Spiritual Virtue in the Age of AI and Algorithmic Media

The theological depth behind the five moves: virtues, gifts, fruits, and charisms, and what each demands of the systems that now form our attention.

Executive Summary

This report uses Catholic theology as the most systematically articulated framework for distinguishing virtues, gifts, fruits, and charisms, while comparing Orthodox, Anglican, Reformed, Wesleyan/Methodist, and Pentecostal/Charismatic emphases. These traditions share a biblical center (especially Isaiah 11, Galatians 5, Romans 12, John 15, and 1 Corinthians 12 to 14) but differ in terminology, sacramental framing, and the relative attention given to sanctification, liturgical reception of the Spirit, or extraordinary spiritual gifts.

The fundamental distinction:

In Catholic theology, prudence, justice, fortitude, and temperance are cardinal moral virtues; faith, hope, and charity are theological virtues directed immediately toward God. The seven gifts (wisdom, understanding, counsel, fortitude, knowledge, piety, and fear of the Lord) perfect the virtues and dispose the believer to follow the Spirit. The fruits are the mature effects of life in the Spirit. Charisms, whether conspicuous or ordinary, are given not primarily for the recipient's private perfection but for service and ecclesial upbuilding.

AI systems cannot literally possess theological virtues, receive the gifts of the Holy Spirit, manifest sanctifying fruit, or exercise a charism. Those realities pertain to human persons and communities in relationship with God. AI can, however, be designed and governed so that it supports or obstructs human virtue. Catholic reflection on AI warns against equating computational performance with embodied, relational, moral, and spiritual intelligence, and against assigning machines a status or authority that obscures human responsibility.

The central analytical finding is that AI and media should be understood not merely as tools or information channels but as environments of formation. They shape attention (what people repeatedly notice, ignore, fear, desire, and remember), habituation (what responses become easy, automatic, socially rewarded, or emotionally satisfying), social imagination (how people perceive neighbors, enemies, authority, truth, success, and human worth), institutional incentives (what organizations optimize, measure, promote, and tolerate), and practical judgment (whether people retain or surrender the work of discernment).

Empirical research supports concern about these mechanisms. Out-group hostility is a strong predictor of social-media engagement, and an audit of an engagement-ranking system found that it amplified emotionally charged and divisive material more than timelines based on users' stated preferences. This does not prove that all algorithmic media necessarily corrupt character, but it shows that an objective such as engagement can reward patterns contrary to peace, gentleness, temperance, justice, and charity.

A spiritually informed AI-ethics framework should therefore unite three layers:

Virtue cannot replace law, technical risk management, or institutional accountability. Conversely, compliance cannot replace character. Principles remain aspirational when workers lack honesty, courage, care, prudence, and organizational permission to resist harmful incentives.

Theological Architecture Across Christian Traditions

The categories overlap, but they are not interchangeable. The confusion is common because the English word gift may refer broadly to any divine grace, specifically to Isaiah's sevenfold gifts, or to Pauline charisms such as prophecy, teaching, administration, healing, service, or discernment.

Charisms are structurally different from virtues and fruits. A person can possess an impressive capacity to teach, organize, communicate, heal, lead, or persuade while exercising it immaturely or without love. Paul's treatment of Corinth is decisive: manifestations are diverse, are given for the common benefit, and must be ordered by love, intelligibility, mutual edification, and communal discernment. No charism is exempt from discernment.

Technical brilliance, persuasive communication, visionary entrepreneurship, and model-building ability are capacity-like goods, not proof of moral maturity. The presence of highly capable engineers does not establish that a product is just, prudent, truthful, or ordered to human flourishing.

Major Christian perspectives

Across these traditions the common principle is clear: capacity is morally underdetermined. It becomes beneficial only when incorporated into truthful judgment, self-command, love, service, communion, and accountability. That principle transfers directly to AI: greater technical capability increases the need for stronger moral formation and governance; it does not reduce it.

Virtuous Leadership

Alexandre Havard rejects the idea that leadership is fundamentally a technique, temperament, public position, or magnetic personality. Leadership is a dimension of character available to people in many stations of life: achieving greatness by bringing out greatness in others.

Applied to AI, this corrects two opposite failures. Technocratic smallness says: our job is merely to optimize the metric, ship the model, or comply with the rule. Technological grandiosity says: our system will solve education, loneliness, politics, religion, or human meaning. Magnanimity asks whether the project serves a genuinely great human good. Humility asks whether the organization understands its limits, dependencies, externalities, and responsibility to people who did not choose the system. Humility without magnanimity can become timidity; magnanimity without humility becomes domination.

The Two Works Called Media Mindfulness

Hailer and Pacatte define media mindfulness as critical thinking about media joined to Gospel values. Media messages are constructed rather than transparent windows onto reality; they select, frame, omit, symbolize, and communicate values. The appropriate response is neither uncritical consumption nor blanket denunciation, but inquiry, reflection, dialogue, and discerning action. A concise form of their method:

Indick approaches media as psychological environments that shape modes of attention, behavior, and thought. Media mindfulness is not total disengagement but deeper awareness of how linguistic, numerical, musical, bodily, natural, and digital environments affect human functioning, with emphasis on changing patterns of engagement rather than assuming insight alone will overcome unhealthy habits.

The accounts synthesize into four questions. Hailer and Pacatte ask: what does this media message mean, what values does it carry, and how should faith respond? Indick asks: what kind of attentional and psychological environment am I inhabiting, and what is repeated exposure doing to me? Virtue ethics asks: what kind of person am I becoming through this pattern of attention and action? AI governance asks: who designed this environment, what does it optimize, who benefits, who bears risk, and what remedies exist?

AI and Media as Environments of Formation

AI and algorithmic media challenge virtue not because technology is intrinsically opposed to spiritual life, but because these systems intervene at the points where character is formed: attention, repetition, imagination, relationship, judgment, and social reward. Pope Francis named this condition directly: "all of us, but especially the young, are immersed in a culture of zapping," and "without the wisdom of discernment, we can easily become prey to every passing trend" (Gaudete et Exsultate 167). He also named its ecclesial form: verbal violence "even in Catholic media" (115). Holiness, in his account, now passes through the discernment of screens.

An AI system's apparent linguistic competence can obscure the difference between generating an answer and possessing practical wisdom. Antiqua et Nova argues that human intelligence is embodied, relational, affective, social, moral, and open to truth and the good in ways not captured by computational task performance. Four consequences follow:

• AI can simulate counsel without possessing the gift of counsel. It may present options or retrieve traditions, but it cannot replace conscience, prayer, lived experience, pastoral responsibility, or communal discernment.
• AI can produce kindness-like language without possessing charity or the fruit of kindness. The moral quality of the interaction depends on the intentions, design, truthfulness, consequences, and human relationships surrounding it, not merely its tone.
• AI can classify ethical concepts without possessing prudence. Prudence is not rule retrieval; it requires perception of morally salient particulars, rightly ordered desire, experience, and accountable choice.
• AI may extend a human charism without receiving a charism. A gifted teacher, translator, administrator, artist, or evangelist may use AI to extend service, but the tool itself is not the spiritual recipient.

Challenges to particular virtues and fruits

The empirical evidence on engagement systems illustrates why temperance and peace matter at the design level. Hostile references to political out-groups are especially engagement-producing, and a preregistered audit found that engagement-based ranking amplified more emotionally charged, partisan, and out-group-hostile content than a chronological feed or a ranking based on users' stated preferences. The ethical problem is therefore not reducible to bad users; it also concerns what the system makes easy, visible, and rewarding.

AI can nevertheless support virtue when it functions as a scaffold rather than a substitute: presenting counterarguments before publication, helping users identify unsupported assertions, translating materials for underserved communities, making documents accessible, prompting reflection on affected stakeholders, retrieving primary theological sources, or structuring a post-incident examination.

An Integrated Ethical Framework

Virtue ethics asks a question often neglected by rule-centered frameworks: what must human beings and organizations become in order to use power well, especially when rules are incomplete and incentives are adverse? Shannon Vallor argues that rapidly changing technosocial conditions require cultivated qualities such as honesty, self-control, humility, justice, courage, empathy, care, civility, perspective, flexibility, magnanimity, and integrating practical wisdom. Thilo Hagendorff makes this operational for AI practice: justice, honesty, responsibility, and care as basic dispositions, with prudence and fortitude as second-order virtues needed to perceive the right course and act despite pressure or institutional resistance.

This mapping is analogical, not an equation of theology with compliance. The Rome Call, the UNESCO Recommendation, the OECD Principles, the NIST AI Risk Management Framework, and the EU AI Act operate in legal, policy, or professional registers. They do not confer spiritual virtue. They nevertheless encode principles substantially consonant with dignity, justice, truthfulness, responsibility, solidarity, prudence, and care.

Why virtue and compliance need each other

A principle-only approach can fail because principles are abstract and may conflict in concrete cases; organizations can formally comply while manipulating scope, metrics, or documentation; workers may recognize a problem yet lack courage or authority to act; incentives may reward silence, speed, engagement, and market capture; and unforeseen situations require judgment rather than mechanical rule application.

A virtue-only approach can also fail because good intentions do not eliminate bias, error, power asymmetry, or structural harm; character judgments can become subjective or protect charismatic leaders from scrutiny; affected people need enforceable rights, not confidence that developers are benevolent; complex systems require measurement, documentation, audit, technical assurance, and remedy; and organizational incentives can overwhelm individual virtue.

The stronger architecture: virtuous purpose + formed practitioners + accountable institutions + technical controls + enforceable rights + continual evaluation of outcomes.

Cases and Historical Lessons

Corinth as an early governance case

The Corinthian controversy in 1 Corinthians 12 to 14 can be read as an early case of governing powerful, unevenly distributed capacities within a community. Paul does not abolish gifts because they can be misused. Instead, he establishes an ordered framework:

The lesson for AI is that plural expertise requires both freedom and order. Innovation should not be extinguished, but neither may capability exempt itself from discernment. The more excellent way is not anti-technical; it is the governance of technical gifts by love and the common good.

Engagement ranking and the vice of the objective function

A 2025 audit of an engagement-ranking system provides a contemporary case in which the optimization target itself became morally formative. Engagement ranking selected more divisive, emotionally charged, and out-group-hostile posts than users said they preferred. The platform did not need an intention to cultivate hostility; the combination of objective function, behavioral signals, and human vulnerabilities was sufficient to amplify it. A system can generate no explicitly prohibited item and nevertheless produce a harmful ecology through repeated selection and reward. Virtue-oriented evaluation therefore asks: What behavior does the metric reward? What emotional states increase system success? What dispositions become easier through repetition? Does the system privilege immediate impulse over reflective preference? Who becomes visible as a neighbor, rival, object, threat, or commodity? What would count as a good fruit after six months of use?

Catholic media teaching as a historical continuity

The Second Vatican Council's Inter Mirifica recognized that mass media influence not only individuals but entire societies. Communio et Progressio emphasized media's capacity to advance human development, solidarity, and proclamation, and later Church documents addressed the internet as a new field of communication, education, participation, and moral responsibility. This history cautions against two simplifications: technological rejection (the Church has repeatedly affirmed media's potential for education, evangelization, solidarity, and progress) and technological innocence (because media shape culture and consciousness, users, creators, institutions, and authorities have corresponding duties of formation, truthfulness, justice, and responsibility). Pope Francis extended this tradition to AI, arguing that AI should serve the highest human potential rather than compete with it, and calling for a wisdom of the heart capable of integrating technology into fully human communication.

The Rome Call in organizational practice

Providence, a United States healthcare organization, became a Rome Call signatory and implemented it through ethical guidelines, discernment practices, ethics consultations, research, and a dedicated AI ethics subcommittee. A virtue statement becomes operational only when attached to named decision-makers, recurring review, access to domain and ethics expertise, escalation and consultation mechanisms, documentation, research and evaluation, and responsibility for revising practices. The limitation is equally instructive: voluntary frameworks depend on internal commitment and external scrutiny, and should complement rather than substitute for legal duties, professional standards, procurement conditions, audits, and rights of appeal.

Practical Integration Roadmap

The same path applies to an AI project and, at a smaller scale, to personal media habits:

1. Name the human good and true purpose.
2. Identify people, communities, and vulnerable groups affected.
3. Select the virtues, gifts, and fruits most relevant to the context.
4. Map risks to dignity, truth, justice, attention, freedom, and communion.
5. Translate virtues into product requirements, habits, and accountable roles.
6. Test with affected users, domain experts, and adversarial reviewers.
7. Ask: are the legal, technical, and virtue thresholds met? If not, return to step 4.
8. Deploy gradually with meaningful human oversight.
9. Monitor harms, benefits, and formative outcomes.
10. Conduct an examen, repair incidents, and revise incentives. Repeat.

Personal practice: a rule of life for AI and media

A useful weekly examen: What repeatedly captured my attention? What did I come to love, fear, envy, or despise? When did AI increase understanding or service? When did it make me passive, hurried, vain, harsh, or less present? What fruits appeared in my relationships? What truth needs correction, what harm needs repair, and what habit should change?

AI may assist by organizing the examination or retrieving records, but it should not be treated as confessor, spiritual director, authoritative interpreter of providence, or substitute for a trustworthy human and ecclesial relationship.

Organizational leadership

Protected dissent is particularly important. An organization does not possess fortitude merely because its values page praises courage. Fortitude exists organizationally when employees can delay or stop a harmful launch without predictable career punishment, and when leaders publicly reward truth-telling before a crisis rather than praising whistleblowers after one. Humility likewise requires structures: external evaluation, red-team access, consultation with people outside the product team, limitation statements, uncertainty reporting, and willingness to abandon a technically impressive system when the use case is not socially justified.

AI product and design practice

1. Purpose and limits. State the human capability the system should strengthen and the decisions it must not make. Specify contexts in which the system is inappropriate, including uses requiring pastoral, clinical, judicial, or fiduciary judgment.
2. Truth and provenance. Distinguish generated content from verified information. Support citations, source inspection, change histories, and correction. Do not use a confident conversational style to conceal uncertainty.
3. Meaningful human oversight. Define what the human reviewer knows, can inspect, can override, and is accountable for. A nominal human in the loop who lacks time, authority, competence, or alternatives does not constitute meaningful oversight.
4. Friction at morally significant moments. Add pauses, confirmations, counterevidence, or secondary review before publishing high-reach content, impersonating a person, making consequential decisions, or transferring sensitive data.
5. Non-manipulative interaction. Avoid deceptive anthropomorphism, false claims of feelings or consciousness, coercive emotional attachment, shame-based retention, and designs that imply spiritual or moral authority.
6. Privacy and restraint. Collect the minimum data necessary, offer understandable controls, limit retention, protect sensitive inferences, and make the privacy-preserving path the default.
7. Care for vulnerable users. Age-appropriate design, crisis escalation, accessibility, clear boundaries, and routes to qualified human assistance. Include vulnerable groups in testing and governance.
8. Reflective preference. Measure not only what people click but what they endorse after reflection, whether the interaction completed their stated purpose, whether they regret the use, and whether the system strengthened or weakened agency.
9. Continuing evaluation. Monitor changes in use, population, model behavior, downstream integration, and organizational incentives across the lifecycle.

Governance and Policy

A virtue-informed public policy does not attempt to legislate holiness. It creates the conditions in which truthful, just, prudent, and responsible conduct is possible, and in which vice, negligence, manipulation, and abuse encounter meaningful constraints.

A recurrent policy error is to regulate only model outputs while leaving the underlying business objective untouched. A platform can remove particular harmful posts while retaining an optimization system that systematically rewards outrage, compulsive attention, or polarization. Governance must therefore examine recommender objectives, advertising incentives, data practices, interface patterns, and the distribution of power. Pope Francis specifically argued that AI-regulation debates must include those commonly unheard in global decision-making, including poor and powerless persons. This corresponds to justice, humility, solidarity, and the Pauline insistence that apparently weaker members of the body cannot be treated as dispensable.

Priority Reading

AI should not be evaluated only by what it can produce, but by what its creation and use ask human beings to love, notice, practice, tolerate, become, and build together.

A project ordered toward virtue therefore evaluates at least four kinds of fruit: the quality of the artifact, the character and culture of its builders, the agency and relationships of its users, and the justice and flourishing of the wider community. Technical excellence is one good among these, not their substitute and not their judge.