Artificial intelligence is increasingly used to help companies decide whom to hire, whom to lend to and which customers deserve closer scrutiny. Governments and businesses can use algorithms to process large amounts of information faster than any individual could, which makes automated decision-making attractive in areas where speed and efficiency matter.
But efficiency becomes more complicated when an algorithm makes a mistake.
If an AI system rejects a qualified job applicant, denies someone access to credit or treats a person unfairly because of biased data, responsibility cannot simply be assigned to the machine. Economist Jesus Felipe argues that this is one of the fundamental questions societies will have to confront as artificial intelligence becomes more deeply embedded in economic life.
Felipe, Distinguished Professor at the Carlos L. Tiu School of Economics of De La Salle University, discusses the issue in his study, Could Pope Leo XIV and Karl Marx See Eye to Eye? Comments and Questions on MAGNIFICA HUMANITAS. The paper examines Pope Leo XIV’s encyclical Magnifica Humanitas, particularly its arguments about artificial intelligence, technology, employment and the economic system.
One of the encyclical’s central arguments, as Felipe summarizes it, is that AI cannot be treated as a purely technical issue once it begins to affect people’s lives.
Automated systems are already involved in decisions concerning employment, credit, public services and reputation. While these systems can improve efficiency, the encyclical warns that they can also create new forms of exclusion, reproduce stereotypes and make harmful decisions difficult to challenge.
An algorithm is not neutral
A common assumption is that computers may be less biased than humans because they rely on data rather than emotion. The problem is that the data itself reflects the world from which it was collected, while the design of a system reflects choices made by people.
Felipe summarizes Pope Leo XIV’s argument that technology is “never neutral” because it reflects those who design, finance, regulate and use it.
The same principle applies to AI models. Decisions must be made about what information a system should measure, which outcomes it should optimize and how different people or situations should be classified. Those choices can matter greatly in financial services.
Suppose a bank uses an AI model to assess whether someone deserves a loan. The system may examine income, employment history, spending behavior and other indicators that appear objective. But if the underlying data contains historical inequalities or the model relies on variables that indirectly disadvantage certain groups, the resulting decision can reproduce those inequalities at scale.
The same risk arises in recruitment. A company may use software to screen thousands of applications and identify the candidates most likely to succeed. Yet if the model learned from previous hiring decisions that contained biases, automation does not necessarily remove those biases. It can simply make them faster and less visible.
Felipe’s paper notes that AI systems may “reflect and reinforce stereotypes or ideological bias,” while decisions about what an algorithm measures, ignores and optimizes inevitably embody choices and priorities.
Someone still has to answer for the decision
The problem becomes harder when nobody fully understands how a particular result was produced.
The encyclical notes that even the people who develop advanced AI systems may have only a limited understanding of how those systems arrive at particular outputs. Felipe summarizes the Pope’s description of modern AI as being more “cultivated” than “built,” since developers construct the environment in which the system learns rather than programming every individual decision.
Felipe’s study stresses the need to identify responsibility at every stage of AI development and use. Someone must be able to account for a decision, explain it, monitor its consequences, allow affected people to challenge it and correct the harm when something goes wrong.
This creates an important principle for businesses: delegating a decision to software does not mean delegating accountability.
A bank cannot reasonably argue that a customer was rejected because “the algorithm said so.” Neither should an employer be able to hide behind an automated hiring system when a candidate is unfairly excluded.
The company still selected the technology, determined how it would be used and decided how much authority it would receive.
AI could create a new kind of information gap
Automated decision-making also changes the balance of power between institutions and individuals.
A company may have access to enormous datasets, sophisticated algorithms and teams of specialists who understand how those systems operate. The person affected by the decision may know little more than whether an application was accepted or rejected.
This asymmetry becomes particularly troubling when the decision affects someone’s livelihood or financial future.
Felipe’s study highlights the encyclical’s call for transparency and accountability when algorithms influence credit, personnel selection, services or economic opportunities. It also argues for legal frameworks, independent oversight and informed users so that technological change is not controlled solely by those who possess data, infrastructure and computing power.
The challenge is to preserve the efficiencies AI can provide without allowing complicated systems to become a convenient shield against responsibility.
Companies will undoubtedly continue to automate decisions because the economic incentives are powerful. AI can process applications more quickly, detect patterns that humans might miss and reduce the cost of handling large numbers of transactions.
But the more consequential the decision, the harder it becomes to argue that human judgment should disappear completely.
An algorithm can recommend whether someone should receive a loan or advance to the next stage of a job application. It can identify risks and rank candidates. What it cannot do is accept moral responsibility for the consequences.
As AI gains more influence over economic decisions, the most important question may therefore become less about what machines are capable of deciding and more about which decisions society is willing to let them make without meaningful human accountability.
