Replicating the digital welfare state

One of the topics I particularly enjoyed during my MSc (how is that nearly 10 years ago) was policy transfer - how policy ideas and instruments seem to ‘move’ across governments, geographies and time periods.

It’s a whole field of academic theory so I won’t try to summarise it all here, but there are some really interesting drivers behind policy transfer such as pre-existing relationships between countries and governments, who is considered to be most credible and worthy of learning from, and how compatible different political and policy environments are.

There is also the question of policy translation: recognising that successfully adopting a policy from elsewhere is not a copy and paste job but a process of negotiation, adaptation and power dynamics. Policy transfer and translation can take place between countries/ governments, and also within a country.

In this post, I’m taking the digitisation and automation of the welfare state as an example of international policy transfer and translation. The rollout of digitised and automated systems within welfare states continues to spread globally, despite growing evidence of their potential flaws (see this previous post for example) and growing scepticism of technologies like AI. Why is this?

Many countries have similar priorities in their welfare state policies: cut spending, improve efficiency, reduce fraud, make the customer experience smoother. Digitisation and automation can genuinely contribute to these priorities in many cases, so naturally politicians and policymakers will look to introduce and scale up technological solutions. Many dedicated civil servants are looking to technology to help them find solutions to complex challenges.

However, there is also a generous amount of hype about tech solutions, particularly AI, at the moment. For politicians and political influencers, embracing AI is an easy way to be seen as innovative and entrepreneurial; being, or appearing to be, behind the curve is to be avoided at all costs. Normal levels of scrutiny and healthy scepticism therefore tend to be greatly reduced.

Much of the AI hype is driven by the firms that build the AI tools themselves, many of which operate internationally. Clearly they are aiming to maximise sales by pitching their products to as many territories as possible, while also using case studies of early adopters to motivate those lagging behind.

If you’re lagging behind, you’re definitely not keeping up with the Joneses. For those less familiar with old-fashioned British sayings, this means the social pressure, real or imagined, to compare favourably with ones neighbours. In policy terms, this can mean looking at the national or international reputation of a friend (or rival) and deciding that their digital welfare policy is the envy of everyone, and therefore needs to be replicated. ‘Learning from what works’ is probably the official explanation rather than envy of course, but policymakers do need to be really cautious about the extent to which ‘what works’ is replicable across vastly different welfare state contexts.

Finally, one of the critical dimensions of policy translation is power: who holds it in the administration that’s looking for policy solutions, what’s the power differential between the places that are transmitting and receiving policy ideas, and where does the epistemic power (the power to influence what others think and believe) reside? In the case of digital welfare state policy, there are significant imbalances between those designing and implementing policy and those who influence them, and those who are experiencing or critiquing the downstream effects.

This jumble of factors are, for now at least, keeping the digital welfare state high on the agenda. I’ll be interested to see if there’s a backlash at any point, either from a growing weight of evidence of failure and social harm, or a failure to realise the predicted savings and efficiency, or both.