Catholic Health Care and AI Ethics: Algorithms for Human Flourishing

Abstract

Artificial Intelligence (AI) contributes to common goods and common harms in our everyday lives. In light of the Collingridge dilemma, information about both the actual and potential harm of AI is explored and myths about AI are dispelled. Catholic health care is then presented as being in a unique position to exert its influence to model the use of AI systems that minimizes the risk of harm and promotes human flourishing and the common good.

Introduction

Artificial Intelligence (AI) technology is rapidly evolving. Throughout the world, in many different settings and industries, AI is contributing both common goods and common harms. AI powered technological advances can enhance life for many. At the same time, however, the unscrupulous application of AI systems can exacerbate marginalization and contribute to the degradation of human life.

After introducing some basic terms and application of AI in the health care setting, I explore the Collingridge dilemma and the concept of data colonialism. I will then seek to dispel some common ideological myths about AI and data ethics. Inspired by the application of Catholic social teaching to the experience of globalization, I conclude by establishing that Catholic health care is uniquely positioned to implement and promote AI systems that minimize harm and contribute to human flourishing throughout the world.

Definitions

• • Artificial intelligence (AI) refers to the capability of computers to demonstrate intelligent behavior (Matheny et al. 2019, 14).

• • A key form of AI is machine learning (ML), a type of modeling technique that can learn and predict a target state (Matheny et al. 2019, 15).

• • An algorithm “is a set of instructions of how a computer should accomplish a particular task” (Caplan et al. 2018, 4).

• • Big data refers to datasets that are too large to be processed by traditional databases (Davis and Patterson 2012, 6).

• • Deep learning is a type of ML that processes extremely large data “to find relationships and patterns that humans are often unable to detect” (Taulli 2019, Ch. 4).

AI is deployed throughout the health care industry in many ways that promote health and contribute to the common good. Basic AI applications include those that help predict the future, identify and/or diagnose in the present, and develop an understanding of the past. Predictive AI tools can be used in health care to review large datasets that can project readmission rates, mortality, and sepsis, to support efficient allocation of resources (Topol 2019, 48). Another common application of AI systems in health care is diagnostic imaging. An AI system can be trained to identify lung nodules, liver masses, head trauma, or pancreatic cancer, for instance (Topol 2019, 44). This AI application can enhance and speed up the detection and diagnosis of diseases in real-time. Finally, an AI system designed for public health can combine geospatial and person generated data from the past, like GPS location data, to explore contributing factors to community health outcomes on the population level (Mooney and Pejaver 2018, 96). These are examples of how AI can promote the common good by analyzing data from the past, assisting with diagnoses in the present, and predicting the future. However, it is also critical to be aware of the common harms it can foster.

Common Harms of AI in Health Care

Collingridge Dilemma

In the field of responsible innovation, many concerns of academics and policy makers about the development of science and technology can be summed up by the Collingridge dilemma. Named after David Collingridge, the researcher who coined it in 1980, it states, “‘attempting to control a technology is difficult…because during its early stages, when it can be controlled, not enough can be known about its harmful social consequences to warrant controlling its development; but by the time these consequences are apparent, control has become costly and slow’” (Genus and Stirling 2018, 3). AI is an excellent example of this dilemma in action. Spurred on by the anticipation and promise of the technology, AI systems were built and implemented before their implications could be fully comprehended. When the developer has the power to control the technology, there is not enough information available to know how it might need to be controlled. The corollary to this double-bind problem is that, after enough information about a technology’s implications, the power to change it is greatly diminished.

Olya Kudina and Peter-Paul Verbeek deepen this dilemma by adding the additional layer of “value dynamism.” In their exploration of privacy around the development and testing of Google Glass, they note “…when we develop technologies on the basis of specific value frameworks, we do not know their social implications yet, but once we know these implications, the technologies might have already changed the value frameworks to evaluate these implications” (Kudina and Verbeek 2019, 293). During the testing phase of these eyeglasses, which featured a camera that could be connected to the internet to facilitate live video streaming of what the wearer was seeing, societal privacy values underwent “technomoral change” (Kudina and Verbeek 2019, 294). People wearing Google Glass encountered increased levels of privacy concerns, as it was unclear to those they encountered how the technology was being used. Some venues prohibited them by declaring “Glass-free zones” (Kudina and Verbeek 2019, 298). Kudina and Verbeek propose a “sociotechnical experimentation approach” replacing the “anticipation [of new technologies] with responsible experimentation” as a way of navigating the Collingridge dilemma in light of technology’s constant impact in the world (Kudina and Verbeek 2019, 309). Realizing that the impact of new technology can never be fully known, its introduction should be approached as carefully as a science or medical experiment, within a principlist bioethical framework of nonmaleficence, beneficence, respect for autonomy, and justice to minimize risk of harm and side effects (Kudina and Verbeek 2019, 296). This call for responsible experimentation transcends all AI systems development, particularly those related to the health care sector.

Data Colonialism

In addition to the tension between information and power articulated by the Collingridge dilemma, concerns of data colonialism are coming into focus. The pace of datafication, or the “transformation of [human] social action to online quantified data,” allowing for tracking and real-time analysis is rapidly increasing (Van Dijck 2014, 198). Nick Couldry and Ulises Mejias notice how the practice of datafication appropriates human life and name it “data colonialism.” This provocative term refers to “the extension of a global process of extraction that started under colonialism and continued through industrial capitalism, culminating in today’s new form: instead of natural resources and labor, what is now being appropriated is human life through its conversion into data” (Couldry and Mejias 2019b, ix). They continue by identifying how this results in the degradation of life, as people are subject to continuous surveillance, data extraction, and ultimately become fuel for for-profit production (Couldry and Mejias 2019b, ix).

Shoshana Zuboff uses the term “surveillance capitalism” to describe the often unexpected and invisible ways in which data is extracted, commodified, and controlled, creating new markets of behavioral prediction and behavioral modification (Zuboff 2015). In order for this capitalism to thrive, there must be “cheap nature,” that is, easily appropriated natural resources that exist in abundance (Couldry and Mejias 2019a, 339). This would then justify describing data as “exhaust,” as the data are generated simply existing and are “not capable of being owned by anyone” (Couldry and Mejias 2019a, 340; Zuboff 2015, 79). Couldry and Jun Yu describe data as “a raw material with value” (Couldry and Yu 2018, 4476).

In the extreme, there is a movement called “big data exceptionalism,” in which “data collection is so pervasive, it should be exempted from legal regulation” (Couldry and Yu 2018, 4474). Should this position become policy, the outcome would mirror key components of historical colonialism, including the appropriation of resources (extracted data from daily human life), extremely unequal economic relations (global technology platforms requiring user consent to participate), unequal distribution of benefits (technology companies collecting enormous amounts of revenue), and the spread of related ideologies (data collection is so ubiquitous, nothing can be done about it) (Couldry and Mejias 2019b, 4). Almost any major modern technology company (Apple, Facebook/Meta, Google, etc.) has practices that align with these colonial characteristics. The hardware and software products of these companies constantly record the activity of their users. This data collection typically occurs after the user agrees to a lengthy terms of service document. If there is any disagreement with these terms, the user is effectively left without recourse, except to not use the product. The data collection drives revenue, usually through sales of focused advertisements, based on extracted user data. Facebook, for instance, received over $84 billion USD in the first three quarters of 2021 (Facebook 2021). This 9-month revenue total is greater than the 2020 annual gross domestic product of most countries in the world (The World Bank 2021). These colonial elements depend on the promulgation of one or more ideologies that facilitate the resource extraction and perpetuate the economic and social marginalization. Dataism, the belief in the objectivity of quantified data and trust in the companies who extract, process, and utilize personal data, is an example of an ideology that perpetuates the unequal economic and social relationships that are core to data colonialism (Van Dijck 2014, 198).

Ideological Myths

In the next section of this paper, three ideological myths about AI and data ethics are explored: (1) datafication is inevitable, (2) we can objectively trust “big data” and algorithms, and (3) privacy is no longer a concern once it is “de-identified.” These myths illustrate conventional wisdom about the emergence of AI technologies. If left unchecked, these myths will continue to cause harm. Their pervasiveness in the understanding of those using and benefiting from AI systems will further numb cultures to harmful bias that can occur when these systems are misused. Human dignity is harmed when appropriation and marginalization occur. We can begin to avoid this degradation of life once we understand some of the ideological myths that perpetuate data colonialism.

Datafication Is Inevitable

Dataism, defined above, is a belief and trust in the structure that makes datafication possible. It is likely that most people unconsciously understand that their activity will just naturally flow into a data stream primed for extraction. The more personal technology becomes, the more natural datafication becomes (Couldry and Mejias 2019b, 16). In time this “legitimate means to access, understand and monitor people’s behavior [becomes] a leading principle” (Van Dijck 2014, 198).

In the context of health care delivery, datafication is commonplace. Patients are continuously connected to electronic monitors and electronic medical records collect multiple types of digitized patient data. In a health care setting, the access, monitoring, and effort to understand the data collected during a patient’s stay is limited to their individual encounter. The goal of the health care provider is to contribute to the care and healing of the individual while in their care.

In responding to the concept of datafication, Mejias and Couldry note that “by naming a process (datafication), we also invoke its limits” (2019, 7). There are somethings that ultimately cannot be quantified or reduced to data. Additionally, they note the ineffectiveness of a single individual opting out of a technology platform and suggest developing “a larger awareness of ourselves as the objects of datafication [as this] can contribute to creating challenges and alternatives to the growth of datafication” (Mejias and Couldry 2019, 7).

We Can Objectively Trust “big data” and Algorithms

The trust that is commonly placed in “big data” suggests that, from a large pile of “raw data,” conclusions with a mathematical certainty can be gleaned. Van Dijck notes that making sense of patterns within large amounts of data “requires critical interrogation: why do we look for certain patterns in piles of metadata, in whose interests, and for what purposes?” (Van Dijck 2014, 202). The purpose with which the data was collected and the lens through which it is analyzed and interpreted are critical in determining its value.

In health care, algorithms, which increasingly depend on large amounts of data with which the model can be trained and implemented, can perpetuate bias and lead to real harm. In their comprehensive survey on bias in machine learning, Ninareh Mehrabi and colleagues provide meaningful insight into the many ways bias can be experienced through data and the use of algorithms (Mehrabi et al. 2019). Mehrabi lists six forms of algorithmic bias and twenty-three types of bias in data (Mehrabi et al. 2019, 4–11). One of the types, population bias, is described as “when statistics, demographics, representatives, and user characteristics are different in the user population represented in the dataset or platform from the original target population” (Mehrabi et al. 2019, 5).

In 2019 Ziad Obermeyer, et al., describes racial bias in a predictive algorithm intended to identify patients in need of high-risk care management. Patients identified by this algorithm become candidates for a program that seeks to “improve the care of patients with complex health needs by providing additional resources, including greater attention from trained providers, to help ensure that care is well coordinated” (Obermeyer et al. 2019, 447). The assumption is: “those with the greatest care needs will benefit the most from the program” (Obermeyer et al. 2019, 447). Past data, including health inputs, outputs, and outcomes, is used to build a prediction model for future patients. In their study, Obermeyer, et al. identify little bias in analysis of cost. The bias is instead found in their risk prediction function. A risk prediction is calculated by developing a “comorbidity score,” based on the number of active chronic diagnoses (Obermeyer et al. 2019, 448). In the algorithm studied by the authors, Blacks have a higher burden of illness than Whites. In fact, Blacks had 26.3% more chronic illnesses than Whites (Obermeyer et al. 2019, 448). The algorithm itself did not use race. It did, however, utilize total medical expenditures and makes a prediction about future health costs as a proxy for health needs. Generally, this is fair. But health need does not always correlate to health expenditures. (Obermeyer et al. 2019, 451). The nature of the costs also differs.

This study demonstrated that Blacks have fewer inpatient surgical and outpatient specialty costs, but more costs related to emergency room use and dialysis. When accurate predictions are based upon costs, it means that the prediction is racially biased. Reasons for this include: poor patients have greater barriers to access care and are often sicker, reduced trust in the health system by Black patients, direct discrimination, elements of the physician/patient relationship, etc. Black patients are underrepresented in these data. Correcting for this bias would “increase the percentage of Black patients receiving additional help from 17.7 to 46.5%” (Obermeyer et al. 2019, 447)

The level of dependence and certainty physicians and health systems place on these algorithms is critical. If a racially biased algorithm is the sole method for patients to get screened into a higher level of care, then the algorithm can perpetuate racial disparities. The physician and care team should be educated on the limitations of the algorithm and ensure there are mechanisms in place to correct for the potential of contributing to health disparities. The physician’s misplaced trust in the algorithm could diminish the patient’s trust in their provider as well as the entire health care industry. In part for this reason, when the American Medical Association (AMA) uses the term “AI,” it refers to “augmented intelligence” (American Medical Association 2019). In doing so, it reinforces the identity of the AI system as a tool, subordinate to the clinician or provider.

Privacy Is No Longer a Concern Once Data Is “de-identified”

Health care personnel in the United States are very familiar with the Healthcare Information Portability and Accountability Act (HIPAA). This law is intended to allow the sharing of patient data only as necessary for treatment, payment, or health care operations, unless authorized by the patient (CDC 2019). If the data are de-identified, HIPAA no longer applies and it can be shared freely (Ohm 2009, 1737). This lack of regulation for de-identified or anonymized health data provides a false sense of security.

In practice, it is becoming increasingly easy to re-identify de-identified data. After an analysis of 1990 census data, it was determined that 87.1% of people in the United States could be uniquely identified suing their five-digit ZIP code, birth date, and sex (Ohm 2009, 1719). With the proliferation of datasets available the ability to cross-reference datasets to identify people renders the terms “anonymize” and “de-identify” misleading. Instead, the word “scrub” is recommended, as it does not imply the successful removal of identifiable information from data, but only indicates effort (Ohm 2009, 1744). Privacy will continue to be a concern whenever data is used, as “data can be either useful or perfectly anonymous, but never both” (Ohm 2009, 1704).

Dispelling these ideologies is among the first steps to addressing practices of data colonialism. This is the type of information that can catalyze a preventative response to potential harm as AI technology continues to evolve.

Globalism and Catholic Social Teaching

As globalism continued to emerge in the early twenty-first century, Kenneth Himes considered the resources available in Catholic social teaching that could assist with the ethical assessment and navigation of the neoliberal capitalist project (Himes 2008). In doing so, he made a few observations that are applicable to this project. First, he quoted Pope John Paul II’s initial assessment that globalization, “a priori, is neither good nor bad. It will be what people make of it” (Himes 2008, 269). Additionally, Himes clarified a few generalities of globalization: it is not fully formed or developed, it is shaped by human choice and action rather than predetermined, and it is subject to ethical assessment because it is a set of processes guided by human persons (Himes 2008, 269).

These generalizations also hold true to AI and data ethics. AI, a priori, is also amoral. As with most technologies or tools, the creators and operators of AI systems can use them for good or evil. It is also important to note that AI significantly evolved within the milieu of globalization and is intertwined with the free market tendencies of this prevailing form of economic liberalism. Himes’ description of globalism as “new forms of interaction and interdependence,” could be another way to describe datafication around the world (Himes 2008, 273). Considering these similarities, it may be prudent to consider an ethical approach to AI as a subset of globalization. Although not a perfect mirror image, there are enough parallels and connection points to serve as a useful way to begin an exploration of how Catholic social teaching can impact AI to promote human flourishing.

Catholic Health Care is Positioned to Lead

The Catholic health care ministry is particularly well suited to navigate the opportunities and challenges of AI’s proliferation throughout the world. The theory and theology upon which the ministry is founded provides a firm foundation and, after generations of putting this to practice has earned influence which can be used to further promote human flourishing and the common good. In this final section, the theory, praxis, and influence of Catholic health care will be explored to inform its approach to discerning and implementing AI systems.

Theory

Catholic health care benefits from a vast tradition of theological and applied ethics as it discerns how the ministry will continue the healing ministry of Jesus Christ. The founders of Catholic hospitals were inspired by Gospel stories of Jesus restoring the health of those who were sick, infirm, and marginalized. This remains true for the leaders of Catholic health care today.

Of the many examples of Jesus’ healing actions, stories about the healing of lepers (Luke 5:12-16, 17:11–19) illustrate the wholistic experience of healing that God can provide. During the time of Jesus’ ministry, lepers were those who suffered from some type of skin ailment, with visible scales or rashes (Martin 2017, 67). In addition to the physical infirmity related to this form of leprosy, lepers were identified by their torn clothes and were required to announce their presence by calling out “Unclean!” so that members of the community could keep their distance (Martin 2017, 67). Lepers suffered from physical illness and were excluded from their community. As a result of Jesus’ healing, the physical disease was cured, health was returned, and relationships to their family and community were restored. With these acts of healing on both the individual and community levels, Jesus demonstrated a deep respect for human dignity and promoted the common good. Today, the Catholic health ministry continues this healing ministry by providing compassionate care to all those in need and curing disease whenever possible.

The principles of human dignity and the common good are at the root of Catholic social teaching, a body of teaching that began to take its current form in the late nineteenth century. Observing the “dramatic social, political, and cultural impact” that emerged within the industrial revolution, the church became concerned about justice in the face of the conflict between capital and labor (Pontifical Council for Justice and Peace 2004, 88). Rerum Novarum, the encyclical written by Pope Leo XIII in 1891, identified social ills and reflected upon them through the lens of the Gospel. In the document, Leo XIII established “the Catholic doctrine on work, the right to property, the principle of collaboration instead of class struggle as the fundamental means for social change, the rights of the weak, the dignity of the poor and the obligations of the rich, the perfecting of justice through charity, [and] the right to form professional associations” (Pontifical Council for Justice and Peace 2004, 89). Over the subsequent decades, Popes have continued the living tradition of discerning cultural, economic, and political events in light of the Gospel. The articulation of these principles continues to evolve, yet these core concepts remain: respect for human dignity, particularly those who are vulnerable, and the common good.

Catholic health care is one way the church continues to live out these principles. In the Introduction to Part One of The Ethical and Religious Directives for Catholic Health Care Services (ERDs), the first three principles named in describing the ministry’s social responsibility are (1) promote and defend human dignity, (2) care for the poor, and (3) contribute to the common good (United States Conference of Catholic Bishops 2018, 8). The ERDs offer concrete guidance on how to apply the Gospel and Catholic social teaching in the context of health care operations.

Similarly, the church has recognized the emergence of AI technology and reflected upon it in light of the Gospel and principles of Catholic social teaching. In February 2020, the Rome Call for AI Ethics (RCAIE) was publicly announced and signed by representatives of the Pontifical Academy for Life, Microsoft, IBM, the Food and Agriculture Organization (FAO) of the United Nations, and the Italian Ministry of Innovation. This document recognized the emerging significance of artificial intelligence on ethics, education, and rights. It offers six principles as “fundamental elements of good innovation” (Pontifical Academy for Life 2020, 6). If followed, these principles of algorithm ethics (or “algor-ethics”), transparency, inclusion, responsibility, impartiality, reliability, and security/privacy, will help to guide the creation of AI tools that will “serve the entire ‘human family’” (Pontifical Academy for Life 2020, 3). The “human family” entails a respect for the human dignity of all, including natural environments, with a particular concern for those who are most vulnerable. These principles should be at the forefront in all stages of AI system development (Table 1).

Table 1.

Rome Call for AI Ethics (2020).

Principles	Description
Transparency	In principle, AI systems must be explainable
Inclusion	The needs of all human beings must be taken into consideration so that everyone can benefit and all individuals can be offered the best possible conditions to express themselves and develop
Responsibility	Those who design and deploy the use of AI must proceed with responsibility and transparency
Impartiality	Do not create or act according to bias, thus safeguarding fairness and human dignity
Reliability	AI systems must be able to work reliably
Security and Privacy	AI systems must work securely and respect the privacy of users

The respect for human dignity and care for the vulnerable are recognizable as key elements of Catholic social teaching. Pope Francis, sending a greeting to those gathered, read by the President of the Pontifical Academy for Life, Archbishop Vincenzo Paglia, frames these principles of “algor-ethics” in light of the dignity of the person, justice, subsidiary and solidarity (Pope Francis 2020, 3). Noting the benefits and technological advancements anticipated in the potential of AI, the pope is careful to identify areas of concern as well. These AI systems are not neutral. Rather, they can contribute to exploitation, the loss of freedom and the expansion of inequities. For these reasons, “it is not enough simply to trust in the moral sense of researchers and developers of devices and algorithms” (Pope Francis 2020, 2). An ethical framework, with public-private support is necessary to achieve the benefits of AI while avoiding the pitfalls and temptations that come with a new technology of this potential.

Pope Francis is particularly concerned with the socio-economic impact of AI. He warns of users—people—being “reduced to ‘consumers’” (Pope Francis 2020, 1). This reduces both the humanity and the agency of the individual that subjected them to the owner of the algorithm. The consolidation of wealth into the few is certainly dangerous to the individual, but it also poses “grave risks for democratic societies” (Pope Francis 2020, 2).

The RCAIE takes care to understand the impact of AI system on the environment. On one hand, AI can contribute to agricultural innovations and play a role in feeding those who are hungry throughout the world (Pontifical Academy for Life 2020, 4). AI can also shape an individual’s experience of the world by contributing to how living conditions are created (Pontifical Academy for Life 2020, 4). There is a risk with this environmental impact to make decisions that benefit only the wealthiest and leave the vulnerable members of the community without a voice or an avenue of participation. A sustainable approach to the environment promotes the common good. The RCAIE reinforces the appropriate relationship between humanity and technology: “AI-based technology must never be used to exploit people in any way, especially those who are most vulnerable. Instead, it must be used to help people develop their abilities (empowerment/enablement) and to support the planet” (Pontifical Academy for Life 2020, 4).

It is notable that the first signatories of this document include international private technology companies, governmental and intergovernmental organizations, in addition to the Vatican’s Pontifical Academy for Life. The development of AI systems is vast and rapidly evolving. A significant partnership with a diverse slate of influential, wide-reaching organizations will be necessary to ensure that AI technology serves humanity and promotes the common good. The RCAIE demonstrates an important milestone in the church’s discernment and response to AI technology. Catholic health care, too, should study, sign on, and integrate the RCAIE. Additionally, the ministry should expect any collaborator or vendor related to AI systems to sign on as well (Baric-Parker and Anderson 2020, 478).

Praxis

Catholic health care has a long tradition of effectively integrating its mission and core values into its organizational operations. Although variation exists in the particular language used for Catholic health system mission statements and organizational values, the Catholic Health Association has summarized most into a Shared Statement of Identity for the Catholic Health Ministry. This statement includes the commitments: (1) promote and defend human dignity, (2) attend to the whole person, (3) care for poor and vulnerable persons, (4) promote the common good, (5) act on behalf of justice, (6) steward resources, and (7) serve as a ministry of the church (The Catholic Health Association 2021). To ensure the integrity of these identity commitments, a few key integrative practices are commonly utilized in health care ministries, including the appointment of mission leaders, providing formation for ministry staff, and conducting discernments when significant decisions must be made.

When Catholic hospital ministries were originally established, they were led by members of their founding congregations. Members of these congregations, usually women, would have leadership roles in the hospital. As administrative, nursing, or food service leaders, these sisters would integrate their faith-based mission in all that they would do. For generations, this was effective. However, after the Second Vatican Council, most Catholic religious orders saw steady declines in membership. This decline resulted in the need to hire non-religious laypersons to lead these Catholic hospitals (Wall 2016, 4). As a result, a new leadership role emerged to support the integration of the ministry’s mission, values, and identity with decision-making and operational activities (Wall 2016, 177).

A significant responsibility of the mission leader is to provide formation to ministry staff (Gremmels 2019, 32–33). Respecting that not all employees are Catholic or Christian, formation is provided at all levels of the organization to ensure everyone is empowered to make decisions and operate in a way that reflects the mission, values, and identity of the health ministry. This inclusive practice welcomes all from any faith tradition or religious practice, provided they are willing to participate in the ministry’s mission.

When an organizational decision needs to be made that will have a significant impact on resources, the service provided to a community, and/or ministry reputation, a mission discernment is conducted. The goal of a discernment is to facilitate a dialog with key stakeholders in which all focus on “responsible decision making in light of the organization’s mission and values” (Marceau 2003, 40). Similar to a checklist in a surgery suite, an effective mission discernment can reduce decisional errors, minimize unnecessary costs, and promote participation in a shared decision-making process. An additional benefit of a mission discernment is context and language that can assist in crafting both external and internal communications about the decision. When a difficult decision is carefully discerned, the ministry is best empowered to communicate how a decision reflects the mission, values, and identity of the organization. Many organizations allow financial contexts to be the driving force in decision-making. This is different for Catholic health care. The ministry is never exempt from difficult decisions like making a reduction in staffing or closing a particular hospital ministry. Finance is always an important factor, but it is not the only one. The discernment process helps to ensure a wholistic, mission driven approach is applied to decision-making, ensuring the finance elements are in proportion to other factors, such as commitment to the community served, quality improvement, impact on staffing, etc.

These integrative practices empower Catholic health care to ensure the Gospel, Catholic social teaching, and additional guidance such as the RCAIE, are reflected in the hospital ministry’s operations. Bioethicists and mission leaders should be engaged to facilitate discernments and formation programs that engage with health ministry leaders and other stakeholders in decision-making and implementation related to AI technology. Industry wide, AI systems have been developed and deployed. This is true for Catholic health care as well. An integrated mission driven approach to AI in health care does not mean AI will be prohibited or refused. Rather, it makes it possible to implement AI technology in a manner that reflects the long-standing tradition of bringing the theology and applied ethics offered by the Catholic church to health care praxis.

Influence

Catholic health care’s ongoing commitment to integrating the Gospel and principles of Catholic social teaching in their delivery of health care services have garnered respect throughout the communities they serve. The influence this reputation affords comes through the ministry’s commitment to the communities in which they serve, its voice as an advocate for those who are poor and vulnerable, and its significant market share.

This reputation is perhaps most concretely apparent in Catholic health care’s identity as anchor institutions. Simply by remaining in the community they serve, Catholic hospitals deploy their resources to hire, and purchase goods and services, thereby investing in their neighbors. Recognizing this innate opportunity to invest in their community, Catholic health systems have collaborated with secular health systems on this work in the context of the Healthcare Anchor Network (HAN). Not only do members of HAN share best practices on how to leverage hiring and purchasing resources to more effectively contribute to a healthy community economy, but they collaborate in the space of public policy advocacy (Bich Ha Pham and David Zuckerman 2020). The “Healthy Communities Policy Framework” features many elements that demonstrate respect for human dignity and promote the common good, including employment, education, safe neighborhoods, affordable access to nutritious foods, affordable housing, and safe sustainable transportation (Bich Ha Pham and David Zuckerman 2020, 65).

Catholic hospitals have, in both word and deed, cared for the poor, vulnerable, and uninsured in their communities (Wall 2016, 116–17). The health care ministry and United States Catholic church leadership have regularly advocated for public policies that expand access to health care. The 1981 pastoral letter, Health and Health Care, states “Every person has a basic right to adequate health care. This right flows from the sanctity of human life and dignity that belongs to all human persons, who are made in the image of God… Special attention should be given to meeting the basic health needs of the poor. With increasingly limited resources in the economy, it is the basic rights of the poor that are frequently threatened first. The church should work with the government to avoid this danger” (United States Catholic Conference 1981, 17–18). This policy position, rooted in the Gospel and Catholic social teaching, is one way the health care ministry advocates in union with the church to be a voice for the voiceless and promote the common good.

The Catholic health ministry also has significant industry market share. The nationwide ministry has a presence in each of the fifty United States, operates over six hundred hospitals, employs over five hundred thousand full-time employees, and cares for more than one in seven patients, resulting in over five million annual admissions (The Catholic Health Association of the United States 2021). In addition to being illustrative of the church’s mission to serve those in need, these numbers represent significant industry market share. A benefit of this market share is influence with elected officials and policy makers, at the local, state, and national levels. This type of market share also results in significant amounts of patient data that is extremely valuable to developers of health care AI applications.

The Catholic health ministry has an obligation to both participate in the marketplace in a way that models health care operations that integrate principles of Catholic social teaching and influence collaborators and competitors to do the same. If Catholic health care models stewardship of patient health data and implementation of AI systems that reflect the Gospel, Catholic social teaching, and the RCAIE, a significant amount of those served in the health care system would benefit. It is very possible that this approach to AI ethics will influence legislation and regulations developed by policy makers and elected officials. This approach may influence other health care organizations and AI developers participating in the industry as well.

Conclusion

In many ways, AI technology is promoting the common good while also contributing to some common harms in the world. As this new technology continues to evolve, there is opportunity to avoid the Collingridge dilemma. The experience of data colonialism and harmful bias in the implementation of AI systems provides us with enough information to control the emerging technology so that it might reduce harm and ensure it is applied in the service of humankind. Today, Catholic health care finds itself in a unique position at the right time in history. The health ministry benefits from a solid theological foundation, has the experience of integrating this theology with its operations, and can leverage its integrity to advocate for policies that minimize the risk of harm presented by AI technology. Catholic health care must utilize its theory, praxis, and influence to implement AI systems that contribute to human flourishing to continue the healing ministry of Jesus Christ.

Biographical Note

Michael Miller Jr. serves as System Vice President of Mission and Ethics for SSM Health in St. Louis, Missouri. During his 12 years of progressively responsible mission leadership roles in Catholic health care, Michael has established himself as an effective health care executive and trusted colleague. He provides ministry formation, clinical ethics consultation, mission discernment facilitation and community benefit leadership, that empowers health care leaders to manage the tensions—and embrace the opportunities—of operating a health care business as a ministry of the Catholic Church. Michael holds graduate degrees in Theology and Bioethics and is pursuing a Doctorate in Health Care Mission Leadership (D.HCML) from Loyola University Chicago. He is certified in Healthcare Ethics Consultation via the HCEC Certification Commission of the American Society for Bioethics and Humanities.

ORCID iD

Michael Miller Jr https://orcid.org/0000-0001-8813-8191