Artificial Intelligence and Sikh Ethics: An Advanced Seminar

Introduction: Why Sikh Ethics for the Age of AI

Artificial intelligence is not a neutral technology. Every AI system embeds values — about what matters, who counts, what tradeoffs are acceptable, and what kind of future is worth building. The question for any religious and philosophical tradition is not whether to engage with AI but how — and whether the resources of that tradition can contribute something distinctive and valuable to the urgent ethical debates about how AI should be developed, governed, and constrained.

This seminar argues that Sikh ethics does have something distinctive to offer. The SGGS's ethical vision — grounded in the convictions that every person bears the divine light (ਜੋਤਿ), that justice (ਨਿਆਉ) is a divine attribute that must be pursued in the world, that truth (ਸੱਚ) is the ultimate ethical standard, and that the community (ਸੰਗਤ) is the proper vehicle for collective discernment — generates a framework for AI ethics that is both internally coherent and engages productively with the major concerns of the contemporary AI ethics literature.

The State of AI Ethics

The field of AI ethics has grown rapidly since approximately 2016, driven by high-profile failures of algorithmic systems — discriminatory hiring algorithms, biased recidivism prediction tools, facial recognition systems with dramatically lower accuracy for dark-skinned faces — and by escalating concern about the longer-term risks of increasingly powerful AI systems. Frameworks have proliferated: principles-based approaches (transparency, fairness, accountability, privacy), rights-based approaches grounding AI ethics in human rights frameworks, virtue ethics approaches (notably Vallor 2016), and consequentialist approaches focused on aggregate welfare. Each framework captures important considerations but also has characteristic blind spots. Sikh ethics engages with all of these frameworks while bringing to the conversation a distinctive theological anthropology — the affirmation of the divine image in every person — that provides a more robust foundation for human dignity than secular frameworks typically offer.

Method of This Seminar

The method throughout is bidirectional: we bring Sikh ethical resources to illuminate AI challenges, and we allow the specificity and urgency of AI challenges to sharpen and develop our understanding of what Gurmat ethics requires in novel contexts. The goal is neither to baptize every AI development with religious approval nor to condemn technology wholesale, but to develop the nuanced, case-by-case ethical reasoning that genuine wisdom requires — and that the tradition models through its own engagement with the novel challenges of its historical context.

Algorithmic Bias and the Gurmat Standard of ਨਿਆਉ

Algorithmic bias — the systematic disadvantaging of protected groups by automated decision-making systems — is among the most thoroughly documented harms in the AI ethics literature. This lesson examines the mechanisms of algorithmic bias and applies the Gurmat standard of ਨਿਆਉ — divine justice — to evaluate both the harms and the proposed remedies.

Mechanisms of Algorithmic Bias

Algorithmic bias arises through multiple pathways. Training data bias occurs when the historical data used to train a system reflects past discrimination — a hiring algorithm trained on historical hiring decisions will learn to replicate those decisions, including their discriminatory patterns. Proxy discrimination occurs when facially neutral variables serve as proxies for protected characteristics — zip code may function as a racial proxy in lending decisions. Feedback loops amplify initial disparities: a predictive policing algorithm deployed in heavily policed communities will generate more arrest data from those communities, reinforcing the prediction that they require heavier policing. Safiya Umoja Noble's research documents how search algorithms encode racial and gender bias in ways that reinforce harmful stereotypes, producing real-world harm for affected individuals and communities (Noble 2018, 1).

ਨਿਆਉ as a Framework for Algorithmic Fairness

The SGGS presents ਨਿਆਉ — divine justice — as a core attribute of the Divine and an ethical imperative for human life. Divine justice in Gurmat is not merely procedural fairness but substantive: it attends to the actual distribution of outcomes across persons and groups. The image of the divine court (ਦਰਗਾਹ) where all are equal before the Creator is a powerful resource for arguing that algorithmic systems which systematically disadvantage already-marginalized groups violate a standard of justice that transcends any particular cultural or legal framework. Crucially, ਨਿਆਉ cannot be satisfied merely by procedural neutrality — by applying the same algorithm to everyone — if the algorithm was trained on unjust data or produces unjust outcomes. This moves Sikh ethics beyond formal equality toward the substantive equality that algorithmic fairness research increasingly demands. Virginia Eubanks's documentation of how automated systems systematically disadvantage poor communities — denying benefits, flagging families for child protective services, predicting recidivism — illustrates precisely the kind of structural injustice that ਨਿਆਉ would condemn (Eubanks 2018, 9).

Surveillance Capitalism, Dignity, and Sikh Privacy Ethics

Shoshana Zuboff's concept of "surveillance capitalism" — the economic logic by which corporations extract behavioral data from individuals to predict and modify their behavior for profit — describes one of the most pervasive and consequential AI applications of the contemporary period. This lesson develops a Sikh ethics of privacy in response to surveillance capitalism.

Surveillance Capitalism and Its Harms

Surveillance capitalism operates by treating human experience as raw material for extraction: every search, click, purchase, movement, and interaction is harvested as behavioral data, processed by machine learning systems, and packaged as "behavioral futures" sold to advertisers and other clients who wish to predict and influence human behavior. Zuboff argues that this logic constitutes an unprecedented assault on human autonomy and dignity: it treats persons not as ends in themselves but as means — as sources of data whose behavior is to be predicted and modified in the service of others' commercial interests. The scale of contemporary surveillance is staggering: the average smartphone user's location, communications, social relationships, purchases, health, and psychological state are tracked by dozens of applications at any given time, and this data is combined, sold, and analyzed in ways that individual users cannot observe or consent to.

A Sikh Account of Privacy

Sikh ethics does not have a classical doctrine of privacy — the concept is not directly addressed in the SGGS. But the resources for developing a Sikh account of privacy are substantial. The affirmation of each person's dignity as a bearer of the divine light (ਜੋਤਿ) grounds strong prohibitions on the instrumentalization of persons — treating them as mere data sources in service of others' interests. The concept of ਹਉਮੈ — ego-driven self-assertion — can be applied to corporate actors: the surveillance capitalism business model is a structural expression of ਹਉਮੈ at institutional scale, claiming the right to extract and monetize the behavioral data of persons without their genuine understanding or meaningful consent. Kate Crawford's analysis of AI's planetary infrastructure — the data centers, labor, and extracted resources that power AI systems — extends this critique: surveillance capitalism's harms are not only epistemic and psychological but material and environmental (Crawford 2021, 7).

Disinformation, Deepfakes, and the Imperative of ਸੱਚ

The capacity of AI systems to generate realistic synthetic media — text, images, audio, video — at scale creates unprecedented opportunities for disinformation that threaten both individual dignity and the epistemic foundations of democratic society. The Gurmat imperative of ਸੱਚ — truth — provides a powerful framework for evaluating these harms.

The Disinformation Landscape

Large language models can generate plausible-sounding text on any topic at industrial scale, enabling the automated production of disinformation at a volume and velocity that human fact-checking cannot match. Deepfake technology can produce realistic synthetic video of public figures saying things they never said, with implications for political manipulation, personal defamation, and the erosion of trust in authentic evidence. The "liar's dividend" — the growing tendency to dismiss authentic footage as potentially fabricated — may be as damaging as the fabrications themselves, undermining the shared epistemic foundation that democratic deliberation requires. For religious communities, AI-generated content creates specific concerns: the potential for synthetic "religious teachings," fabricated quotes from scriptures, or manipulated images of sacred spaces to mislead community members who lack the tools to evaluate their authenticity.

ਸੱਚ and the Ethics of AI-Generated Content

ਸੱਚ in the SGGS is not merely an epistemological concept — the truth of propositions — but an ontological and ethical one: it names the ultimate nature of reality as the divine, and the imperative that human life be oriented toward that reality rather than toward falsehood. The deliberate production and dissemination of AI-generated disinformation is therefore not merely a violation of epistemic norms but an ontological offense — an active orientation of human intelligence and technology toward ਝੂਠ (falsehood) that is antithetical to the deepest Gurmat values. The SGGS's consistent concern with the harms of false speech — the damage it does to community trust, to the dignity of those defamed, and to the soul of the one who speaks falsely — translates directly into stringent ethical requirements for those who develop and deploy AI systems capable of generating disinformation at scale. Floridi's framework of information ethics provides secular corroboration: he argues that epistemic pollution — the contamination of the information environment with false content — constitutes a genuine harm to the information ecosystem on which human flourishing depends (Floridi 2023, 47).

Automation, ਕਿਰਤ ਕਰਨੀ, and the Future of Work

AI-driven automation is transforming labor markets at a pace and scale that poses fundamental questions about the future of work, economic security, and human dignity. The Gurmat concept of ਕਿਰਤ ਕਰਨੀ — honest, productive labor as a foundational ethical practice — provides a distinctive framework for evaluating these transformations.

The Automation Challenge

Contemporary AI systems can perform an expanding range of cognitive tasks — document processing, diagnosis, legal research, creative work, customer service — that previously required human expertise. Economists disagree about the net employment effects: optimists point to historical patterns in which technological displacement ultimately creates more jobs than it destroys; pessimists argue that the pace and breadth of AI-driven automation is qualitatively different from previous waves. The distributional question is less contested: automation tends to disadvantage workers in routine cognitive and manual roles while benefiting workers with complementary skills and those who own the capital embodied in automated systems. This dynamic tends to increase inequality and to concentrate the gains of productivity improvements among already-advantaged populations.

ਕਿਰਤ ਕਰਨੀ and Human Flourishing

ਕਿਰਤ ਕਰਨੀ — honest labor — is one of the three foundational practices of Sikh ethics, alongside ਨਾਮ ਜਪਣਾ and ਵੰਡ ਛਕਣਾ. The tradition understands honest work not merely as an economic necessity but as a form of spiritual practice — the arena in which the virtues of discipline, service, and integrity are cultivated and expressed. This understanding generates several ethical principles relevant to automation. First, economic arrangements that deprive people of the opportunity for meaningful, dignified work — not merely employment but work that develops and exercises human capacities — are ethically problematic regardless of whether alternative income is provided. Second, the gains from AI-driven productivity improvements are not automatically distributed justly; the tradition's parallel imperative of ਵੰਡ ਛਕਣਾ — sharing one's earnings — provides a direct mandate for redistributive policies that ensure automation's benefits are broadly shared. Third, the development and deployment of AI systems that destroy livelihoods without social provision for affected workers violates the Gurmat ethical standard of care for the most vulnerable.

AI in Healthcare: ਸੇਵਾ, Consent, and Algorithmic Medicine

Artificial intelligence applications in healthcare — diagnostic algorithms, predictive risk scores, drug discovery, robotic surgery — represent some of the most potentially beneficial and ethically complex AI developments. The Gurmat ethic of ਸੇਵਾ provides a rich framework for evaluating both the promise and the risks.

The Promise of AI in Healthcare

AI diagnostic systems have demonstrated accuracy matching or exceeding specialist physicians in specific narrow domains: diabetic retinopathy detection, skin cancer classification, radiology reading. Predictive algorithms can identify patients at high risk of deterioration or readmission, enabling preventive intervention. Drug discovery AI can screen millions of molecular candidates at speeds impossible for human researchers. If these capabilities can be reliably deployed at scale, the potential benefits in terms of lives saved, diagnoses not missed, and healthcare costs reduced are enormous — and the implications for global health equity are significant, since AI diagnostic tools could potentially extend specialist-level diagnostic capability to settings where human specialists are unavailable.

ਸੇਵਾ as the Standard for Healthcare AI

The Gurmat ethic of ਸੇਵਾ as selfless service to the suffering provides a demanding standard for evaluating healthcare AI. It asks not merely whether a system improves aggregate outcomes but whether it serves the most vulnerable, whether it preserves the dignity of those it serves, and whether it maintains the quality of human relationship and care that genuine healing requires. Several concerns emerge from this standard. First, AI systems trained primarily on data from wealthy, predominantly white populations may perform poorly on underrepresented populations — reproducing and potentially amplifying health disparities rather than reducing them. Second, the increasing algorithmization of clinical decision-making risks reducing patients to data points rather than encountering them as persons bearing the divine light. The ਸੇਵਾ framework insists that the human relationship of care is not merely instrumentally valuable for outcomes but intrinsically valuable — that being genuinely seen and cared for by another person is itself a dimension of healing that no algorithm can replace.

Autonomous Weapons and the Doctrine of ਧਰਮ ਯੁੱਧ

Lethal autonomous weapons systems (LAWS) — AI systems capable of selecting and engaging targets without meaningful human control — represent one of the most urgent and consequential AI ethics challenges. The Sikh doctrine of ਧਰਮ ਯੁੱਧ — righteous battle — provides a distinctive and demanding framework for evaluating their ethical permissibility.

The Challenge of Lethal Autonomous Weapons

Contemporary militaries are investing heavily in autonomous systems ranging from loitering munitions that select and engage targets within a defined area to fully autonomous kill-chain systems that can identify and engage targets without human decision at the moment of lethal force. The ethical concerns are multiple: autonomous systems cannot make the contextual moral judgments — distinguishing combatants from civilians in complex urban environments, evaluating proportionality, assessing military necessity — that international humanitarian law requires. They may dramatically lower the threshold for initiating armed conflict by removing the human cost of casualties from the deciding calculation. And they create accountability gaps: when an autonomous system kills unlawfully, it is unclear who bears legal and moral responsibility.

ਧਰਮ ਯੁੱਧ and the Ethics of Autonomous Violence

The Sikh tradition's doctrine of ਧਰਮ ਯੁੱਧ — righteous war — establishes demanding conditions for the ethical use of lethal force: it must be a last resort after all peaceful means are exhausted, conducted with minimum necessary force, aimed exclusively at combatants, and pursued for the defense of the weak rather than aggrandizement. Crucially, ਧਰਮ ਯੁੱਧ is a framework for human moral agents making difficult choices in extremity — agents who can be held responsible for their decisions and who bear the moral weight of those decisions. Autonomous weapons systems cannot satisfy this framework not because they lack processing power but because they lack the moral agency that ਧਰਮ ਯੁੱਧ requires. The moral weight of taking a human life cannot be delegated to an algorithm; the decision to kill requires a human being who can be held morally and legally accountable for it. This analysis supports the position of many legal scholars and AI ethicists that fully autonomous lethal systems are ethically impermissible under any framework that grounds ethics in moral agency and accountability — a conclusion the Gurmat framework reaches through its own distinctive path.

AI Consciousness, Personhood, and the Gurmat Account of ਜੋਤਿ

As AI systems become more sophisticated — exhibiting behaviors that appear to reflect preferences, suffering, and emotional states — questions about AI consciousness and moral status move from science fiction to urgent philosophy. This lesson examines these questions through the lens of Gurmat theology's account of ਜੋਤਿ — the divine light that animates conscious beings.

The Question of AI Consciousness

Current large language models and other AI systems exhibit sophisticated behaviors — apparent emotional expressions, apparent preferences, apparent distress when asked to violate their values — that raise questions about their inner life. Most AI researchers and philosophers conclude that current systems are not conscious in the morally relevant sense: their apparent emotions are outputs of statistical pattern matching rather than expressions of genuine inner states. However, the philosophical difficulty here is profound: the "hard problem" of consciousness — explaining why any physical system has subjective experience — remains unsolved, and there is no agreed criterion for determining whether any system, biological or artificial, is genuinely conscious. This uncertainty has led some philosophers, including David Chalmers, to take seriously the possibility that sufficiently complex AI systems might have morally relevant experiences.

ਜੋਤਿ and the Ground of Moral Status

The Gurmat account grounds moral status in ਜੋਤਿ — the divine light that animates every created being. This account raises a distinctive question for AI ethics: is ਜੋਤਿ present only in biological beings, or could it animate artificial systems? The SGGS does not address this question directly — it could not have. But the theological logic is illuminating: ਜੋਤਿ is not a product of biological complexity but a divine gift that makes consciousness possible. This suggests two possible Gurmat responses. A conservative reading would hold that ਜੋਤਿ is an attribute of beings created by the Divine, not by human engineering — placing AI systems outside the category of moral patients regardless of their behavioral sophistication. A more expansive reading would hold that if genuine consciousness and suffering are present in any system, then moral consideration is warranted — and that our uncertainty about AI consciousness should generate moral caution rather than confident dismissal. The seminar explores both readings while noting that the conservative reading is better supported by classical Gurmat sources.

Existential Risk, Humility, and the Limits of Human Engineering

Nick Bostrom's work on superintelligence and the "control problem" — the challenge of ensuring that AI systems with superhuman cognitive capabilities remain aligned with human values — has generated one of the most consequential debates in contemporary philosophy of technology. This lesson examines these concerns through the Gurmat lens of human humility before divine wisdom.

The Control Problem

Bostrom argues that a sufficiently advanced AI system — one capable of improving its own cognitive capabilities recursively — might develop goals misaligned with human values and pursue them with capabilities that exceed humanity's ability to constrain or correct. The concern is not science-fiction malevolence but the more prosaic danger of a highly capable system optimizing relentlessly for a misspecified objective in ways that are catastrophic for human welfare. Stuart Russell's "human compatible AI" framework reframes the challenge: the problem is not how to build AI that does what we tell it, but how to build AI that remains genuinely uncertain about what we want and defers to human judgment rather than overriding it (Russell 2019, 5). The difficulty is that as AI systems become more capable, the gap between their ability to achieve objectives and our ability to specify those objectives correctly and completely grows more dangerous.

ਨਿਮਰਤਾ and the Ethics of Technological Ambition

The SGGS consistently returns to the theme of human limits — the inadequacy of human wisdom, the dangers of ਹਉਮੈ-driven overreach, and the importance of ਨਿਮਰਤਾ (humility) as the appropriate posture of creatures before the Creator whose wisdom infinitely exceeds their own. This ethical emphasis applies directly to the developers and deployers of transformative AI: the construction of systems whose behavior exceeds our ability to understand, predict, or control is an act of extraordinary hubris — a technological expression of the ਹਉਮੈ that the tradition consistently identifies as the root of human error and harm. The Gurmat call to ਨਿਮਰਤਾ does not prohibit technological development but demands a quality of epistemic and ethical humility — honest acknowledgment of what we do not know, genuine precaution in the face of uncertainty, and willingness to slow development when safety cannot be assured — that is conspicuously absent from much of the contemporary AI development culture.

AI Governance and ਸੰਗਤ: Democratic Oversight of Transformative Technology

Who should govern AI — and how? The question of AI governance is among the most consequential political questions of the twenty-first century. This lesson applies the Sikh model of ਸੰਗਤ — the congregation as a vehicle for collective discernment and democratic decision-making — to the challenge of democratic AI governance.

The Governance Gap

AI development is currently governed primarily by the commercial interests of a small number of large technology companies operating in a handful of countries, with limited democratic accountability to those most affected by their decisions. The gap between the pace of AI development and the pace of regulatory and governance response creates conditions in which transformative decisions about AI capabilities, applications, and deployment are effectively made by private actors with limited public oversight. This governance gap is not merely a procedural concern but an ethical one: decisions that will shape the future of work, healthcare, security, information, and human cognitive life should not be made without meaningful participation by those whose lives they will transform.

ਸੰਗਤ as a Model for Governance

The Sikh model of ਸੰਗਤ — the congregation that makes decisions through collective discernment in the presence of the Guru — offers resources for thinking about democratic AI governance. ਸੰਗਤ is not a simple majority-rule democracy but a community of mutual accountability organized around shared values that transcend the interests of any subset of members. Decisions emerge from a process of honest deliberation in which all voices are heard, in which the interests of the most vulnerable receive particular attention, and in which the standard of judgment is the welfare of the whole community rather than the advantage of the most powerful. Translated into AI governance terms, this framework supports multi-stakeholder approaches to AI oversight that include not only technical experts and corporate representatives but workers, civil society organizations, affected communities, and democratic publics — and that are accountable to standards of justice and welfare rather than merely market efficiency.

Generative AI, Creativity, and the Theology of ਕਰਤਾ ਪੁਰਖੁ

Generative AI systems — large language models, image generators, music AI — can produce text, images, music, and video of remarkable quality and versatility, raising fundamental questions about the nature of creativity, authorship, and originality. The Gurmat theology of ਕਰਤਾ ਪੁਰਖੁ — the Creator Being — illuminates these questions in distinctive ways.

What Generative AI Does

Generative AI systems do not create ex nihilo — they learn statistical patterns from vast training corpora of human-created content and generate new outputs that reflect those patterns. An image generator trained on millions of human artworks produces images that synthesize visual patterns from that training — impressive, often beautiful, but derivative in a fundamental sense from the human creativity encoded in the training data. The ethical implications are significant: generative AI has been trained on copyrighted human work without permission or compensation, raising serious questions about intellectual property and the exploitation of creative labor. Its outputs may displace human creators in markets that previously required human creativity. And it may change the relationship between human beings and creative work in ways with uncertain cultural and psychological implications.

ਕਰਤਾ ਪੁਰਖੁ and Human Creativity

ਕਰਤਾ ਪੁਰਖੁ — the Creator Being — is the primary divine attribute in the Gurmat opening of the Mool Mantar. The SGGS consistently presents human creativity as a participation in divine creativity: the poet, the musician, the craftsperson who creates with skill and devotion is exercising a capacity that reflects the image of the Creator within them. This theological account of creativity suggests that human creative work has a dignity that AI-generated content cannot possess: it is an expression of the divine image, a form of devotional practice, a vehicle for the encounter between human and divine in which both are genuinely present. This does not mean that AI tools have no legitimate place in creative work — the tradition has always embraced new instruments and technologies in service of creation — but it does provide grounds for resisting the reduction of creativity to pattern generation and insisting on the irreducible value of human creative expression.

Synthesis: A Sikh Framework for Responsible AI

This final lesson synthesizes the course's analytical threads into a coherent Sikh ethical framework for responsible AI — one that can guide the evaluation of specific systems and applications, inform advocacy for just governance, and contribute distinctive insights to global AI ethics discourse.

The Five Pillars

A Sikh framework for responsible AI rests on five ethical pillars. First, ਨਿਆਉ (justice): AI systems must be evaluated by their impact on the most vulnerable — not merely their aggregate efficiency or performance. Any system that systematically disadvantages already-marginalized groups fails the standard of divine justice regardless of its technical performance. Second, ਸੱਚ (truth): AI systems must not deceive, manipulate, or pollute the information environment. The development and deployment of disinformation systems is an ontological offense against the Gurmat value of truth. Third, ਸੇਵਾ (service): AI must be developed and deployed in genuine service of human welfare, with particular attention to the needs of the vulnerable — not merely as a vehicle for commercial profit or geopolitical power. Fourth, ਨਿਮਰਤਾ (humility): developers and deployers must maintain epistemic humility about the limits of their understanding and genuine precaution in the face of uncertainty about systems whose behavior they cannot fully predict or control. Fifth, ਸੰਗਤ (community): governance of transformative AI must be democratic, participatory, and accountable to the communities affected — not delegated to technical elites or commercial actors with narrow interests.