AI and Philosophy: A Necessary Alliance

The Importance of Philosophy in AI Design

Companies often embark on auditing their artificial intelligence (AI) systems after they have already set their objectives. However, philosophy should come into play much earlier, at the design phase, to prevent notions of work or performance from becoming rigid and unquestionable systems.

In 2026, a document issued by the Vatican highlighted a truth often overlooked in AI strategies. Pope Leo XIV emphasized that every design choice of an AI system reflects a certain vision of humanity. This observation directly addresses organizations that use AI for tasks such as sorting applications, assessing risks, or recommending actions. Behind these applications lie preconceived ideas about work, merit, and trust. Once these ideas are translated into data and rules, they become infrastructures that are difficult to challenge. Thus, AI governance often begins too late, after the system's objectives have been accepted as self-evident.

Defining Problems through AI

When an organization entrusts a problem to AI, it also conveys a way to define that problem. John Dewey, in his work Logic: The Theory of Inquiry, explains that a problem forms during an inquiry that delineates an indeterminate situation and gives it direction. The formulation of a problem already guides action, and this idea is crucial when the initial framing influences thousands of decisions.

Take the example of a company seeking to improve its customer service. It might define the problem as excessive interaction duration and ask AI to reduce the average processing time. However, a shorter exchange could indicate a quick resolution, but also the abandonment of a customer or the avoidance of a complex case. The same problem could have been framed in terms of trust, clarity of offers, or sustainable resolution. Each framing leads to a different intervention. AI can speed up the processing of a request, but the definition of that request remains a human responsibility.

The same dilemma arises in recruitment. Identifying the "best candidates" requires defining what talent is. Should immediate fit for the position be prioritized, or learning potential, uniqueness of background, or the ability to transform a function? Each answer requires different data and produces distinct rankings. The system then operationalizes this definition, often with a consistency and speed that gives it the appearance of being self-evident.

Research by Rachel Etta Rudolph, Elay Shech, and Michael Tamir shows that the design of machine learning systems engages a form of conceptual engineering. Modifying a model can amount to changing how an institution uses social categories. AI thus industrializes not isolated decisions but the concepts that make them possible. A rough definition, once limited to a few local judgments, can become a rule applied on a large scale. A metric chosen for its availability can gradually transform into the official definition of success.

This is where philosophy finds its role. Technique allows for understanding the system, its capabilities, and its limits. Management organizes its integration into work. Ethics examines rights, responsibilities, and potential harms. Philosophy intervenes much earlier. It clarifies concepts, compares purposes, and makes explicit the conception of the person that the project is about to institutionalize. It prevents a managerial preference, a professional convention, or a convenient indicator from quickly becoming a calculated truth.

Implementing Philosophical Due Diligence

To fulfill this function, philosophy must move beyond commentary and enter the decision-making process. Every significant AI project should undergo philosophical due diligence before the use case is validated. This procedure would aim to verify that the problem, concepts, and purposes entrusted to the system are robust enough to be automated. It should precede the feasibility study and be included in the file submitted to the body responsible for authorizing the project.

• The first examination would focus on the problem formulation. What situation is the company actually seeking to transform? Why is AI a relevant response? What other formulations have been considered? A decrease in loyalty can be interpreted as a lack of personalization, a degradation of service, a loss of trust, or a value proposition that has become less relevant. As long as these hypotheses remain conflated, automation risks accelerating a response without clarifying the question.

• The second examination would concern the central concept. What term will be converted into a score, ranking, or recommendation? How does the organization define quality, risk, satisfaction, or merit? What competing definitions could reasonably be defended? A metric represents a purpose without ever exhausting it. Philosophical due diligence thus forces the company to distinguish what is easily measurable from what truly holds value.

• The third examination would focus on the conflicting goods. Accelerating a process may reduce attention to exceptions. Standardizing a decision can enhance consistency while diminishing the value of situated judgment. Personalizing an offer can increase its relevance while trapping the user in the continuity of past behaviors. Philosophy does not resolve these tensions on behalf of decision-makers. It makes them visible early enough for trade-offs to be accepted before becoming technical routines.

• The final examination would concern contestability. Who will be able to challenge the initial definition of the problem? Will the affected individuals only be able to signal an error, or also discuss the categories that organize the decision? Comprehensive governance must be able to revise the mission entrusted to AI, not just correct its results. Responsibility lies as much in the quality of execution as in the legitimacy of the pursued objective.

Philosophy as a Strategic Lever

This work of clarification does not slow down strategy. It prevents it from too quickly transforming a fragile intuition into technological investment. A request for automation may mask a lack of priority, a conflict between multiple objectives, or a degradation of the organization. By subjecting the initial framing to discussion, philosophy sometimes reveals that the expected response is less about a model than about a transformation of work, service, or the relationship with the customer.

This approach also improves the quality of indicators. Organizations tend to confuse what is easily quantifiable with what truly matters. Once an objective is converted into a metric, the risk arises that this metric becomes the objective itself. Philosophical due diligence keeps the distinction between the sought value and its numerical approximation open.

Finally, it protects the diversity of interpretations. Lisa Messeri and Molly Crockett have shown that AI tools can produce illusions of understanding and foster intellectual monocultures. An organization may generate more analyses while exploring fewer ways to understand its activity. The fluidity of responses then creates an impression of closure. Philosophy reintroduces alternatives. It asks what hypothesis has become invisible, what concept artificially groups different realities, and what purpose has been dismissed before being genuinely examined.

A study conducted with 319 professionals also shows that the use of generative AI shifts critical effort towards verifying and integrating responses. This shift can be useful, but it leaves an underlying question intact. Does the proposed framework deserve to be accepted? An answer can be factually accurate and serve a poor problem. Philosophy allows for controlling what precedes fact-checking, namely the construction of the question, the definition of concepts, and the justification of the purpose.

Integrating Philosophy as an Essential Skill

Philosophy should not be added as a peripheral general culture. It must become a practical skill. Knowing how to problematize a request, distinguish multiple definitions of the same concept, clarify conflicting goods, construct the strongest objection to a decision, and justify a purpose. These exercises can be integrated into AI projects, leadership training, and investment processes.

As models become accessible to everyone, the advantage shifts towards the quality of the missions entrusted to them. An organization can have powerful tools and remain trapped in poor concepts. Another may use the same tools with greater discernment because it has better formulated the problem, better defined the sought value, and better organized the contestation.

Training in AI thus requires four complementary skills.

• Technique allows for understanding systems and their limits.

• Management enables organizing their integration into work.

• Ethics governs rights, responsibilities, and consequences.

• Philosophy allows for choosing ends, clarifying concepts, and preserving the possibility of discussing the mission before automating its execution.