AI - friend or foe for social security administration?

Scrolling through the endless stream of AI related articles and posts on Linked In, I come away dazed and confused.   Is the current round of AI hype the beginning of the end as no quarter is given by commentators spreading doom and gloom, or the opening of a new chapter with cause for celebration?

Reflecting on forty years at the intersection of social security and digital technology, my observation is despite the incredible progress made in automation, not much has changed when it comes to effectively addressing wicked social problems.  This is not a criticism of social security administration, merely reflecting the reality that technological innovation can’t solve everything - and we are silly to think it somehow will.

The social security system maintains the social safety net, provides insurance against social risks and on the whole has a reassuring stabilizing effect during periods of economic instability. It stands above whatever technological change is happening around it. Technology has improved efficiency and effectiveness of social programs but it is not a cure-all for the social challenges we face.  If the doomsayers are right, AI may give rise to economic instability and social disruption. If and when it happens, the social security system and the people who make it work will be there to mitigate the impacts.

As we pass through the current phase of AI hype, we should note it hasn’t dropped out of the sky like a rain song. Recent developments in generative AI represent a big-step change (and this is a very big one) and the tech sector is good at making us believe this is the biggest step ever. Perhaps it is but so what – change is all around us.

Previous big steps such as the personal computer, the internet, smartphones, big data, have all brought their fair share of calamitous predictions in terms of their impact on society. This phase, at this stage of the cycle, appears no different. Will there be significant impact ? Yes. Will it all be for the better? Probably not. Unintended consequences? Without doubt. Social change? Jury is out. A big benefit to social security administration? Most likely.

The tech sector has brought a never-ending stream of continuous improvement in automation over the past 60+ years, with a mix of big and small steps. The current AI led big step should be observed in the context of the ongoing automation journey, a journey social security administrators have largely managed to control while achieving positive and enduring results.  On occasions when social security administrators have overreached with automation, they have been called out and held accountable, noting there have been high profile failures in recent times across various jurisdictions. There have been instances of significant harm from misguided intent and unintended consequences. Generally however, social security administration has held course in keeping humans in control of human services and avoiding the technology trap of inflated expectations.

So will AI get us closer to addressing the human issues of social disadvantage? Amongst the predicted economic disruption and social changes, AI does offer many opportunities to improve social security administration (service delivery and policy development) leading to better social outcomes.  But can William Beveridge’s five giants of inequality as he named them in 1942: idleness, ignorance, disease, squalor, and want be beaten by this next wave of AI led technological change, when other waves have failed?  While digitisation and digitalisation has enabled dramatic improvements, particularly in service delivery, closing out on the wicked social problems remains as elusive as ever.  

Like previous waves of technological change, there is risk and reward.  How we rock and roll with managing the risks will determine the reach and impact of the rewards. We shouldn’t lose sight of the tech industry as commercially motivated to innovate and promote technology led change as a stairway to heaven. Potential risks are often downplayed rather than addressed head-on. 

The tech industry, like the health, defence and nuclear sectors before it, shouldn’t lose sight of societies right to manage these risks through regulation, governance, plain old common sense and a whole lotta love for our fellow human beings. Managing risks sometimes requires trade-offs. For example people are generally happy to wait for new drugs to be properly tested before they are released to market. Why should a patentable AI algorithm or machine learning program be treated any differently? And when a tech industry executive says ‘regulation stifles innovation’ then remind them that within the competitive market where their companies thrive, winners innovate because of, rather than be constrained by social norms, wise legislation and sound regulatory frameworks.

Social security administrators have in the main been wise users of technology. They understand technology hype can fuel false promises. Metaverse for social security anyone? Convenient, more personalised, 24x7 digital service delivery is no substitute for compassionate and fiscally responsible social policy. Social security administrators, aided and assisted by responsible AI and ethical data management, remain well placed to balance competing interests and priorities of individuals and society at large when designing social programs.

Returning to the opening question, AI, friend of foe to social security administration: it is neither. It is a (big) part of the on-going automation journey. If you allow social security administration to be managed as a logistics exercise for service delivery efficiency, then AI may well be a foe.   If you incorporate responsible and ethical data management services starting within the policy domain, then AI can be your best friend.  

Social security is effective largely due to the dedication and commitment of the people who administer it, from policy makers to front line service delivery and the people who support them.  Program efficacy is compromised when social security administrators get caught up in tech hype. Mitigating these risks is made possible by ensuring AI based systems use data where there is confidence in the source, provenance, accuracy, relevance and the who, what, how and when it was collected. In tech terms this is the metadata.

The secrets to policy level innovation lie within the metadata layer. It is through metadata that disconnected data sources can be linked and harmonised. Metadata management is therefore an effective assurance mechanism for ensuring AI based systems operate responsibly and ethically. 

Meanwhile, the business of social security administration rocks on and in general when it comes to technology, the song remains the same.

#AI #metadata #digitalisation #digitalisation #socialsecurity #ledzeppelin #socialservices #socialprotection #knowledgegraphs

Note: No generative AI tools were touched or harmed in the making of this article