The Ethics of AI
The Mirror That Thinks: What AI Reveals About Moral Responsibility
When an algorithm denies someone a loan, a job, or a medical diagnosis, the question isn't whether the machine was wrong — it's whether anyone is still responsible for being right.
The Idea
Moral responsibility has always depended on a simple assumption: somewhere in the chain of cause and effect, there is a person who chose. Law, ethics, and our deepest intuitions about fairness all rest on this. But AI systems are quietly dissolving that assumption in ways we haven't fully reckoned with. This isn't the familiar worry about robots taking jobs or superintelligences going rogue. It's subtler and already here. When a hospital deploys a predictive model to triage patients, or a court uses risk-scoring software to influence sentencing, the decision is real and consequential — but the reasoning is often opaque even to its creators. Philosophers call this the 'problem of many hands': when responsibility is distributed across engineers, data scientists, procurement officers, and executives, it can evaporate entirely. Everyone did their part; no one owns the outcome. What makes this genuinely philosophically new is that AI doesn't just automate decisions — it launders them. A human judge who makes a biased ruling can be questioned, can reflect, can change. A model encodes its biases invisibly, at scale, with the appearance of objectivity. The outputs arrive wearing the costume of mathematics, which we culturally associate with neutrality. But a model trained on historical data doesn't describe the world as it should be — it predicts the world as it was, and then helps make it so again. The ethical question isn't just 'is this AI fair?' It's 'who is obligated to ask?'
In the World
In 2016, a journalism outlet called ProPublica published an investigation into a software tool called COMPAS, widely used in American courts to assess the likelihood that a defendant would reoffend. The algorithm's outputs were influencing real sentencing decisions — how long someone spent in prison. ProPublica's analysis found that Black defendants were nearly twice as likely as white defendants to be incorrectly flagged as high risk when they went on to commit no further offence. The company behind COMPAS disputed the methodology; statisticians argued about which definition of 'fairness' was even the right one to apply. The debate remains unresolved. What is not disputed is this: judges were being handed a number, generated by a proprietary model whose inner workings they could not inspect, and that number was shaping decisions about human freedom. None of the individuals involved — the judges, the software company, the court administrators — saw themselves as the responsible party. The judge deferred to the data. The company said it was a tool, not a decision-maker. The administrators said they were following approved procedure. Nobody chose the outcome. Or rather, everybody did — just in fragments small enough to feel blameless. This is what the philosopher Helen Nissenbaum meant when she wrote about how computerisation erodes accountability: it doesn't remove humans from the loop so much as it gives every human in the loop a reason to believe they're not the one who counts.
Why It Matters
Most of us will never design an algorithm or sit on a sentencing panel. But we live in systems shaped by both. And the habits of mind that let institutions off the hook — the deference to data, the comfort of procedure, the assumption that complexity is the same as neutrality — are not confined to courtrooms or hospitals. They show up wherever we stop asking who decided and why. Mindfulness, in its most practical sense, is about noticing what we'd rather not notice. Applied to technology, that means resisting the sedation that comes with smooth interfaces and confident outputs. It means asking, when a recommendation feels authoritative, whose values were baked into its training. It means holding space for the possibility that 'the system said so' is not, and has never been, a complete moral answer. The philosopher of technology Albert Borgmann once wrote about how devices conceal the processes that produce them. AI is perhaps the most extreme version of this concealment ever built. Which makes the practice of asking — insistently, even inconveniently — not just intellectually honest but genuinely ethical.
A Question to Ponder
When you defer to a system — a recommendation, a score, an automated decision — are you trusting the output, or are you quietly relieved that the responsibility for being wrong belongs to someone else?
Get a new one of these every morning.
Start learning with Thinkable