forget that feedback loops are often very long, how many Zeppelin engineers do we now need?
The links between data and situational awareness are something which GDS began to explore. It feels very much like Terra Incognita in the government space at the moment but it requires a clear purpose. Every large dataset will reach the point where the chance of finding a positive match reaches unity. Being able to distinguish between the laws of mathematics and the complexities of reality is the core challenge.
The key thing to remember about economics is that it is always right in theory and wrong in practice.
Linking it with AI is brave…
On project managers, the fundamental question is do you want project managers or project deliverers? The former is about ritual, the latter about reality. The Senior Civil Service understands and rewards the former, it is deeply uncomfortable with the latter.
The Junior Researchers section is deeply flawed. People are not resources to be worked into the ground. Burning out at the age of 21 will ruin your life.
On communications, I have long been of the belief that every major project needs what we would call a seanchaí, a combination of an oral historian and a story teller who captures the narrative and the lessons learned.
On the policy section, all the successful programmes I have delivered have had mixed teams with all the core people working together. The artificial segmentation of the Civil Service into cadres is a nonsense.
As for the “Super-talented weirdos”, the Senior Civil Service is a monoculture and reacts very badly to those outside that culture. I have seen and indeed experienced the extreme bullying and harassment of those who do not fit into the classical mode. Creating a more diverse culture will require very supportive and empathetic leadership and management.
Readers will note that I have not commented on the recruitment process or the political dimensions to the blogpost. That is deliberate.
Cummings strikes me as someone who needs an editor, not just in the sense of redlining prose, but someone tough enough to provide a buffer for staff and to say “No”.
Ironically, his blog post seems to posit a theory of government closest to Project Cybersyn…