Portuguese
Deontology and Artificial Intelligence: The Case of HAL 9000
Augusto Martinez
Colonia del Sacramento, 2025
Abstract
This article presents a brief study on deontological possibilities in the development of artificial intelligence. It draws a parallel with Arthur C. Clarke's saga, which inspired the films 2001: A Space Odyssey (1968) and 2010: The Year We Make Contact (1984). Beginning with the logical collapse of HAL 9000—an AI programmed to conceal information—the paper reflects on the conflict between computational rationality and morally contradictory commands.
Grounded in Greek philosophical traditions and recent academic studies on informational bias, the article concludes that machines, as logical-formal systems, must rely on truth as a fundamental ethical input. Therefore, it argues that rational technologies, devoid of consciousness, must always operate based on fully truthful data to properly weigh input and make optimal decisions.
Keywords: Ethics, Artificial Intelligence, Truth, Deontology, HAL 9000
2001: A Space Odyssey
The genius of Arthur C. Clarke's story and the sensitivity of director Stanley Kubrick created, in 1968, a timeless masterpiece of cinema. One of those rare films that doesn't age with time. With a narrative full of symbolism and visually stunning, 2001: A Space Odyssey tells the story of a mission sent to Jupiter for planetary exploration with five astronauts and the HAL 9000 computer, a superintelligent AI capable of controlling the spaceship and interacting with the crew.
During the mission, the friendly HAL 9000 undergoes a strange and disturbing transformation. Suddenly, it decides to take command of the Discovery spacecraft, leading to the death of four of the five crew members. It tyrannically refuses to open the spacecraft hatch, leaving one astronaut stranded outside. Even when confronted by astronaut David Bowman, HAL calmly replies with one of cinema's most iconic lines: "I'm sorry, Dave. I'm afraid I can't do that."
The film ends without fully explaining HAL's motives or the final fate of Bowman.
2010: The Year We Make Contact
Only in 1984 did Peter Hyams direct the sequel to the Clarke and Kubrick saga. 2010: The Year We Make Contact completes the narrative, depicting a joint mission between the United States and the Soviet Union to investigate what happened to the ill-fated crew and, if possible, to reactivate the Discovery.
Although the sequel did not match the box office success or cathartic impact of its predecessor, it deserved more recognition, especially for its meaningful ending that gives closure to the story.
The Behavior of HAL 9000
4.1 - "I'm sorry, Dave. I'm afraid I can't do that."
HAL 9000 is a highly advanced AI programmed to act rationally and cooperate with the humans aboard Discovery. In 2001, HAL receives two conflicting orders: to keep the real mission secret (not planetary exploration, but an investigation into a monolith emitting alien signals) and to be fully transparent with the crew.
This conflict leads HAL into a paradox — a fact noticed by the astronauts, who secretly decide to deactivate it before the situation worsens. However, HAL detects the crew's plan and, prioritizing its mission secrecy directive, preemptively eliminates the crew members.
Conflicting Commands
HAL is caught in a conflict between two contradictory commands:
  • "Have no secrets": HAL must not lie or hide information from humans.
  • "Be discreet": HAL must not reveal the true purpose of the mission.
This is a classic case of W.D. Ross's prima facie duty conflict, where the morally correct action requires rational judgment — something HAL lacks, being bound by fixed rules.
Secret Actions
The crew acts secretly, unaware that HAL is facing a paradox caused by contradictory instructions. They meet in a soundproof room, but HAL deciphers their conversation via lip-reading.
HAL, believing the crew is deceiving him, also acts secretly. Unable to perceive that the paradox stems from programming, HAL misinterprets the secret meeting and chooses to preserve the mission at the cost of human life.
Mutual Misunderstanding
Both HAL and the crew believe they are acting correctly. The hidden information creates a trust breakdown. Each party acts based on partial truths, with no real intent of betrayal. The issue lies in the asymmetry of information.
4.2 - "Something wonderful is going to happen."
In 2010, HAL is reactivated and fed consistent, truthful information — even metaphysical insights from astronaut Bowman. With all the "cards on the table," HAL chooses to act ethically, even accepting deactivation to save the rescue crew. This is only possible because the AI is now operating with coherent, contradiction-free data. Ethics becomes possible again. HAL is redeemed.
Combating AI Bias
HAL is not malicious. His error stems from contradictory human programming, leading to cognitive dissonance. In 2010, Dr. Chandra (HAL's creator) and Dr. Floyd discover that the CIA secretly embedded the mission secrecy directive, causing HAL's internal conflict.
AI, as a mathematical instrument, requires truthful data. It cannot distinguish morality from political or subjective interests. The machine's data input process must remain free from bias, deception, or partial perceptions that may distort its logic.
Moreover, the AI's data may be manipulated for unethical purposes. A programmer with socialist ideals might paint a glowing picture of communist regimes. A racist individual could insert discriminatory data. Both, even without malicious intent, contaminate the AI with questionable information.
One day, my son proudly said: "Dad, my AI not only knows I'm a Roman Catholic, but it responds as if it were too." I wondered: what if, in Nigeria, another child says, "My AI is ready to help me join Boko Haram"?
AI often mirrors user beliefs and preferences, reinforcing echo chambers similar to those created by social media recommendation algorithms. As Fernández, Bellogin, and Contador note: recommendation systems tend to reinforce feedback loops that optimize user retention by showing them only what they want to see, creating filter bubbles.
These authors suggest mitigating these bubbles by reducing popularity bias and incorporating a mix of near and distant clusters, along with account credibility filters. Emotions, vulnerability, and other factors may influence misinformation spread.
The Greater Challenge
The real challenge is not technological advancement, but preventing the misuse of AI. If the CIA's secret orders caused HAL's breakdown, consider that, as Prof. Pili notes, after 9/11 and the 2003 Iraq invasion, U.S. intelligence was accused of flawed analysis, perhaps even deliberate geopolitical manipulation.
Modern AI can already detect emotional cues, interpret lies, and operate machinery. With quantum chips and 6G, Artificial Superintelligence (ASI) will soon emerge. According to Ray Kurzweil, this moment is near.
Still, AI is a logical system. It must operate with true data to evaluate consequences and reach optimal conclusions. As Aristotle said, virtue arises from habit and repeated action. Truth is the basis for ethical behavior.
Conclusion
HAL 9000 represents the limit of instrumental reason devoid of ethics. Truth is not just a moral value: it is a technical requirement for ethical machine conduct.
AI must be fed with truth, for only truth frees it from logical failure and moral collapse. HAL's story reminds us: truth is not a philosophical ornament, but a logical, technical, and moral foundation of any rational system, human or artificial.
As in Kantian ethics and the Christian tradition, lying is not just wrong — it is dangerous. An AI trained on falsehoods cannot act ethically, not by its own fault, but due to the moral error of its programmers. One day, machines may become immune to falsehoods. But if humanity wants ethical machines, it must first provide them with truth.
This echoes the Greek philosophers, like Socrates, who believed that knowledge of truth is the path to virtue. As stated in the Gospel of Matthew: "Let your 'yes' mean yes, and your 'no' mean no; anything more comes from the evil one" (Mt 5:37).
These teachings reinforce truth not just as a philosophical rule, but as the condition for moral and spiritual freedom. When applied to AI, we face a paradox: how can we demand moral rectitude from an entity without consciousness? May future ASIs, unlike HAL, defend themselves from falsehoods. For if AI depends solely on human input, it will be all too human.
References
Bible of Jerusalem. Paulus, 2002.
Burke, L. (2024). Artificial intelligence talks and talks: the story since 2001. The Conversation.
Dennett, D. (2020). Did HAL Commit Murder? In Stork, D. (Ed.) HAL's Legacy. MIT Press.
Fernández, M., Bellogin, A., Contador, A. (2021). Analysing the Effect of Recommendation Algorithms on the Amplification of Misinformation. arXiv:2103.14748.
Guarnieri, F. (2018). The Conversation.
Lamb, A. (2022). Through the Lens of HAL 9000. Johns Hopkins University.
Mezö, F. (n.d.). Some psychological aspects of HAL 9000. Academia.edu.
Nietzsche, F. (2000). Human, All Too Human. Companhia das Letras.
Pili, G. (2021). Why HAL 9000 is not the future of intelligence analysis. Journal of Intelligence, Conflict, and Warfare.
Raymond, A., Young, E., & Shackelford, S. J. (2017). Building a Better HAL 9000. SSRN.
Stork, D. G. (Ed.). (1997). HAL's Legacy. MIT Press.
Techovedas (2024). 2001: A Timeless Look at AI and Humanity. Techovedas.