I’ve long found fiction to be one of the most fruitful ways to explore technology. This essay series grows out of my love for the humanities—as a consumer and creator.
SPOILER ALERT: My discussion involves spoilers, so consider yourself warned.
AI predictions (or more generally known as algorithmic predictions), are becoming increasingly common. As applied to human behavior, AI predictions are deeply problematic. It’s hard to imagine a discussion about AI predictions of human behavior that doesn’t reference Minority Report. The 2002 movie by Stephen Spielberg adapts a 1956 story, “The Minority Report” by Philip K. Dick. Both of these works contribute significantly to the debate about the ethics of AI predictions of human behavior.
Minority Report
This movie was based on a story by Philip K. Dick called “The Minority Report” published in 1956. It was directed by Stephen Spielberg and starred Tom Cruise, Colin Farrell, Samantha Morton, and Max von Sydow.
The movie is set in 2054. The protagonist, John Anderton, works in a special law enforcement unit (Precrime) that uses psychics — called “precogs” — to predict future crimes. People are arrested before the crime occurs.
My favorite parts of the movie involve some of the smaller diversions. The movie depicts a future surveillance society that is not cold and drab like Orwell’s world of Big Brother. This world is filled with flashing screens and an endless barrage of stimulation. The surveillance isn’t just by the government but also by businesses.
The movie depicts the risks of biometric identification. Iris scanners are everywhere, and when the protagonist walks by a TV monitor, he is addressed by his name and a targeted ad is delivered to him. Anderton must replace his eyes to evade biometric surveillance.
The main action of the movie is precipitated when Anderton, who runs the precrime unit, discovers that the precogs have predicted he will murder a man (Leo Crow). He doesn’t initially know why, but because the predictions are so accurate, he flees before he can be arrested.
The movie explores the theme of the limitations of prediction and free will. The precogs who had a vision of the murder had a correct vision, but the movie shows that the vision can be interpreted in different ways.
Anderton’s killing of Crow was a set up to create the conditions for Anderton to want to kill Crow by making Anderton think Crow was the man who kidnapped and murdered his son.
Crow wanted Anderton to kill him so his family could get a big payout. Anderton was intent on killing Crow but then changed his mind. Crow and Anderton scuffled, and the gun went off, killing Crow. The precogs saw the vision, but the key was how it was interpretated and what was going on in the minds of Anderton and Crow. The movie shows the limitation of prediction, which foretold only part of the story. The precog predictions were incomplete because they didn’t reveal Anderton’s mental state, which is a key component for a crime. The prediction only foretold the act, but crimes require both an act (actus reus) and a mental state (mens rea).
As Hideyuki Matsumi and I wrote in our article, The Prediction Society:
The movie depicts a dystopian future, where people are punished before they do anything wrong. People are at the mercy of predictions about their future, with no ability to contest them. When Anderton learns about the prediction that he will commit murder, he doesn’t try to argue because he knows it is futile. Instead, he runs.
We are not yet living in the harrowing world of Minority Report, but we are well along the path toward it—and in some cases, we are coming uncomfortably close to it. Every day, incarceration decisions are made based on algorithmic predictions of future crimes. Every day, people are denied jobs or loans based on predictions about things they haven’t yet done. The escalating collection of vast quantities of personal data, gathered from a burgeoning number of connected devices, is being fed into more and more algorithms, generating countless predictions on a scale hitherto unimaginable.
Today, algorithmic predictions are being used with increasing frequency, in an ever-expanding range of domains, with too much confidence, and not enough accountability. Ironically, future forecasting is occurring with far too little foresight. . . .
The predictions [in Minority Report] are highly accurate—far more accurate than a criminal trial might be. Yet, even so, punishing people for crimes they haven’t yet committed crosses an ethical line and is fundamentally at odds with basic concepts of fairness. It is hard to imagine scenarios where it would be ethically acceptable to punish a person for future predicted wrongdoing. . . .
The use of algorithmic predictions for matters involving humans is fraught with problems. On the surface, these predictions seem beguiling. They gleam with the promise of higher accuracy and less bias than human predictions. But algorithmic predictions are, in reality, power dressed up with math. They are used not to see the future but to shape it. They draw from a particular construction of the past and aim to give it an iron grip on the future. The entities using algorithmic predictions are not predicting the future to understand it but to control it.
When entities gain control over people’s future, people lose control. Organizations make decisions and choices based on predictions, and people’s own decisions and choices no longer matter.
In life imitating fiction, today’s AI predictions are used to punish people for future crimes they may or may not commit. Courts widely use algorithmic prediction tools to generate risk scores for recidivism, which impacts the length of criminal sentences.
In Wisconsin v. Loomis, 881 N.W.2d 749 (Wis. 2016), a defendant challenged the use of an algorithmic prediction that he had a high risk of committing future crimes, leading to a significantly longer sentence. He argued that the prediction violated his Constitutional right to due process because he didn’t receive an “individualized sentence” because the prediction was based on recidivism statistics of other people (predictions assume that people will behave the same as other people with similar characteristics). The Wisconsin Supreme Court rejected the defendant’s due process challenge because the trial judge had the ultimate discretion to decide the defendant’s sentence.
I find this case quite troublesome, and Minority Report shows why. A person was punished based on crimes he didn’t yet commit. In fact, it’s worse than the movie, where psychics actually saw visions of people committing crimes. In Loomis, the prediction was based not on a vision of the future, but on past data involving other people, not the defendant. In other words, the kind of AI predictions used in criminal sentencing punish people based on crimes that other people have committed.
More broadly, AI predictions diverge from those in Minority Report because they are not really predictions like those of an oracle or psychic, who purportedly “see” the future.
AI predictions rest on the assumption (which isn’t always true) that patterns in the past will persist into the future. What they see is the past—they look backward.
Philip K. Dick’s Story, “The Minority Report”
Philip K. Dick’s story “The Minority Report” was originally published in Fantastic Universe in 1956.
It has held up remarkably over the past 70 years.
Although Spielberg’s 2002 movie takes many elements from Dick’s story, there are several key differences.
First, the murder victim in the story is General Leopold Kaplan, who aims to discredit the Precrime unit and carry out a military coup. Anderton kills him to stop his plan and protect Precrime. In the movie, the victim is a man named Leo Crow who is part of a plot to frame Anderton. Crow impersonates a pedophile who killed Anderton’s son and tries to entice Anderton to kill him. Unlike the story, Anderton doesn’t commit murder. He makes a last-minute decision not to do so, manifesting his free will. Crow and Anderton scuffle, and the gun goes off killing Crow. The precog visions were correct, but they didn’t account for Anderton’s lack of intent, which is an essential element of murder.
Unlike the movie, the story focuses on multiple time-paths. Precog reports are often not unanimous (hence the “minority report” in the title). The reports diverge based on the possibility of Anderton changing his mind upon seeing the reports and thinking about them. The first report predicts he’ll murder Kaplan. The second report concludes he’ll change his mind after reading the first report. The third report predicts Anderton’s final decision. The story shows that prediction doesn’t just foretell the future but actively shapes it. Prediction can become a self-fulfilling prophecy.
In the movie, all the precogs predict the killing. As Robert Batey observes, the lack of a true minority report in the movie is “the ultimate betrayal of Dick’s story, as the possibility of a minority report becomes little more than a red herring in the movie.” Dick’s story is more ambiguous about the existence of free will than the movie.
For Dick, the different reports account for the changes that the other reports cause, and although Anderton alters the course of his actions, it all might be pre-ordained.
According to Robert Batey, “Perhaps the short story is best understood as a depiction of one aspect of Werner Heisenberg’s renowned uncertainty principle: The fact of measurement always alters the item being measured.”
Batey provides fascinating insight about the genesis of the movie and its departures from Dick’s story. He quotes from Gary Goldman, who wrote the original script for the movie before it was rewritten by Scott Frank and Jon Cohen :
The movie . . . doesn’t go to the roots of Phil Dick’s story. . . . The basic sentiment of the movie is that the U.S. Constitution and our current ideas of civil rights are more important than having absolute truth . . . . These are good lessons but not what Philip K. Dick was writing . . . . Steven [Spielberg] took it as a given that there had to be free will, that the system was bad because it violated the constitution . . . . Dick was willing to question everything.
Ethics, Free Will, and Dostoyevsky
The movie takes a stronger moral stance than Dick’s story about the ethics of prediction; in the movie, it becomes clear that punishing people based on predictions of future crimes is bad. But both the story and the movie illustrate a terrifying dystopian world where our agency is lost, usurped, or perhaps even non-existent.
In Fyodor Dostoyevsky’s Notes from Underground (1864), the underground man raises concerns about a future day when “all human actions” will be “calculated . . . like a table of logarithms.” He complains that “everything will be so precisely calculated and designated that there will no longer be any actions or adventures in the world.”
Interestingly, the underground man suggests that if people lack free will, then they can no longer be held accountable for their behavior. In The Brothers Karamazov (1880), Dostoyevsky further develops this concept through Ivan Karamazov, though less as a problem of free will and more as one of divine forgiveness for sin. Ivan struggles with a tension in his views—he wants people to be punished, not forgiven, but he also believes nobody is to blame for their sins: “What do I care that none are to blame and that I know it—I need retribution, otherwise I will destroy myself.”
Minority Report takes a different approach to the question of culpability and free will, where the potential lack of free will doesn’t eliminate being held responsible for one’s behavior. While Ivan can’t accept a world where people aren’t punished for terrible wrongdoings, the world of Minority Report, where people are punished before they’ve even done anything wrong, is also unacceptable.
Whether there’s ultimately free will or not, there are problematic consequences to embracing its non-existence.
The Dangers of Al Predictions About Human Behavior
AI predictions (or algorithmic predictions) raise several problems that the literature illuminates. In our article, Matsumi and I identify at least four problems with algorithmic predictions:
The Fossilization Problem. Algorithmic predictions reify certain facts from the past by casting them into the future, making the past persist and harder for people to escape from the past.
The Unfalsifiability Problem. Algorithmic predictions are often un-verifiable because the events they are predicting haven’t yet occurred. Algorithms can’t be challenged for falsity, which is the law’s main vehicle for allowing individuals to contest algorithms.
The Preemptive Intervention Problem. When preemptive decisions or interventions are made based on future forecasting, the feedback loop to assess whether or not the forecasting was accurate dissipates, making it difficult or impossible to evaluate the accuracy of a prediction.
The Self-Fulfilling Prophecy Problem. Algorithmic predictions can turn into a self-fulfilling prophecy because decisions based on them further what they predict.
Matsumi and I are not the only ones concluding that AI predictions about human behavior are irreparably problematic and should be should not be used in most circumstances.
Arvind Narayanan and Sayash Kapoor argue: “Accurately predicting people’s social behavior is not a solvable technology problem, and determining people’s life chances on the basis of inherently faulty predictions will always be morally problematic.” Additionally, they note that the “cost of flawed AI is not borne equally by all. The use of predictive AI disproportionately harms groups that have been systematically excluded and disadvantaged in the past.”
Katrina Geddes argues that AI predictions ignore “the input of the underlying individual.” Instead of relying on individuals as a source of “information about their intentions, motivations, and moral capabilities,” AI predictions treat “data subjects not as unique individuals, but as patterns of behavior.” When used for determining criminal punishment, “algorithmic prediction effectively punishes the underlying individual for membership of a statistical group.”
Carissa Véliz contends that generative Ais “are fortune tellers, as opposed to truth tellers, and the ultimate prediction machine is also the ultimate bullshitter. . . . [A]lthough predictions appear to be descriptive claims, they are in fact veiled prescriptive assertions—they tell us how to act.”
Here’s a cartoon I created a while ago that depicts the due process problems of algorithmic predictions. In many cases, they are impossible to disprove.
Bibliography
Robert Batey, Minority Report and the Law of Attempt, 1 Ohio St. J. Crim. L. 689 (2004):
Philip K. Dick, “The Minority Report” (1956)
Fyodor Dostoevsky, Notes from Underground (Richard Pevear & Larissa Volokhonsky trans., 1993) (1864)
Andrew Guthrie Ferguson, Policing Predictive Policing, 94 Wash. Univ. L.R. 1109, 1112 (2017)
Katrina Geddes, The Death of the Legal Subject: How Algorithms Are (Re)constructing Legal Subjectivity, 25 Vand. J. Ent. & Tech. L. 1, 3 (2023)
Hideyuki Matsumi, Predictions and Privacy: Should There be Rules About Using Personal Data to Forecast the Future?, 48 Cumb. L. Rev. 149 (2017)
Hideyuki Matsumi & Daniel J. Solove, The Prediction Society: Algorithms and the Problems of Forecasting the Future, 2025 U. Ill. L. Rev. 1 (2025)
Minority Report (Steven Spielberg director 2002)
Arvind Narayanan & Sayash Kapoor, AI Snake Oil: What Artificial Intelligence Can Do, What It Can’t, and How to Tell the Difference (2024)
Daniel Solove, On Privacy and Technology (2025)
Carissa Véliz, Prophecy: Prediction, Power, and the Fight for the Future, from Ancient Oracles to AI (2026)
Other Notable Things
Stage Adaptation of Minority Report
There was a stage adaptation of Minority Report.
Carissa Véliz gave a brilliant TED Talk, Beware the Power of Prediction
Daniel J. Solove is the Bernard Professor of Intellectual Property and Technology Law at the George Washington University Law School. He is the founder of TeachPrivacy, a company that provides workforce privacy, cybersecurity security, and AI training to companies and organizations around the world. He is the author of 10+ books and 100+ articles.
He was written a children’s fiction book about privacy, The Eyemonger, and is currently writing Black Mirror style short stories. Be on the lookout when (or if) they are published.
You can follow his events, writings, training, cartoons, and resources by subscribing to his free weekly newsletter.
Thanks for the shout out :)