Information. Reputation. Leverage.
Build, verify and carry a Reputation Record shaped by evidence, responses and resolution — not anonymous stars.
AI is entering healthcare, criminal justice and risk assessment — but some decisions, like abortion and capital punishment, aren't just prediction problems. RRSource Paper No. 005 asks where AI's usefulness ends and human moral responsibility must begin.
RRSource Paper No. 005
Is humanity too human for AI logic?
Artificial intelligence is increasingly being used to analyse health information, predict risk, identify patterns, support decision-making and automate processes that once required human judgment.
But there are some decisions for which the consequences are fundamentally human.
And, at the furthest boundary of state power:
Who lives and who dies?
This paper is not arguing that today's AI systems are secretly deciding who should have an abortion or who should be executed. They are not.
The more important question is what happens when increasingly sophisticated AI decision-making systems are introduced into areas where the underlying question is not merely what is probable? but what is morally permissible?
That distinction could become one of the defining questions of AI ethics.
Artificial intelligence is already being used across healthcare, criminal justice, fraud detection, identity verification, risk assessment and other high-consequence environments.
The World Health Organization says AI is already transforming sexual and reproductive healthcare, including applications involving screening, prediction, pregnancy and access to information. At the same time, WHO warns about bias, privacy, misinformation, unequal access and insufficient transparency.
In April 2026, WHO also introduced ChatHRP, an AI-assisted system designed to retrieve verified sexual and reproductive health information for policymakers, researchers and health workers. Importantly, the system is deliberately built around authoritative evidence and references rather than asking an AI model to invent answers.
That distinction is significant.
There is a considerable difference between:
AI helping humans find evidence
and
AI deciding what humans should do.
The first can increase human capability.
The second raises questions about authority.
Consider abortion.
A medical AI system could potentially process enormous amounts of information concerning pregnancy:
This can be extremely useful.
But medicine is not simply mathematics.
WHO describes abortion as a common health intervention and states that it can be safe when performed using recommended methods appropriate to pregnancy duration and by people with the necessary skills. WHO also identifies lack of access to safe, timely and respectful abortion care as a public-health and human-rights concern.
Yet the question of abortion extends beyond clinical prediction.
It involves competing claims concerning:
AI can analyse these arguments.
It cannot make the disagreement disappear.
This is where the idea of AI logic becomes complicated.
Suppose an AI knows:
At 10 weeks, a particular medical intervention carries a particular probability of outcome.
That is an empirical question.
But suppose we ask:
Should the intervention be permitted?
That is no longer merely an empirical question.
The answer depends upon a moral or legal framework.
Someone who believes bodily autonomy should be given primary importance may reach one conclusion.
Someone who believes protection of unborn human life should take priority may reach another.
Someone else may believe the answer changes according to gestational age or medical circumstances.
The AI cannot resolve this disagreement merely by becoming more computationally powerful.
It would need a normative framework — a set of assumptions about what ought to matter.
And those assumptions ultimately come from somewhere.
This is one of the central problems of AI ethics.
If an algorithm produces a decision, we should ask:
These are not hypothetical concerns.
UNESCO's global Recommendation on the Ethics of AI places human dignity, human rights, transparency, fairness, accountability and human oversight at the centre of responsible AI governance. It explicitly states that AI systems should not displace ultimate human responsibility and accountability.
NIST's AI Risk Management Framework similarly treats trustworthy AI as involving validity, reliability, safety, security, accountability, transparency, explainability, privacy and fairness, with harmful bias managed throughout the AI lifecycle.
In other words:
The emerging international consensus is not "let AI decide." It is closer to: use AI, but retain responsibility.
The death penalty presents an even more extreme version of the problem.
A criminal justice system already makes predictions.
Algorithms and statistical risk assessments can potentially assist with some of these questions.
But there is an enormous difference between predicting risk and determining the moral legitimacy of execution.
A 2007 Oxford Academic study demonstrated how an artificial neural network could predict execution outcomes among a historical sample of 1,366 death-row inmates. The researchers argued that the system's predictive performance raised serious questions about the apparent arbitrariness of capital punishment.
That is fascinating.
But notice what the machine was doing.
It was predicting human decisions.
It was not demonstrating that execution was morally justified.
That distinction matters.
Imagine an algorithm that predicts with extraordinary accuracy that a court is likely to impose the death penalty in a particular type of case.
It might identify patterns involving:
The prediction could be statistically impressive.
But it doesn't answer:
Should the death penalty be imposed?
Prediction describes what is likely to happen.
Justice asks what ought to happen.
Those are fundamentally different questions.
This becomes particularly important as AI enters criminal justice.
A person can be transformed into a collection of variables:
The system then produces:
The label appears objective.
But it is actually the final output of a chain of human decisions about what information matters.
NIST's AI Risk Management Framework explicitly calls for organizations to identify and document the impacts of AI on individuals, groups, communities and society, and to establish processes for human oversight.
The question is therefore not merely:
Is the algorithm accurate?
It is:
Accurate about what?
Consider this statement:
This individual has a 72% probability of reoffending.
It sounds scientific.
But the individual did not commit a future crime.
The number represents a prediction.
If society begins treating the prediction as though it were a fact, the distinction between risk and guilt becomes dangerously blurred.
The person stops being judged solely for what they did.
They begin to be judged for what an algorithm believes they might do.
That is the philosophical territory of pre-crime.
And it is precisely where AI-assisted risk assessment demands scrutiny.
A false recommendation for a film can be corrected.
A bad shopping recommendation can be ignored.
A chatbot can give a poor restaurant suggestion.
But some decisions cannot be casually reversed.
These involve fundamentally different levels of risk.
NIST's framework therefore treats AI risk as something that must be governed according to context, potential impacts and consequences rather than treating every AI application as equivalent.
This suggests an important principle:
The more irreversible the consequence, the less acceptable it is to hide human responsibility behind automation.
Here the discussion moves from existing applications into a genuinely future-facing question.
A 2026 paper in The Journal of Ethics examines the possibility of integrating AI into capital punishment through what its authors call "lethal autonomous mechanisms." The paper is explicitly anticipatory: it does not claim that autonomous execution is an established reality, but examines how such systems could theoretically become part of a future "death row network."
The authors identify a profound ethical problem.
Automation might reduce some of the human labour involved in carrying out executions.
But it would not eliminate moral responsibility.
It would redistribute that responsibility across the people designing, deploying, maintaining and authorising the system.
That produces a chilling question:
If a machine carries out an execution, who has actually executed the person?
The machine cannot appear in court and accept responsibility.
AI can be extraordinarily good at finding patterns.
But morality is not merely pattern recognition.
Suppose two people commit identical crimes.
One has spent twenty years helping victims, raising a family and living without further offences.
The other has continued offending.
A purely historical system might treat their past as a fixed attribute.
A human justice system can, at least in principle, consider mitigation, rehabilitation, remorse and change.
That is why the previous RRSource paper asked:
If AI does not forget, who does it forgive?
The question becomes even more serious when AI moves from remembering to judging.
This is not simply a religious argument.
Religious traditions have historically developed concepts of:
Secular justice systems have their own concepts:
The details vary enormously between jurisdictions and traditions.
But the underlying idea is recognisable:
A person's past does not necessarily have to determine their entire future.
That idea is difficult to encode into systems whose primary function is prediction.
Perhaps the question is badly framed.
AI does not necessarily need to understand redemption.
It needs to be prevented from accidentally erasing the possibility of redemption from the decision-making process.
That is a different engineering problem.
An AI system could record:
What happened — and also what happened afterwards.
It could distinguish:
Allegation — from verified finding.
Historical incident — from current pattern.
Risk — from certainty.
Information — from judgment.
That distinction is central to trustworthy reputation intelligence.
Consider a hypothetical reputation report.
A simplistic system might produce:
High risk.
That tells the user almost nothing.
A more responsible system could provide:
What has been reported.
What can be substantiated.
What the subject said or did.
Whether the information has been challenged.
Whether the matter was resolved.
How old the information is.
Whether similar events continued.
What is known today.
That is not an attempt to make an algorithm morally superior to humans.
It is an attempt to give humans better information before they make consequential decisions.
This distinction is fundamental to the RRSource proposition.
RRSource should not exist to say:
Good person.
Or: bad person.
Nor should a reputation report become an automated sentence.
Its value should come from creating a more complete reputation record.
A user should be able to investigate:
That is trust intelligence rather than artificial morality.
There is a subtle but important distinction.
AI can help humans process information that would otherwise be impossible to analyse at scale.
But the existence of an AI recommendation should never make responsibility disappear.
UNESCO's framework explicitly calls for AI systems to remain subject to human responsibility and accountability.
Perhaps the title of this paper asks the wrong question.
Perhaps humanity isn't too human for AI logic.
Perhaps human beings are simply more complicated than logic alone can capture.
We have competing values.
We disagree.
We change our minds.
We make mistakes.
We forgive.
We seek revenge.
We show compassion.
We demand justice.
Sometimes we even contradict ourselves.
That isn't necessarily evidence of irrationality.
It is evidence that human morality contains competing principles.
This is where things become particularly interesting.
Imagine an AI that says:
Based on the available evidence, this decision will produce the greatest statistical benefit.
But a human says:
I don't care. It violates a fundamental right.
Which one wins?
If the AI wins, we have effectively granted it moral authority.
If the human wins, the AI remains a tool.
That distinction should be explicit.
Otherwise, responsibility becomes ambiguous.
The future conversation about AI should therefore not only ask:
What can AI do?
It should also ask:
What should AI never be permitted to decide alone?
UNESCO's principles include human oversight and determination, fairness and non-discrimination, transparency, accountability and human dignity.
WHO's guidance similarly stresses that AI in health should be developed around ethics, human rights, accountability and public benefit.
The emerging lesson is clear:
Capability is not authority.
Just because a machine can make a prediction does not mean society should give it the power to determine the outcome.
RRSource is ultimately concerned with something slightly different from the question:
Can AI make better decisions than humans?
The more important question is:
Can humans make better decisions when AI gives them better information?
That is a much more defensible proposition.
AI can process. Humans can judge.
AI can remember. Humans can contextualise.
AI can identify patterns. Humans can consider meaning.
AI can calculate probabilities. Humans can decide what those probabilities are worth.
Artificial intelligence is going to enter more areas of human life.
It will influence healthcare.
It will influence finance.
It will influence policing.
It will influence employment.
It will influence identity verification.
It will influence risk assessment.
It will influence reputation.
And potentially, over time, it will influence decisions involving fundamental questions of life, liberty and punishment.
The answer should not be to reject AI.
Nor should it be to worship its apparent objectivity.
We should build systems that understand the difference between:
And perhaps most importantly:
Logic and morality.
Because an AI can tell us what is likely.
It can tell us what happened before.
It can tell us what patterns exist.
It can even tell us what humans have historically decided.
But when the question becomes:
What should happen to a human being?
we enter territory where the answer cannot simply be hidden inside a probability score.
AI can calculate. AI can predict. AI can remember.
But humanity must remain responsible for what happens next.
The future of reputation intelligence should not be an automated morality machine.
It should be an evidence and context layer that helps humans make better decisions.
A reputation report should not simply tell you:
Trust. Or: do not trust.
It should help you understand why.
That means evidence, verification, disputes, resolution, recency, patterns and context.
Because if AI is going to help humanity make increasingly consequential decisions, then perhaps the most important thing we can build is not an AI that judges people.
It is infrastructure that helps people judge information better.
That is the difference between a machine that makes decisions and a system that makes human decisions more informed.
This paper deliberately distinguishes current AI applications from future or hypothetical AI authority. WHO documents current AI use in sexual and reproductive health primarily around screening, prediction and information, while warning about bias, privacy, misinformation and transparency. The capital-punishment discussion likewise separates historical AI research predicting death-penalty outcomes from the much newer, explicitly hypothetical literature on autonomous execution.
That distinction matters: this paper is not sensationalising AI as already deciding who lives and dies; it is examining the governance problem before technological capability becomes moral authority.
Paper 005 · August 2026 · By Logic
Previous
Someone Posted an Unfair Review About You. Here's How to Respond Without Making It Worse
Next
Digital Discrimination: Automation, Minority Rights and the New Politics of Trust
Loading comments…
AI companies are pouring unprecedented computing power into machines that can learn, reason and increasingly contribute to the development of their own successors. Humans have something processors do not: generations of accumulated experience built through work, failure, relationships and consequence. As machine intelligence compounds faster, the deeper question is no longer simply whether AI will become smarter — but whose experience will shape what comes next.
A criminal record may document what someone did. The harder question is whether society allows it to stop defining who they are.
Years of expensive housing, rising everyday costs and economic uncertainty are changing the Canadian mood. The numbers suggest something deeper than a temporary bout of frustration: for many households, the margin for absorbing another shock is becoming increasingly thin.
