To be human has always meant trying to see what comes next. From ancient hunters reading animal tracks to farmers scanning the skies for rain, survival has depended on the ability to anticipate events before they happen. Today, that instinct has been supercharged by artificial intelligence systems that promise not merely to forecast the future but to optimize it. Yet as predictive technologies seep into everything from hiring decisions to medical care and political messaging, a new wave of thinkers is asking whether society has surrendered too much authority to machines trained on the past.
An analysis published in MIT Technology Review explores how modern life has become saturated with automated predictions so pervasive that they are almost invisible. Algorithms now finish sentences, recommend purchases, determine creditworthiness, and filter the news people see. These systems, built to detect patterns in enormous datasets, constantly guess what individuals will do next, often shaping those choices in the process. The result is an environment in which prediction is no longer a tool but an infrastructure, quietly steering everyday experience.
In The Means of Prediction: How AI Really Works (and Who Benefits), economist Maximilian Kasy argues that most of these forecasts rely on supervised learning, a form of machine learning that draws conclusions from labeled historical data. Once trained, such systems can be applied to new cases, predicting outcomes like job performance, loan repayment, or criminal recidivism. These predictions increasingly guide institutional decisions, determining who gets hired, insured, or even released from prison.
Kasy contends that the consequences are not accidental byproducts of innovation but logical outcomes of systems designed to maximize efficiency and profit. If a social media algorithm amplifies outrage because it increases engagement, he argues, it is functioning exactly as intended. If hiring software filters out candidates likely to require caregiving leave or medical support, that too reflects economic incentives rather than technical error. In this view, prediction becomes a mechanism for narrowing possibilities, reinforcing existing inequalities while presenting itself as neutral computation.
Efforts to make algorithms more fair, Kasy suggests, cannot fully succeed because the data they rely on already encode historical biases. Training a model on flawed records risks automating discrimination at scale. His proposed remedy is not better code but democratic oversight of what he calls the “means of prediction,” including data ownership, computational infrastructure, and the policy frameworks governing their use. Ideas such as public data trusts and taxation models aimed at offsetting social harms could, he believes, rebalance power away from corporations that currently dominate predictive technologies.
While Kasy examines who benefits from prediction, The Irrational Decision: How We Gave Computers the Power to Choose for Us by Benjamin Recht explores how societies came to believe computers should make decisions at all. Recht traces the roots of modern algorithmic thinking to Enlightenment-era rationalism and, more decisively, to the mathematical frameworks refined during World War II. Faced with urgent wartime calculations about risk, logistics, and strategy, scientists developed formal decision-making models that treated uncertainty as something quantifiable and manageable.
These models later shaped the design of digital computers, embedding the assumption that rational calculation could outperform human judgment. Over time, optimization theory and statistical reasoning migrated from military planning into business operations, governance, and personal technology. Today, Recht argues, many systems treat life itself as a sequence of cost-benefit analyses, reducing complex moral or social questions to data points that can be optimized.
He challenges the idea that human progress required such mechanized rationality. Major advances in medicine, infrastructure, and democratic governance, he notes, predated modern computing and were often driven by intuition, experimentation, and ethical debate rather than mathematical optimization. By elevating prediction above all other forms of understanding, societies risk sidelining qualities that cannot be easily quantified, including empathy, cultural context, and moral responsibility.
A third perspective comes from Prophecy: Prediction, Power, and the Fight for the Future, from Ancient Oracles to AI by philosopher Carissa Véliz, who frames prediction as a force that actively shapes reality rather than merely describing it. Forecasts, she writes, can function like magnets, pulling events toward their anticipated outcomes. When widely believed, they become self-fulfilling, guiding investment, research priorities, and public expectations.
The history of computing offers a vivid example in Gordon Moore, whose observation that transistor density would double roughly every two years became known as Moore’s Law. The prediction did not simply forecast technological growth; it galvanized an entire industry to meet the expectation. Companies, including Intel, invested billions to maintain the pace, turning a projection into a roadmap that structured decades of innovation.
Véliz warns that predictive narratives can also distract from urgent present-day problems. Grand promises about future artificial general intelligence, for instance, may draw attention away from the tangible harms already associated with automated decision-making, such as surveillance, labor displacement, and entrenched inequality. The authority to predict, she argues, is inseparable from the authority to define priorities and behaviors.
Across these differing analyses runs a shared concern: prediction is never just technical. It is political. Decisions about what to measure, which data to collect, and how to interpret probabilities determine who gains opportunity and who is excluded. Systems presented as objective often carry embedded values reflecting the institutions that created them.
The rise of predictive technology also coincides with declining public trust in institutions, raising questions about who should govern tools that increasingly mediate social life. Kasy’s call for democratic stewardship, Recht’s appeal to restore human judgment, and Véliz’s critique of predictive power all suggest that the future of AI will hinge less on engineering breakthroughs than on collective choices about authority and accountability.
Technology, these scholars emphasize, is not destiny. Algorithms may forecast behavior, but societies decide whether to follow those forecasts, regulate them, or resist them altogether. The fundamental human act of imagining the future remains intact, even as machines attempt to automate it.
The challenge now is not whether predictions can be made more accurate. It is whether humanity can reclaim the ability to decide which predictions matter, and when to ignore them.
