AIiiiii!! How AI won't fix the world's problems with bureaucracy
AI is going to transform economies and governments but we need to answer basic questions about accountability, about the need for regulations and about the degree to which social control is warranted
“Every revolution evaporates and leaves behind only the slime of a new bureaucracy.” - Franz Kafka
It is widely acknowledged that the development of AI ushers in a revolution in the way we work, play and learn. As Ben Sixsmith tells us it has already ushered in a new generation of grifters and scammers as well as an emerging industry of spotters able to see ‘AI-produced slop’ when it arrives. Elsewhere Louis Mosley, Palantir’s UK boss, is writing that the AI revolution will see the end of bureaucracy:
“It is the lanyard class that has most to fear from this technology, because AI is better than any human when it comes to paperwork. Blue collar will replace white collar in the hierarchy of value.”
I appreciate that Mosley works for a company that’s main energy is directed to the bureaucratic systems of “...government agencies, militaries, and corporations…” so you’d expect its bosses to be right on brand with ‘we kill bureaucracy’ messaging. It is also true that, for many facing the daily frustration of bureaucratic uselessness, the annihilation of the lanyard class presents a huge and appealing outcome, you can almost feel the sense of liberation building. But Mosley, and plenty of other AI promoters, are not arguing for the reduction of bureaucracy or the elimination of regulations. Were there none of these things then Palantir wouldn’t have much of a business model remaining.
No, Palantir and other similar businesses, are the living embodiment of the Kafka quotation I opened with. They present their AI systems as a revolution (and this is probably true) but do nothing to remove or reduce the sclerosis of bureaucracy and regulation: it is just the computer saying no rather than a human. I know this isn’t what they’re saying, it’s more about how AI can provide insights and analysis to large organisations and the message (familiar from each iteration of computing power) that this revolution will empower front line workers. Use of AI to complete the paperwork doesn’t remove that paperwork, instead it shifts the accountability for bureaucratic activity even further away from individual decision-makers and onto “The System”. When something goes wrong - and it is definitely a ‘when’ not an ‘if’ - the sub-contracting of oversight to AI means that nobody, least of all the front-line manager, is accountable for the error. To return to Kafka, we arrive at the perfection of his bureaucratic nightmare rather than any sort of liberation:
“Surveyor, in your thoughts you may be reproaching Sordini for not having been prompted by my claim to make inquiries about the matter in other departments. But that would have been wrong, and I want this man cleared of all blame in your thoughts. One of the operating principles of authorities is that the possibility of error is simply not taken into account. This principle is justified by the excellence of the entire organization and is also necessary if matters are to be discharged with the utmost rapidity. So Sordini couldn’t inquire in other departments, besides those departments wouldn’t have answered, since they would have noticed right away that he was investigating the possibility of an error.”
In the bureaucracy of AI, Kafka’s Surveyor wouldn’t even get the opportunity to appeal before unhelpful humans, but would instead be endlessly frustrated by computer-generated refusals to accept error or accountability. By the ‘slime of a new bureaucracy’. The problem with administrative systems isn’t efficiency but purpose. Saying you can replace lanyard-wearing HR officers with AI sounds great but does nothing to eliminate the bureaucratic reasons behind the explosion in the numbers of those lanyard-wearers.
The lanyard-wearer exists for two main reasons: firstly to ensure compliance with (the organisation’s interpretation of) laws and regulations enacted by the state; and secondly to provide protection for front-line managers by shifting accountability from them to the (relatively anonymous) system itself. The existence of AI doesn’t eliminate either of these bureaucratic functions and, worse, it leaves questions of accountability entirely unanswered. It is clear that large bureaucratic systems, and especially such systems within the public sector, already lack any real sense of decision-maker accountability. By replacing the people collectively used as cover for egregious decisions (compliance officers, human resources, DEI, legal services) with AI systems, either the accountability lands on the desk of the blue-collar front line worker and his supervisor, or else responsibility disappears into the ether, lost in the administrative certainty of the AI.
Most costs of regulation and bureaucracy fall on the individuals, families and businesses constrained by the rules, not on the administering of those rules. Replacing the man with the clipboard by a computer screen doesn’t get rid of the clipboard, it just gets rid of the person holding it. Those planning rules, licensing conditions, equalities act requirements, reporting obligations, environmental directives and pricing interventions will still be there, we’ll still be hobbled by regulations (probably including a whole new bunch of regulations around our use of AI, social media and the Internet). It is just that those regulations will be administered by an AI system programmed and prompted by someone you don’t know and will never meet, and overseen by a tiny cadre of administrators who will, without doubt, adopt the view that the AI is infallible, that there is no possibility of error and that your appeals to fairness or justice are specious.
Mosley, in an observation of surprising naivety, concludes that “...Friedrich Engels was right all along: the state won’t be abolished, it will simply ‘wither away’.” This is to suppose, ignoring the warnings of Kafka and the wisdom of Hayek, that the state is defined by the people who work for the state. Each change in the world, technical, social or economic, leads to the call for more regulations, for new rules and for additional laws. With each of the resulting acts creating the need for new bureaucracy. If you really want to reduce bureaucracy, the way to do so is to eliminate regulations, repeal laws and have fewer rules. But, and I appreciate this undermines Palantir’s business model which is predicated on the continued existence of bureaucracy, simply hanging the lanyard on the computer does not solve the problem of bureaucracy.
There’s no doubt that AI is going to transform economies, entertainments and governments but we need to answer basic questions about accountability, about the need for regulations and about the degree to which social control is warranted. If many of the state’s decisions are made by AI, who is accountable? Are the regulations we impose on business, families and individuals justified? How much autonomy should people enjoy and does that include not being filmed, spied on, tracked and ID’d all the time? Lots of the chatter about AI ignores these problems and prefers the sort of outlook pioneered by the likes of Palantir where intrusion and external regulatory oversight is justified by a sort of Benthamite panopticon, a MaxU world where the greatest good is defined as the greatest good for the state.
Using AI to administer bureaucracy doesn’t reduce bureaucracy. If we are to get the benefits from the latest iteration of technological revolution, we must start with answering this question, not with arguments for marginal and modest savings in the cost of administering the state.