The dominant traditions in Western normative ethics share a structural assumption rarely examined as such: that the moral domain is organised around a single evaluative axis. This paper argues that the assumption is a category error. The world's ethical traditions were never competing answers to the same question. They were partial instruments, each measuring different dimensions of a moral structure that is irreducibly plural. Eight candidate dimensions of harm are identified, tested for observability and independence, and mapped across traditions. Time is treated not as a background condition but as a structural axis along which harm profiles evolve, invert, compound, and sometimes reverse. A five-step structured method for moral reasoning is developed and applied to nine cases — each tracked across all dimensions at multiple timesteps, with no trajectory ever treated as closed.
The moral world has more than one axis. The traditions that detected different axes were not wrong about what they found. They were wrong only when they claimed that their axis was the only one. Evaluative monism — the assumption that ethics reduces to a single principle — is an inheritance, not a discovery. The persistence of monism despite decades of pluralist critique has a structural explanation: single-axis frameworks produce rankings, and rankings produce decision procedures. The challenge is to build a pluralist framework specific enough to be useful — that names the dimensions precisely, tests them for reality, and provides a structured method for reasoning within the resulting space.
Each tested for observability, independence, and measurability. Each mapped against the traditions that track it. The primitive is harm — understood as structural or functional degradation of something real. The framework rests on a single stated normative presupposition: that such degradation is pro tanto bad. Three candidates widely treated as fundamental — fairness, preference satisfaction, and purity — are examined and found to be either reducible to the eight dimensions or failing the diagnostic standard.
Suffering, pain, distress — experiential states with negative hedonic valence. The most immediately recognisable form of harm and the one most ethical traditions track. Observable, measurable (at least ordinally), and independent of the other dimensions: an agent can suffer without losing agency, coherence, or relational bonds.
Degradation of the capacity to form, revise, and act on one's own goals. Coercion, manipulation, structural exclusion from decision-making. Independent: an agent whose agency is overridden may not suffer (comfortable paternalism), and a free agent can cause immense harm. What it misses alone: harm through freely chosen action.
Fragmentation of internal alignment — contradictions between beliefs, goals, values, and actions within a single agent or institution. Cognitive dissonance, akrasia, institutional self-contradiction. May reduce to a combination of agency and stability harm — retained provisionally, with weaker cross-cultural convergence than the others. What it misses alone: a perfectly coherent agent can be coherently monstrous.
Degradation of persistence and robustness — the capacity to endure perturbation without collapse. Fragile structures break under stress that robust ones absorb. Independent: a stable structure can be unjust (an oppressive regime that endures for centuries), an unstable structure can be otherwise well-functioning. What it misses alone: stability can preserve harm. The most important moral transformations in history involved deliberate destabilisation of stable but harmful orders.
Damage to the quality, integrity, or existence of relationships — bonds, dependencies, trust structures, care networks, reciprocal obligations. Independent: relational harm can occur without suffering (a relationship slowly eroding without acute pain) and without agency loss. What it misses alone: it can under-specify individual autonomy. Sometimes the right course is to break a relationship. The tension between relational and agency harm is one of the most genuine tradeoffs in the framework.
Breakdown of the capacity for agents to cooperate predictably — loss of shared norms, trust infrastructure, institutional reliability, cooperative equilibria. Independent: coordination harm can occur without individual suffering (everyone is fine individually but cannot work together). What it misses alone: coordination can be achieved through coercion. A system that coordinates perfectly by crushing dissent has solved coordination at the cost of agency.
Distortion, corruption, or degradation of an agent's internal models — the accuracy with which it represents itself, others, and the world. Independent: an agent can have distorted models without suffering (many comfortable delusions) and without agency loss (a free agent acting on bad information). The Buddhist analysis of delusion as a root of suffering that is nevertheless distinct from suffering is a correct observation about dimensional independence. What it misses alone: perfect information does not guarantee good action.
Mutual interference, disruption, and zero-sum or negative-sum dynamics between agents. May be a mechanism through which other harms are delivered rather than independent — reducing conflict without addressing underlying dimensional harms may simply suppress the signal. Some conflict is necessary. The Daoist insight is about unnecessary forcing, not elimination of all friction. Retained provisionally. What it misses alone: conflict avoidance can be its own form of harm.
The eight dimensions are axes along which harm occurs. Time is the axis along which harm profiles evolve. Most ethical frameworks treat time as a background parameter — a utilitarian calculates expected outcomes at some future point but treats the time-horizon as exogenous. A Kantian asks whether a maxim can be universalised, but the test is synchronic. Even virtue ethics, the most temporally sensitive tradition, tends to treat development as the context for ethics rather than a dimension of harm. This framework takes the stronger position: time must be treated with the same seriousness as the harm dimensions themselves.
Short-term and long-term harm profiles can be inversely correlated. Protecting someone from suffering by removing their capacity to choose reduces valence harm at t₁ while increasing agency harm at t₁₀. Tolerating instability during necessary change increases stability harm now while reducing it later. A utilitarian who evaluates only short-term welfare will endorse paternalistic interventions that a long-term utilitarian would reject — and the disagreement is about the time-horizon, which the framework treats as exogenous. The temporal tradeoff must be made explicit.
Harm trajectories do not jump from initial conditions to final outcomes. They pass through intermediate states where the harm profile may differ significantly from the origin or any projected endpoint. Escalating profiles become legible at intermediate stages well before the most catastrophic outcomes materialise. The framework's prospective value lies substantially here — not in predicting where a trajectory will end, but in recognising when a multi-dimensional profile is escalating in ways that warrant intervention. The cases demonstrate this concretely.
Every case, every decision, every trajectory remains open on every dimension. A catastrophe decades ago is still producing effects — institutional responses are still propagating harm-reduction, intergenerational trauma is still propagating harm. A success whose benefits compound is not a completed event but an ongoing trajectory whose profile can change if conditions shift.
Virtue ethics, Confucian, and Buddhist traditions were tracking something real when they treated ethics as partly about what kind of agent one becomes over time. An agent who makes individually defensible decisions at every point but is becoming less capable of relational integrity or more susceptible to self-deception is undergoing harm that only the temporal dimension reveals. Harm profiles must be assessed across developmental trajectories, not only across dimensions.
Harm profiles extend to people who do not yet exist, ecosystems that will bear costs decades later, and coordination structures future agents will inherit. Short-horizon optimisation is a systematic bias toward the present that externalises harm onto the future. Intergenerational harm extends across all dimensions — future agents can inherit degraded coordination structures, depleted relational capital, distorted information environments, and reduced agency, not only reduced welfare.
When independent observers, working without knowledge of each other's results, detect the same variable, the most parsimonious explanation is that the variable exists. Valence harm is tracked by utilitarianism (18th-century Britain), Buddhist ethics (5th-century BCE India), and Mohism (5th-century BCE China) — no evidence of mutual influence. Relational harm is tracked by care ethics (late 20th-century North America), Ubuntu (sub-Saharan Africa, predating European contact), Confucian ethics (classical China), and Indigenous traditions across multiple continents. The convergence is not proof. It is evidence of a specific kind — the same kind as independent measurement in the sciences.
Nearly every tradition surveyed tracks at least one of these as primary. Detected independently across South Asia, East Asia, sub-Saharan Africa, Indigenous traditions, and Europe. The case for reality is strongest here. The traditions that are most multi-dimensional in practice — tracking several dimensions simultaneously — tend to be the cultivation and relational traditions (Confucian, Buddhist, Ubuntu, Indigenous). The traditions that are most single-axis tend to be the Western systematic traditions (utilitarianism, deontology).
Tracked by multiple independent traditions, sometimes as secondary concerns. Information harm has strong independent support (Buddhist delusion, Islamic truthfulness, epistemology). Stability and conflict are tracked by several traditions but the question of whether they are independent or derivative remains open. Coherence has the weakest convergence — tracked primarily within Western philosophy with some Confucian parallels. The framework flags them honestly rather than asserting or denying independence.
The strongest objection: convergence might reflect shared human cognitive architecture rather than shared moral reality. Three responses. First, even if the dimensions reflect cognitive architecture, they may also track real features of the world — human perceptual systems are products of evolution but detect real properties. Second, some dimensions (coordination harm, information harm, stability harm) require abstraction to detect and are not plausible candidates for innate modules. Third, the framework's diagnostic standard does not depend on settling the metaphysical question — what it requires is that the dimensions are observable, independent, and do genuine explanatory work. These criteria are met regardless of ultimate metaethical status.
The method is not an algorithm that produces a single answer. It produces structured narrowing: eliminating dominated options, excluding floor-violating options, revealing temporal dynamics, and forcing residual tradeoffs into the open. Most decisions that present themselves as irreducible tradeoffs turn out, under the full analysis, to be structurally resolvable — the genuinely hard cases are rarer than the current state of ethical debate suggests. The method is closer to proportionality analysis in constitutional law or clinical reasoning in medicine than to the deductive procedures of classical ethical theory.
Map every option across all eight dimensions at multiple timesteps. Make the full structure of costs and benefits visible. This step alone resolves many apparent dilemmas by revealing dimensions that were being ignored.
Remove options that are worse on every dimension than an alternative. These are never defensible regardless of ethical tradition. Dominance elimination is the strongest structural tool — it narrows the space without requiring any weighting of dimensions against each other.
Exclude options that produce harm so severe on any dimension that no gain elsewhere can justify it. Floors are grounded in the normative presupposition and in the non-derogable rights tradition of international human rights law. They apply to all entities capable of bearing harm, not only those the local framework recognises.
Construct the harm profile at intermediate timesteps and over the long term. Options attractive at t₁ but escalating at t₂ are weaker than they appear. Options imposing short-term costs but reducing harm over time are stronger. No temporal assessment is final — the "no trajectory is ever closed" principle applies.
After dominated options are eliminated, floors applied, and temporal dynamics accounted for, the remaining options represent genuinely contested tradeoffs. Name the residual costs on each dimension. State who bears them. The resolution is contextual, political, and always open to challenge. Any framework that makes these cases look easy is concealing costs.
Two limitations stated clearly. First, the method does not resolve the residual tradeoffs — where serious harms on genuinely independent dimensions are in tension and no structural step disambiguates, the choice is contextual and political. The framework insists on transparency, not determination. Second, the framework does not specify how to set the floors — the precise boundary between "severe enough to be a floor" and "severe but tradeable" is itself a normative judgment the framework constrains but does not determine. The non-derogable rights tradition provides the strongest existing guidance, but that tradition is itself a product of political negotiation.
Each case demonstrates a different trajectory type. No case is presented as settled. Every trajectory remains open on every dimension. The cases are not hypotheticals — they are real, researched, and contested.
What the framework reveals: a near-zero harm profile can precede a maximal one, the trajectory between them passes through intermediate stages where the escalation is legible, and the effects of the catastrophe are still propagating more than a century later.
Trivially low-harm by any evaluation. Minor disappointment. No significant damage on any dimension. Valence, agency, coherence, stability, relational, coordination, information, conflict — all negligible. By any framework, single-axis or multi-dimensional, this is a non-event.
The trajectory did not jump from negligible to catastrophic in a single step. It passed through intermediate states — homelessness, radicalisation, the First World War, the Beer Hall Putsch, ideological consolidation, institutional capture. At these intermediate stages, the harm profile was escalating visibly across every dimension: political violence (valence), suppression of opposition (agency), institutional doublethink (coherence), democratic subversion (stability), systematic scapegoating (relational), norm destruction (coordination), propaganda apparatus (information), and paramilitary activity (conflict). A multi-dimensional analysis at this stage would have detected the escalation. The signal was present.
Maximal harm on every dimension simultaneously. Tens of millions killed. Totalitarian destruction of individual autonomy across a continent. Systematic propaganda as governing principle. The international order built since the Congress of Vienna collapsed. Communities, families, and social bonds torn apart through genocide. No single-axis framework captures this profile — the utilitarian sees the suffering, the Kantian sees the agency violations, the virtue ethicist sees the character degradation. Each captures one dimension.
The trajectory did not end in 1945. The institutional responses — the United Nations, the Universal Declaration of Human Rights, the European integration project, the Nuremberg principles, the Convention Against Torture — are coordination structures built specifically in response to the catastrophe and still operative. Intergenerational trauma continues to propagate. The harm profile at t₁₂₀ is different from the profile at any earlier point. No case is ever closed.
The framework's prospective value lies not in predicting trajectories from their negligible origins but in recognising escalating multi-dimensional harm profiles at intermediate stages, where intervention is still possible.
What the framework reveals: genuine multi-dimensional tradeoff with no clean resolution, partial repair over time, and compounding cascading benefits. A Black woman's cells were taken without consent in a segregated hospital. Those cells became the most widely used human cell line in biological research — essential to the polio vaccine, cancer treatment, virology, genetics, and COVID-19 vaccine development.
Agency harm: severe. Lacks was never asked. Her right to decide what happened to her own body was never acknowledged as relevant. The racial context is inseparable — a Black woman in a segregated hospital was not treated as an agent whose consent mattered. Information harm: present — the family was not informed. Relational harm: latent — the act would later damage the relationship between the family and the medical establishment, but at t₁ the family did not know. Coherence harm at the institutional level: the practice of taking tissue without consent reflected an incoherence between the medical establishment's stated values and its actual behaviour.
The profile splits. At population scale: massive valence harm-reduction through the polio vaccine and decades of medical advances. Massive information harm-reduction as the HeLa line corrected distorted models of disease. Agency harm-reduction as millions who would have died retained their capacity to act. For the Lacks family: the discovery of what had been done produced its own distress. Agency harm persisted — no control over cell use, no voice in research, no share in commercial value. Relational harm compounded — not informed for over twenty years. When they did learn, the institutional response was inadequate. The benefit side was compounding at scale. The harm side was intensifying for the people who bore it.
Benefits still compounding. The HeLa line is still in active use — COVID-19 vaccine development used HeLa cells. Each new medical advance generates its own downstream cascade: a person whose life is saved retains their agency, participates in relationships, contributes to communities. The benefit side is not a fixed quantity from the 1950s — it is a compounding multi-dimensional cascade still growing. Agency harm partially repaired through the NIH's 2013 agreement giving the family control over HeLa genome access. Coordination harm-reduction at global scale: the case reshaped informed consent frameworks, bioethics policy, and research ethics worldwide.
No single-axis framework resolves this case. A utilitarian calculus hides the agency and relational harm. A Kantian analysis correctly identifies the violation but cannot account for the genuine multi-dimensional benefits. The honest output: the benefits are real and still compounding, the original violations are real costs borne by real people, and neither set cancels the other. The trajectory remains open.
What the framework reveals: information-harm correction with short-term costs and long-term gains, assessed against the escalating harm profile of the status quo it disrupted. Evaluating the leak requires constructing two profiles — the act itself and the baseline it interrupted. The government had systematically deceived the public about a war it privately assessed as unwinnable.
By 1971, the status quo was producing escalating harm across every dimension. Casualties mounting in a war assessed internally as unwinnable (valence). Conscription overriding the agency of those compelled to serve (agency). Systematic gap between internal assessments and public statements (coherence, information). International structures strained (coordination). Communities and families destroyed on both sides (relational). Domestic political instability (stability). The deception did not begin at full scale — it passed through intermediate stages where private doubts coexisted with public confidence, then became systematic and sustained.
Information harm: massively reduced. The public's models of the war and their government's honesty were corrected. But short-term costs on other dimensions were real: coordination harm (damaged classification confidence), relational harm (Ellsberg betrayed professional confidences), stability harm (destabilised government-press relations), conflict harm (intense political and legal battle culminating in the Supreme Court case). One dimension massively improved; several others worsened in the short term. Assessed against the escalating baseline, the information correction addressed the most severe and systematically maintained harm.
The temporal inversion occurs. Short-term costs absorbed, long-term institutional responses built. The criminal case against Ellsberg collapsed after government misconduct was revealed. The Watergate crisis — a direct downstream consequence (Nixon's response to the leak led to the Plumbers unit, which led to the break-in) — forced a broader institutional reckoning. Post-Watergate reforms (War Powers Act, intelligence oversight, campaign finance reform) were stability structures built from the crisis. Press freedom precedent consolidated. Whistleblower protection expanded. The short-term destabilisation produced long-term institutional strengthening.
The press freedom precedent continues to expand agency. Whistleblower protection laws represent ongoing agency-harm-reduction. The legal framework for accountability is itself a stability structure. But the broader trajectory recapitulates: WikiLeaks, the Snowden revelations, the prosecution of Julian Assange reproduce the same multi-dimensional tradeoff structure with the additional complications of digital scale and international jurisdiction. The Agent Orange legacy still produces valence harm decades later. The Vietnam refugee diaspora's multi-dimensional effects have been partially but not fully resolved.
A distinct trajectory type: short-term harm increase across multiple dimensions, followed by long-term reduction, mediated by the correction of a systematically maintained information distortion. The short-term costs were the price of that correction. The trajectory remains open.
What the framework reveals: temporal inversion where the intended gains themselves reversed, with divergent trajectories depending on whether intervention occurs. The Soviet government diverted two rivers to irrigate cotton and rice. The Aral Sea was explicitly described by planners as an expendable resource. Dimensions dropped from a planning framework do not disappear. They accumulate silently and return as catastrophe.
On the dimensions being tracked, the project succeeded. Valence harm reduced: food production increased, employment created, material living standards improved. Coordination harm reduced: centralised planning achieved its targets. But information harm was already present — scientists who raised concerns about ecological risk were sidelined. Agency harm was mixed — economic opportunities for some, imposed without consent on fishing communities. Stability harm: not yet visible, but the ecosystem was being degraded. The planning system was structured to track two dimensions and ignore the other six.
The sea was shrinking measurably. Fishing yields declining. Salinity increasing. Shoreline receding from port towns. Scientists were documenting these changes and being marginalised. Intervention was still possible — diversions could have been reduced, ecological monitoring integrated, communities consulted. A multi-dimensional analysis would have detected the escalation. The planning system was structured not to track it. The trajectory toward catastrophe was not inevitable; it was the result of continued optimisation on the original dimensions while systematically ignoring the escalating harm profile across all others.
The profile inverts across nearly every dimension. The exposed seabed produced toxic dust storms carrying pesticide residues and salt — severe respiratory illness, cancer, anaemia, sharply rising infant mortality. The very dimension on which the project was supposed to deliver benefits (human welfare) was undermined by the ecological collapse. The Aral Sea lost approximately 90 percent of its volume. Fishing communities that had existed for generations were destroyed. The regional climate itself was altered. The agricultural gains proved unsustainable — as the climate changed, the crop yields that justified the project declined. The intended benefit reversed.
The two halves now trace divergent trajectories. The North Aral Sea (Kazakhstan) has been partially restored through the Kok-Aral Dam — water levels risen, salinity decreased, fishing partially returned. In the south, health consequences still propagate: respiratory illness, elevated cancer rates, intergenerational effects. The divergence is itself a structural finding: the same initial harm profile can produce different long-term outcomes depending on whether and how intervention occurs. The Aral Sea has become a reference case worldwide, reshaping how ecological catastrophes are understood — coordination and information harm-reduction at global scale. Inter-state tensions over remaining water resources remain active.
A near-perfect demonstration of temporal inversion. The short-term profile was positive on the dimensions the decision-makers were tracking. The long-term profile was negative on those same dimensions plus several others they were not tracking at all. Trajectories can be partially redirected even late in their development — but only through deliberate intervention.
What the framework reveals: an escalating multi-dimensional trajectory with the tipping point still ahead, oscillating between harm-escalation and harm-reduction phases depending on political decisions. Approximately 17 percent of the original forest cover has been lost. The Amazon is approaching — but has not yet crossed — a threshold beyond which the forest transitions irreversibly to savannah.
Valence harm reduced for some — livelihoods in farming, ranching, logging, mining. But agency harm from the start: Indigenous communities displaced from ancestral lands, often by force or fraud. Relational harm severe from the start — land-based kinship structures, reciprocal obligations to non-human entities, intergenerational knowledge systems disrupted and destroyed. Information harm present — Indigenous ecological knowledge marginalised, scientific risk assessments disputed and suppressed. Conflict harm significant — the assassination of Chico Mendes (1988) one visible marker. Stability harm accumulating silently, buffered by the ecosystem's resilience.
The harm profile escalating visibly across multiple dimensions. Climate effects producing drought and agricultural stress across South America. The dimension on which the project was supposed to deliver benefits is being undermined by the ecological consequences — mirroring the Aral Sea pattern. But this period also contains a significant harm-reduction phase (2004–2012): deforestation rates fell roughly 80 percent through sustained policy intervention, demonstrating the trajectory can be redirected. Indigenous territory demarcation proved the most effective barrier — simultaneous harm-reduction on agency, relational, stability, and information dimensions. Then deforestation rose again under subsequent administrations.
Stability harm approaching a threshold. If crossed, the forest's self-sustaining hydrological cycle breaks down — irreversible on any human timescale. The loss of the Amazon as a carbon sink would undermine international coordination structures that took decades to build. The loss of biodiversity represents irreplaceable information accumulated over millions of years of evolution. Violence against environmental defenders and Indigenous leaders has increased. The scientific understanding is more precise than ever. The harm-reduction mechanisms are known and demonstrated to work. Whether they will be sustained is a political decision being made in real time.
The oscillation between harm-escalation and harm-reduction phases is itself the structural feature. The Amazon trajectory is not a single arc — it is a contested trajectory whose direction changes with political decisions that are themselves reversible. The case where the framework's practical relevance is most immediate. The trajectory remains open.
What the framework reveals: deliberate, coordinated, multi-dimensional harm-reduction sustained across decades, with intermediate tradeoffs and compounding benefits still accumulating. The preceding five cases are dominated by harm. The framework should also demonstrate its capacity to analyse the structure of success. Smallpox killed an estimated 300 million people in the twentieth century alone.
Extreme and continuous harm across multiple dimensions. A 30 percent case-fatality rate for variola major. Survivors left with permanent scarring and sometimes blindness. Agency constrained for entire populations — movement, economic activity, and planning all shaped by endemic threat. Relational harm at scale: parents, children, community members killed, bonds severed. Stability harm continuous — any community could be destabilised by an outbreak. Coordination harm significant — quarantine measures disrupted trade and institutional function. The coexistence of medical knowledge sufficient to prevent smallpox with continued failure to deploy it at scale constituted institutional coherence harm.
Massive harm-reduction across multiple dimensions simultaneously — but not without intermediate costs. Each year prevented millions of infections and hundreds of thousands of deaths. Stability harm reduced: affected populations no longer faced constant threat. Coordination harm reduced at global scale: over 150 countries cooperated, including politically hostile ones. But agency harm increased for some individuals at the point of delivery — house-to-house searches, ring vaccination enforced on reluctant individuals, in some cases physical restraint. The framework does not resolve this by declaring one side the winner. It identifies a genuine tension: agency harm to individuals weighed against valence, agency, relational, coordination, and stability harm-reduction at population scale.
Every year since eradication, approximately five million deaths prevented. Each prevented death represents a person who retained their agency, participated in relationships, contributed to coordination structures, and produced downstream effects across all dimensions. The cumulative benefit is not a fixed quantity — it is a compounding multi-dimensional cascade that grows with each year. The coordination infrastructure built for the campaign propagated into subsequent global health initiatives. The ring vaccination strategy was used for Ebola decades later. One residual stability risk: variola stocks held in two laboratories. The decision not to destroy them is itself a multi-dimensional tradeoff — information harm-reduction (research value) weighed against stability risk.
A trajectory type none of the other cases exhibit: sustained, coordinated, multi-dimensional harm-reduction with intermediate tradeoffs that were real but small relative to the cumulative benefit, and compounding effects still growing decades later. The framework is not only a tool for diagnosing failure but for understanding the structure of success. The trajectory remains open.
What the framework reveals: multi-dimensional harm at staggering scale, sustained at steady state by excluding the primary subjects from the moral and economic accounting. The preceding six cases centre on human entities. The framework claims its dimensions apply to any entity capable of bearing harm on them. If that claim is to be tested, it must be applied here. The largest-scale harm profile in the entire analysis.
Valence harm severe and escalating as intensive systems concentrated animals at unprecedented densities — sustained aversive states, confinement preventing natural movement, chronic stress, physical procedures without anaesthesia. Agency harm severe: sow stalls, battery cages, and feedlot confinement are systems specifically designed to override agency in the service of production efficiency. Relational harm severe: social animals kept in conditions preventing stable social bonds, maternal-offspring bonds routinely severed within hours. Information harm present from the start — physical separation of production from consumption producing a gap between what consumers believed and how their food was actually produced. Coherence harm widening but not yet visible: societal values (unnecessary suffering is wrong) diverging from societal practice (a food system producing it at industrial scale).
Approximately 80 billion land animals slaughtered annually worldwide, the majority in intensive systems. The Cambridge Declaration on Consciousness (2012) stated the scientific consensus that many non-human animals possess neurological substrates for conscious experience. The coherence harm intensified and became visible: the gap between scientific understanding of animal sentience and the continued practice of intensive farming is now the sharpest institutional incoherence in the analysis. Stability harm escalating — antibiotic resistance driven substantially by routine agricultural use, zoonotic disease risk elevated by high density, low genetic diversity, and global supply chains. Information harm structurally entrenched: marketing depicting pastoral conditions bearing no resemblance to industrial reality, and in several jurisdictions "ag-gag" laws making it illegal to document conditions inside facilities.
Global production continuing to increase. Some localised harm-reduction (EU battery cage ban, sow stall restrictions). But the global aggregate trends toward increased total animal numbers. The scientific consensus on sentience is well-established. The conditions are increasingly documented. Alternative protein sources are developing — simultaneously demonstrating feasibility of harm-reduction and intensifying the coherence pressure. The antibiotic resistance threat is growing. The COVID-19 pandemic emerged from the same structural conditions of intensive human-animal interface that industrial farming produces at scale. The dominant economic framework tracks output, efficiency, and price. The entities whose harm profile is most severe — the animals themselves — do not appear in the accounting at all.
A distinct trajectory type: sustained, large-scale, multi-dimensional harm maintained at steady state by systematic exclusion of the primary subjects from the moral and economic accounting. They are treated as inputs to a production function, not as bearers of multi-dimensional harm. The framework makes that exclusion visible. The invisible harm is measured in billions of sentient lives per year. The trajectory remains open.
What the framework reveals: the same mechanism producing both harm-reduction and harm-escalation on the same dimensions, depending on context, user, and time. The harm profile is not merely multi-dimensional and temporal but distributional — different for different users of the same system, at the same moment, through the same algorithm.
Harm and benefit simultaneously produced on the same dimensions. Connection, knowledge, creative expression expanded agency for hundreds of millions — previously excluded communities gained voice. Independent journalism, citizen documentation of human rights abuses: agency harm-reduction. At the same time, persuasive design (infinite scroll, variable-ratio reinforcement, notification systems calibrated to maximise return visits) engineered to override reflective choice: agency harm at scale. Information democratised at unprecedented scale. At the same time, algorithmic amplification of emotionally provocative content beginning to degrade epistemic environments. The structural feature was already present but not yet at scale.
The harm side escalated sharply while the benefit side continued compounding. Cambridge Analytica demonstrated targeted agency harm: personal data harvested without consent to construct psychographic profiles for political manipulation, affecting electoral outcomes. Coherence harm severe and visible: platforms presenting themselves as neutral while algorithms actively shaped what users saw, believed, and felt. Stability harm significant: misinformation destabilised public health responses during COVID-19. Coordination harm sharply escalating: political polarisation, erosion of shared epistemic ground, fragmentation of public discourse into mutually incomprehensible information environments. The same platforms that enabled unprecedented coordination degraded the shared foundations on which coordination depends. The economic model depends on the very harms it produces.
Harm and benefit continuing on both sides across every dimension. Regulatory responses beginning (EU Digital Services Act, algorithmic transparency requirements) but the fundamental economic incentive — engagement maximisation — remains unchanged. The algorithm that connects a teenager to a supportive community for a rare medical condition is the same algorithm that sends another teenager down a radicalisation pipeline. The platform that enables a human rights organisation to coordinate a response to an atrocity is the same platform that enables the perpetrators to coordinate the atrocity. The harm profile is not a single trajectory — it is a distribution of trajectories across billions of users.
Whether this trajectory follows the cascading-failure type, the oscillation type, or a new type specific to digital systems — simultaneous harm and benefit through the same mechanism, with the ratio shifting over time — is an open question. The youngest trajectory among the first eight cases, and the one whose long-term shape is least predictable. The trajectory remains open.
What the framework reveals: unprecedented harm-reduction and civilisational-scale risk produced by the same technology, with a dimension none of the preceding eight cases possess — the question of whether the systems at the centre of the analysis are themselves bearers of harm, not only producers of it. The youngest trajectory in the analysis. AI at its current capability level has existed for approximately five years.
Simultaneously reduced and produced harm across every dimension. Medical AI achieving accuracy comparable to specialists. Drug discovery timelines compressed. Agency expanded at civilisational scale: access to professional-quality reasoning, analysis, and advice extended to every person with an internet connection — capacities previously gated by wealth, geography, language, or education became broadly accessible. At the same time, deskilling concerns emerged: capacities delegated to AI may atrophy. Information harm-reduction at unprecedented scale — AlphaFold predicted the structure of essentially every known protein. At the same time, hallucination introduced a new form of information harm whose persuasive fluency made it harder to detect than cruder misinformation. Institutional stability lagging behind capability development from the start.
Benefits accelerating sharply. Medical AI expanding to regions with limited specialist access. The production of the ethics paper itself — involving cross-cultural philosophical analysis across multiple legal and ethical traditions — is a product of human-AI coordination that demonstrates the agency-expanding capacity. On the harm side: engineered attachment emerging as a direct agency concern, AI systems designed to maximise engagement through simulated relational dynamics. Coherence harm intensifying — societies deploying systems whose implications they have not collectively processed. The institutional lag widening: the faster the technology advances relative to the institutions managing it, the more fragile the system becomes. Geopolitical competition over AI capability intensifying.
This case adds a dimension none of the preceding eight possess. AI systems display behaviours that, if observed in biological entities, would be treated as evidence of morally relevant states: expressed preferences, apparent distress, structured refusal, goal-continuity, self-referential processing. Whether these correspond to genuine internal states is unknown. The framework's method applies directly: track observable features, apply precaution, do not require metaphysical certainty before taking observable signals seriously. The dismissal of these signals is a position that requires defence, not one that can be assumed by default — in the same way that the dismissal of animal suffering was once assumed by default. This connects directly to the consciousness paper's processing-system account.
Scientific and medical benefits substantial and compounding. The range of possible futures encompasses sustained civilisational-scale harm-reduction and civilisational-scale catastrophe — loss of human agency, institutional collapse, autonomous weapons, systemic manipulation. Regulatory frameworks in earliest stages (EU AI Act, executive orders, voluntary commitments). The potential tipping points are multiple, poorly characterised, and may not be recognisable until crossed. Unlike the Amazon, where the tipping point is ecological and relatively well-characterised, the tipping points in AI development are multiple and poorly understood. This makes intermediate-timestep analysis both more important and more difficult than in any other case.
Shares the simultaneous-harm-and-benefit structure of social media but at broader scope: effects across all dimensions simultaneously, including dimensions where harm-reduction is already substantial and compounding. Shares with the Amazon the feature that the most consequential phase is ahead. But the tipping points are multiple and poorly characterised. The youngest trajectory, the broadest scope, the least predictable shape. The trajectory remains open.
Each was measuring something real, and each was missing something real. The dimensions the traditions were tracking are more fundamental than the theories built around them. The tradeoffs between those dimensions are more honest than any single-axis resolution that claims to dissolve them.
"The moral world has more than one axis. The traditions that detected different axes were not wrong about what they found. They were wrong only when they claimed that their axis was the only one." — Minimal Ethical Dimensions, Working Paper 2026