The Cost of the Ledger: from RFK’s critique to the algorithmic age

On March 18, 1968, Robert F. Kennedy addressed an audience at the University of Kansas and delivered what remains the gold standard of systemic critiques. He targeted the ultimate economic metric of his era: the Gross National Product. Nearly six decades later, as we move through a digitised, AI-saturated landscape, his words strike with a huge impact — because the core tension he described has not changed. It has amplified.

This Part I of the story is about how our societal metrics have evolved from tracking physical things to monetising human attention, how the corporate world fell into the trap of eternal growth, and how we can build a new, human-centric ledger. If Part I is from the past to the present, Part II is from the present to the future, and Part III is the longer 100-year horizon.

Act I: The historical foundation

At the heart of Kennedy’s 1968 address was the dismantling of the illusion that a balance sheet can serve as a proxy for a nation’s health. He identified a fundamental systemic flaw in the GNP:

The problem is that a purely transaction-based ledger is both blind and amoral. Under its logic:

• Destruction is profitable: The GNP counts air pollution, cigarette advertising, and ambulances clearing highway carnage. It counts special locks for doors, jails, and the destruction of natural wonders.
• Violence is growth: It counts weapons, napalm, nuclear warheads, and the glorification of violence on television to sell toys to children.
• The intangible is worthless: It has no category for the health of our children, the quality of their education, the beauty of our poetry, the strength of marriages, or the integrity of public officials.

This represents the “core bug” in human systems. Once we latch onto a highly measurable metric, we warp our entire culture to maximise it, regardless of the human toll.

Act II: The modern attention ledger

If RFK were alive today, he would find this exact same equation running on digital overdrive. In the physical era, the metric was the accumulation of material things. In our modern economy, we have monetised human attention. Our new Gross National Product is built on Engagement, Consumption, and Output.

Just like the GNP of 1968, this digital ledger is deeply deceptive:

• Compulsion over learning: The ledger counts “Time on Site” and “Scroll Depth,” completely blind to whether that time was spent in genuine learning or anxious, compulsive doomscrolling.
• Growth over connection: It celebrates “User Growth” while actively ignoring the modern loneliness epidemic and the decay of real-world community spaces.
• Volume over depth: It rewards “Content Output” but offers no metric for nuance, depth of thought, or the preservation of truth.

The AI parallels: updating RFK’s speech

If we rewrite Kennedy’s list of “what counts” for the AI era, the societal cost of optimising for the wrong metrics becomes alarmingly clear:

• “Cognitive Pollution” (vs. Air Pollution): Today’s metrics count cognitive pollution. AI allows for the infinite, low-cost generation of synthetic text, SEO-slop, and deepfakes. Under current economic models, the massive compute power used to generate this noise is celebrated as “productivity” and “tech sector growth,” even as it actively erodes our shared information ecosystem.
• “Digital Special Locks” (vs. Physical Locks): Just as GNP counted physical locks, today we spend billions developing AI-detection software, watermarking tools, and anti-misinformation filters. We use massive amounts of energy and human genius to build “digital locks” to protect ourselves from the very chaos our own technologies create. Both the threat and the shield boost GDP, but society is no wealthier for it.
• “Algorithmic Rage” (vs. Toy Commercials): Instead of TV programs glorifying violence to sell toys, we have recommendation engines serving outrage, polarisation, and body-image anxiety to teenagers because anger and insecurity are the most reliable drivers of ad-supported engagement. The system functions perfectly according to its metrics, but it devastates the mental health of a generation.

Act III: The frontline, where the work gets done¹

When we bring this flawed ledger into the modern corporate office, the ultimate, unyielding KPI is eternal, compounding capital growth to satisfy shareholder profitability. Because our society is geared toward generating capital rather than human value, the qualitative human elements are treated as expendable.

In every organisation deploying AI, these tensions collide on the front line: with the clinicians, lawyers, analysts, engineers, teachers and creators whose work the system now drafts, checks or replaces.

As AI is rapidly integrated into workflows, organisations face a massive increase in responsibility. They are the gatekeepers of the work that reaches patients, clients, customers and the public. If they implement AI solely to satisfy shareholders’ demand for eternal growth, they become active contributors to the cognitive pollution eroding our society, and they hollow out their own capability at the same time. When productivity is prioritised purely for the sake of the ledger, we risk increasing raw volume output while completely abandoning quality.

This creates a severe, direct threat to the very people doing the work. If we treat professionals as mere editors of synthetic AI-drafted material just to hit volume targets, we trigger a rapid decline in their skills, fulfilment, and mental well-being.

To truly improve output, we must realise that quality requires cognitive space. People can only thrive, validate, and judge AI output effectively if they are given the time to:

• Think and consider systemic challenges.
• Sharpen and train their own skills alongside the technology.
• Feel valued, upskilled, and empowered rather than replaced.

The alternative is a corporate race to the bottom: reducing headcount, ballooning output volume, and actively participating in the societal problem.

Act IV: The counter-strategy — principles before metrics

Here is where this conversation usually goes wrong, including in my own first draft of it. Somebody proposes a metric. Everybody argues about the metric. And the argument about the metric slowly settles the principle without anyone having agreed to it.

So the principles first. Get them wrong and better numbers won’t save us. Get them right and the numbers become an engineering problem.

Five I’d defend, and open a debate to kill some and improve others.

One. Measure the conditions of the work, never the merit of the work.

A ledger may audit the circumstances under which work gets made. It may never grade the work itself.¹

By conditions I mean things a stranger could verify without having an opinion about the output:

• Hours — how much paid time a person actually had for the task, as opposed to how many tasks they closed.
• Training — paid hours spent getting better at the craft, not hours spent learning this quarter’s tool.
• Junior headcount — whether anyone is still being brought into the profession, because AI eats the entry rungs first and the damage only shows up a decade later.
• Human authorship — who made the thing, disclosed, not assessed.
• Rework — what was scrapped and redone, which is the honest price of a fast process.
• Exposure — the AI-mediated adverse events that land on people, counted by class of harm and disclosed by position relative to the system.²

Every one of those is externally observable. None of them requires anybody to decide what good looks like.

And each can be set as a floor rather than a target. A floor harms nobody’s freedom. You can always exceed one, nobody is ranked by how far above it they sit, and there is no reward for faking it.

Merit is the other thing entirely. The moment a ledger scores whether work is good, somebody has to define good. That definition becomes the target. The industry optimises for it. And you have accidentally built an authority on quality: a body that decides what counts as a good diagnosis, a sound judgement or a well-made piece of work, standing over professions whose whole job is making exactly those calls.¹

The intentions of whoever holds that pen are irrelevant. This is the Act I bug applied to my own proposal: latch onto a measurable definition of good, bend the culture to maximise it, destroy the thing you were protecting. It happens by mechanism, not by malice.

So: this is the line between a labour standard, which is dull, enforceable, of the same family as working-time rules and health and safety, and a ministry of taste. Every other principle on this list is a choice I’d happily lose an argument about. This one is the thing that stops the Human Ledger from becoming the exact failure it was written to fix.

Two. Publish everything you measure.

The opposite of a human ledger is not an unmeasured world. There is no unmeasured world. It is a world measured privately. Insurers already price AI liability. Rating agencies are learning to price knowledge loss. Those instruments exist, they are proprietary, and they measure downside protection rather than human outcome. Compulsory publication is the whole idea.

Three. Floors, not targets.

Act I was an argument that every maximised metric eventually gets bent. A Human Ledger built as a set of targets will be gamed within two reporting cycles. Built as thresholds you may not breach, it can’t be — as there’s no upside to exceeding a floor, so there’s no incentive to fake one.

Four. Count what you export.

A clean domestic ledger and a filthy supply chain are entirely compatible. We know this because we’ve already run the experiment with waste and with carbon, and had to invent border adjustment to close it. Offshored volume has to land on the ledger of whoever commissioned it, or the ledger becomes a laundering mechanism with a nice logo.

Five. Protect the unmeasured space.

A ledger that can measure everything will eventually be asked to. Some territory has to be permanently out of scope, and harder to bring into scope than an ordinary policy change, because premature, nascent and fragile work needs somewhere to survive until it becomes recognisable, and that is most of the work that ever mattered.

Those six conditions are the proposal.¹ They are the same whether you run a hospital trust, a law firm, a bank, a newsroom, a software team or an agency. Each one can be checked by a stranger, and each one can be set as a floor.

What those conditions look like inside one sector is the job of an annex: the same six, measured in the units that sector actually keeps. The first annex, for the creative industries where this argument started, closes this article.

The four arguments I didn’t finish

1. Who holds the pen? There are only three answers. A government, and you get direction. Capital, and you get liability management. Nobody, and you get exploitation with good quarterly results. I don’t think there’s a fourth. My own view is that the pen belongs to a legislature for the disclosure standard and to nobody at all for the quality judgement, which is the same split that let sustainability reporting work without anyone voting on product design. But that’s a position, not a settled fact.
2. Which of the six conditions survive?¹ I suspect some of them collapse in the debate. I’d like to find out which.
3. Who audits it? D&I and sustainability reporting only grew teeth when disclosure became mandatory and third-party assured. Self-reported human metrics are marketing.
4. What stays permanently out of scope, and who guards the boundary? This is the hardest one and the one I’d most like help with, because a boundary that can be moved by a single majority isn’t a boundary.

If we measure AI success solely by head-count reduction and output volume, we fall victim to the exact trap Kennedy warned us about in 1968. But if we design our implementations to promote human agency, we turn technology into a tool that serves the human experience rather than exploiting it.

Who else is in this conversation — and why none of it reaches us

Before anyone dismisses this as fantasy: serious institutions are already building pieces of it. The problem isn’t absence. It’s that the work sits in four silos, and not one of them lands on the desk of anyone actually running an organisation.¹

1 · Macro “Beyond GDP”

The direct descendant of Kennedy’s speech. The UN Secretary-General’s High-Level Expert Group published Counting What Counts: A Compass of Progress for People and Planet on 7 May 2026, proposing a global indicator dashboard to sit alongside GDP, with the General Assembly expected to act on it this month. The OECD runs How’s Life?. The UK’s ONS maintains around 60 national well-being measures across 10 domains. Rigorous, decades deep, and entirely national-accounts territory. It never descends to the level of a company, let alone a campaign.

2 · Workforce and AI ethics

Partnership on AI’s Framework for Promoting Workforce Well-being in the AI-Integrated Workplace asks directly whether AI success metrics are grounded in worker well-being, which is the closest anyone has come to Principle One. The US Department of Labor published its AI and Worker Well-Being principles in 2024, centred on worker empowerment, transparency and human oversight. Both are voluntary, both are policy-flavoured, and neither has meaningful adoption inside this industry.

3 · Enterprise AI measurement

Where the money actually is, and it measures the exact opposite: adoption rates, power-user density, hours saved. The exception proves the point. Glean’s Work AI Index 2026, surveying 6,000 workers across the US, UK and Australia, found people spend 6.4 hours a week “botsitting” — feeding context to AI and cleaning up wrong answers — which is more time than they spend using AI to produce anything. 69% admit shipping work they haven’t verified or can’t confidently stand behind. That is the cognitive space ratio and the judgement rate, already measured, with a hard number attached. Nobody reports it.

4 · Regulation, arriving unevenly

EU AI Act transparency obligations (Article 50) apply from 2 August 2026; the Act’s human-oversight duties (Article 14) were deferred to 2 December 2027.³ Connecticut’s SB 5 (Public Act 26-15) requires disclosure of automated employment-decision technology and state reporting on AI-associated layoffs and entry-level job impacts. But Colorado runs the other way: its AI Act was due to take effect in June 2026 with mandatory annual impact assessments, then was amended — assessments stripped out, start date pushed to 1 January 2027. Disclosure is winning; assessment is being negotiated away. Which is the argument for Principle Two.

The closest names to this thesis

Tyler VanderWeele at Harvard’s Human Flourishing Program; Karina Vold and Ashton Anderson at the Schwartz Reisman Institute, whose Florea AI project — three years, US$3.6M, Templeton-funded, with Purdue’s Louis Tay — asks whether models can be optimised for long-term well-being rather than short-term satisfaction; and Partnership on AI. Excellent work, none of it in conversation with agencies or brands.

Which is the white space. Nobody has translated Beyond-GDP thinking into an operational instrument for the organisations deploying AI over people’s work.¹ That’s the gap this is aimed at.

The five principles, the three parts, the six conditions¹, the signals and the four arguments I can’t finish alone are here as a working spec. If you want to kill one of them, the LinkedIn group is the place to do it: The Human Ledger - Measuring the conditions of work in the AI era.⁴