Blog by Geoff Mulgan: “In this piece I describe what I call ‘knowledge twins’, combining data, evidence, foresight, tacit knowledge and more, to enable governments to make better decisions and to learn more effectively. I show how these can build on many existing tools, from evidence syntheses to knowledge graphs, digital twins to systems maps, and how they can help policy-makers realise the full potential of AI. I show how they could evolve over the next decade to make intelligence useful and used, on everything from economic growth to decarbonisation, mental health to poverty, bio and climate risks to the future of care, and become essential cognitive infrastructures for governance.
Understanding what causes what – the grounding for everything governments do
Governments have always been interested in intelligence. But in the past that usually meant secret intelligence about enemies and threats. In the 19th and 20th centuries that interest broadened to include intelligence about the economy, health and population, alongside a recognition that shared intelligence can sometimes be much more valuable than intelligence that is hoarded.
Every part of any government rests on implicit or explicit models of how the world works. If we pass this law, spend money in this way or introduce (or cut) this programme, these are what we think the effects will be. Pre-modern governments were different, as are contemporary autocrats: it might be enough for a law to reflect theology, or the whims of a king or President, or just hope. But in the modern era it’s hard to see why we as citizens would want our money spent, or our freedoms curtailed, for programmes not based on serious assessment of whether they will succeed.
For the same reason much of politics revolves around promises that actions will have predictable effects. So, if a party or candidate promises that their actions can achieve desirable ends such as higher growth, better health, lower crime or immigration, we expect them to be able to say how and we expect to be able to examine their homework to see if it’s convincing.
These causal models get rough and ready scrutiny in an election campaign, and sometimes outside them, mainly through interviews on broadcast and in print media, and on some topics, such as fiscal plans, there are well-resourced independent institutions (like the UK’s Institute for Fiscal Studies or Resolution Foundation) able to scrutinise the assumptions. Indeed, economics is a field where shared intelligence is widely accepted: governments share lots of data, analyses, models and forecasts and have to persuade markets and observers that they know what they’re doing (though of course there are many serious blindspots).
I argue here that these habits of explicit claims, interrogation and openness, should become more normal in other fields, with all departments and agencies seeing themselves as providers of reliable intelligence to the systems they are responsible for. In an era of ubiquitous AI this will be vital for making the most of new generations of technology. But it will also require new institutions, new methods and new mindsets…(More)”.
