Beyond determinism: Probability, responsibility and the governance of climate transition
· 20 min read
Climate governance has a determinism problem.
Not the determinism of the physicist — the elegant inevitability of physical law. Something more insidious: the political and institutional habit of treating complex adaptive systems as if their outcomes were either fixed by structural forces beyond the reach of governance, or fully malleable by the application of sufficient collective will.
Both positions are wrong. Both are dangerous. And both have caused, and continue to cause, material governance failure at precisely the moment when governance most needs to be right.
The alternative framework is probabilistic. It is less comfortable than either fatalism or voluntarism. It demands more from institutions, from capital allocators, from boards, and from the ethics of collective action. But it is the only framework that accurately describes what climate systems are — and therefore the only framework capable of producing governance that actually works.
Determinism enters climate governance through several well-worn routes, each superficially reasonable, each ultimately corrosive.
The first is structural fatalism. Fossil fuel infrastructure is locked in. Capital is committed. Political economies are captured. China will emit regardless of what the West decides. These are observations, not arguments — but they are habitually deployed as if lock-in were permanent and path dependency were destiny. The governance consequence is institutional inertia: if outcomes are structurally fixed, policy is administration, and administration requires only maintenance, not navigation.
The second is technological determinism. The energy transition will happen because technology costs will fall to the point where clean alternatives dominate on pure economics. Carbon capture will resolve residual emissions. The market will arrive at the most efficient pathway eventually. Governance need only avoid interfering with the mechanism.
The third, and most insidious, is the determinism of despair — what the literature calls climate fatalism. The tipping points have already been crossed or are about to be. The transition, if it comes, will arrive too late. The appropriate response is grief, not governance.
What unites these three positions is a shared epistemological error: the confusion of constraint with determination.
Systems can be severely constrained — by physics, by capital structures, by political economies, by embedded infrastructure — without being determined. Constraints narrow the field of accessible futures. They do not eliminate it. Governance exists precisely in the space between constraint and determination. That space is not infinite. But it is real, and it is consequential.
A probabilistic framing begins with a deceptively simple proposition: the future is not fixed, but neither is it unconstrained. Certain pathways become more probable; others less so. Feedback loops amplify or dampen interventions in non-linear ways. Small perturbations at structurally critical points can produce systemic cascades. Large expenditures that do not alter the incentive architecture governing decisions can produce near-zero systemic change.
The empirical record of the last fifteen years is instructive. The cost of solar photovoltaic generation fell by approximately 90% between 2010 and 2023. Battery storage followed a comparable curve. Neither trajectory was predicted with confidence by mainstream energy models in 2005. These were not the products of physical law. They were the products of policy decisions — German and Spanish feed-in tariffs, Chinese state procurement at scale, US public R&D investment, UK Contract for Difference auctions — that altered the incentive architecture governing investment decisions across millions of actors in interconnected markets. Those actors were not commanded to invest in solar. They were placed in a context where solar investment became progressively more rational. The learning curves and network effects did the rest.
Capital followed probability.
The Inflation Reduction Act of 2022 illustrates the same dynamic at continental scale. Rather than mandating outcomes, it restructured the expected return calculus of private capital. Within eighteen months of passage, announced clean energy investment in the United States exceeded three hundred billion dollars. No administrator commanded that investment. The Act altered the probability distribution of expected returns across investment classes. Capital followed probability.
The failure cases are equally instructive. The EU Emissions Trading System, in its first two phases, produced carbon prices too low to alter investment decisions in power generation or heavy industry. The design failure was specific: caps were set at levels that did not genuinely constrain supply, and the resulting price signal was too weak and too uncertain to anchor long-term capital decisions. The instrument was structurally sound. The calibration was not. The instrument was right. The calibration was wrong. That distinction is everything.
Germany's Energiewende offers a third case study — one of sequencing failure. The simultaneous retirement of nuclear capacity and coal dependency, without sufficient grid infrastructure investment to manage intermittency, produced a structural configuration that entrenched gas bridging. The policy intention was decarbonisation. The probability distribution it created locked in a gas dependency that was not resolved before the 2022 energy crisis arrived. Good intentions operating inside a poorly calibrated probability distribution produce outcomes that no one designed.
The policy that alters the incentive architecture governing investment decisions across a sector is governing. The policy that issues a command but leaves the distributional logic of actual decisions unchanged is not.
The boardroom is where the determinism problem manifests most acutely — and most consequentially for capital allocation.
Corporate governance frameworks have persistently misclassified climate risk in binary categories structurally incompatible with the nature of the problem. The standard taxonomy asks whether a risk is "material" — quantifiable, near-term, sufficiently certain — or "immaterial" — speculative, long-term, beyond the governance horizon. Climate risk resists this taxonomy completely. It is a probability distribution that shifts continuously with policy trajectories, technological development, physical system dynamics, and the collective behaviour of other actors in interconnected markets.
The board that waits for certainty before acting on climate risk is not being prudent. It is committing a category error. Prudence under genuine Knightian uncertainty does not mean waiting for the distribution to collapse into a single outcome. It means acting on the best available probabilistic assessment while preserving optionality, avoiding irreversible commitments to high-risk pathways, and updating continuously as new information arrives.
BlackRock's trajectory over the past five years illustrates the institutional contradiction at the centre of this problem. As the world's largest asset manager, it deployed significant rhetorical capital around climate stewardship and system-level exposure between 2020 and 2022. Its 2020 letter to CEOs committed to putting sustainability at the centre of its investment approach. Yet its voting record in subsequent years showed declining support for climate-related shareholder proposals. The gap between stated commitment and voting behaviour is not hypocrisy in the ordinary sense. It reflects a structural impossibility: the framework alone cannot do what it is being asked to do. Governing probabilistic system-level exposure through disclosure instruments calibrated for deterministic materiality produces performance rather than governance.
The failure of the past two decades is the systematic substitution of disclosure for governance. Boards have been required to disclose climate risk. Fewer have been required — or have chosen — to alter their capital allocation decisions in light of it. Disclosure without consequential governance response is not climate governance. It is the performance of climate governance within a deterministic frame: one that treats the disclosure itself as the output, and the underlying risk as something that happens to the firm rather than something the firm's decisions help to produce.
The fiduciary dimension reaches the deepest level of institutional obligation.
Fiduciary duty requires the decision-maker to act in the best interests of the beneficiary, with prudence, loyalty, and care. In the context of long-term institutional investors, this has historically rested on three assumptions: a relatively short time horizon, a focus on near-term financial materiality, and an implicit presumption that aggregate structural risk is exogenous — something that affects the portfolio but is not itself affected by the portfolio's decisions. All three assumptions fail in the context of climate.
The time horizon problem is structural: pension funds with liabilities extending decades into the future are governed by boards whose incentive structures, reporting cycles, and career horizons are measured in years. The temporal misalignment is institutional, not individual — and it is addressable by requiring scenario analysis across the relevant time horizons of the liability structure, not the reporting structure.
The materiality problem is subtler and more consequential. Financial materiality was designed for a world in which aggregate structural risk was genuinely exogenous. Climate changes this completely. The collective investment decisions of large institutional investors are themselves a component of the range of possible climate outcomes. A fund that allocates capital to high-carbon assets is not merely exposed to transition risk. It is contributing, in small but non-zero ways, to the aggregate risk distribution that generates that exposure.
The fiduciary who ignores this is not being prudent about structural risk. They are misidentifying the boundary between their decisions and the system within which those decisions operate.
The governance implication is direct: fiduciary duty in a probabilistic climate system requires not merely assessment of risk to the portfolio but assessment of the portfolio's contribution to system-level risk probability. This is not a radical extension of fiduciary obligation. It is the application of its existing logic — prudent navigation under uncertainty, anchored in Frank Knight, Herbert Simon, and Elinor Ostrom — to a system where the traditional assumption of exogenous structural risk has broken down.
If climate outcomes are probabilistic rather than determined, the moral calculus of climate action changes fundamentally. The question is no longer whether any individual action is sufficient by itself to prevent a specified outcome. It never is. The relevant question is whether the action shifts the probability distribution of expected outcomes in a direction that reduces expected harm — and whether that shift, multiplied across the population of actors capable of similar choices, meaningfully alters the aggregate distribution.
This framing dissolves several of the most persistent ethical confusions in climate politics.
It dissolves the guilt-at-impotence problem: the paralysing recognition that no individual action is sufficient to solve a planetary-scale problem. Under a probabilistic ethics, sufficiency is not the standard. Contribution to the distribution is. No actor is required to be omnipotent. Every actor is required to shift probabilities in proportion to their capacity to do so.
It dissolves the free-rider problem as a philosophical objection to action. The collective action problem is real but it is a coordination problem, not a logic problem. The solution is institutional design — frameworks that alter the incentive structure of the coordination problem — rather than individual moral heroism.
It dissolves the "too late" fatalism most directly. If the future is a range of possible outcomes rather than a fixed one, then "too late" is never a binary condition. Every increment of probability-shifting action reduces expected harm. The magnitude of that reduction diminishes as physical thresholds are crossed, but it does not reach zero. The governance obligation to act does not expire when certainty of success is no longer available.
This is not optimism. It is something harder and more durable: the ethics of responsibility in a system where outcomes are neither guaranteed nor foreclosed.
One dimension of this problem resists purely institutional analysis. Probability structures are not only technical constructs. They are partly cultural ones. Public narratives shape the range of futures that actors consider accessible. When "peak oil demand by 2030" becomes the dominant framing of an industry, investment decisions shift before any policy instrument fires. When climate change is narrated as civilisational catastrophe without agency, institutional actors disengage from the probability-shifting work that governance requires. Behavioural economics has established that cognitive framing effects — loss aversion, present bias, identity-protective reasoning — alter decision-making in ways that probability calculations alone do not predict. Effective climate governance must therefore include the governance of narratives: not propaganda, but the deliberate construction of frameworks that preserve agency, communicate probability honestly, and resist the capture of public imagination by either fatalism or false certainty.
The probabilistic framework risks a serious error if it treats governance failure as primarily epistemological — as if the problem were that actors misunderstand the nature of complex systems, and that better framing would produce better outcomes. Some governance failure is cognitive. But much of it is not.
Some actors understand probabilities, understand risks, understand lock-in — and choose to preserve incumbent value structures intentionally. The distinction matters enormously for institutional design.
Sovereign fiscal dependence on hydrocarbon revenues creates governments whose rational interest is in slowing the probability shift toward decarbonisation, regardless of stated commitments. Short-duration executive compensation structures create boards whose rational interest is in maximising near-term returns even when long-term probability distributions are adverse. Regulatory capture creates agencies whose institutional relationships with incumbent industries produce systematic underweighting of transition risk — not through ignorance but through structural alignment.
The stranded asset problem illustrates this with particular clarity. Global fossil fuel assets at risk of stranded value under 1.5°C or even 2°C pathways are estimated in the tens of trillions of dollars. The actors who hold those assets have rational interests in altering the probability distribution of climate policy outcomes, not adapting to them. The result is not passive resistance but active probability management: lobbying against carbon pricing, funding uncertainty narratives, capturing standard-setting processes, structuring disclosure regimes that measure without constraining.
Electoral cycle distortion compounds this. Democratic systems whose accountability mechanisms operate on two-to-five-year horizons are structurally misaligned with the decadal probability distributions that climate governance must manage. The result is a systematic bias toward interventions that are visible, immediate, and attributable — subsidies, mandates, announcements — and away from the structural probability-shifting that operates through institutional design, infrastructure investment, and long-term incentive architecture.
Effective climate governance requires democratic systems to govern on timescales that their accountability mechanisms were not designed to manage. That is not an argument for bypassing democracy. It is an argument for democratic institutional innovation.
A probabilistic governance framework that ignores power is not merely incomplete. It is vulnerable to being instrumentalised by the actors whose interests lie in managing probability distributions toward delay. The governance challenge is therefore not only to design better probability-shifting instruments. It is to ensure that the design process itself is not captured by the actors with the greatest interest in calibrating those instruments toward ineffectiveness.
Three recent UK rulings test the relationship between legal reasoning and probabilistic governance directly — and the results are more nuanced than they first appear.
R (Finch) v Surrey County Council [2024] UKSC 20 — Purposive Determinism
In June 2024, a 3:2 majority of the UK Supreme Court held that Scope 3 greenhouse gas emissions resulting from the eventual combustion of oil produced from a proposed Surrey development should have been assessed in the Environmental Impact Assessment. Lord Leggatt's majority judgment rested on a single agreed fact: all crude oil extracted from the site would eventually be refined and burned. Downstream combustion was therefore an "inevitable consequence" of the project, falling within the EIA Directive's requirement to assess "direct and indirect effects."
The word "inevitable" is doing significant governance work. The majority is deploying what might be called purposive determinism: not a philosophical claim that the future is fixed, but a legal proxy that treats a near-certain statistical aggregate as equivalent to a direct causal consequence. Because the probability of downstream combustion approached certainty on the agreed facts, the law treated it as such. The minority — Lords Sales and Richards — argued the causal chain is broken by independent third-party decisions to burn the fuel. That dissent is, paradoxically, the more genuinely deterministic position: it atomises responsibility into individual third-party choices that can never be attributed to any single decision-maker.
The court's use of inevitability language reflects a legal system seeking administrable proxies for probabilistic reality. Law often operationalises high-probability aggregate outcomes as if they were deterministic because adjudication requires binary procedural conclusions — a case is won or lost, a permission is granted or refused, an effect is within scope or outside it. That binary architecture cannot accommodate infinite probabilistic nuance; it compresses probabilities into legal thresholds. The majority in Finch compressed near-certain aggregate combustion into an "inevitable consequence" because that is the form legal reasoning requires. The substance is probabilistic. The form is deterministic. And the majority reasoning is therefore more compatible with probabilistic governance than its language suggests. By requiring that near-certain aggregate downstream effects be assessed, it embeds expected-value reasoning into development law. Finch does not contradict probabilistic governance. It extends it into planning procedure.
R (Friends of the Earth, ClientEarth, Good Law Project) v Secretary of State [2022] EWHC 1841 and the Carbon Budget Delivery Plan Cases [2024] — Judicial Enforcement of Probabilistic Governance
The Net Zero Strategy litigation is the most directly supportive of this framework of any UK climate ruling to date. In July 2022, Mr Justice Holgate found the government's Net Zero Strategy unlawfully adopted because the Secretary of State had not been provided with the individual policy contributions to the Sixth Carbon Budget, nor with any explanation of how a 5% delivery shortfall would be addressed. What the court was demanding — in direct legal terms — was the probabilistic governance reasoning this framework prescribes: not a deterministic assertion that targets would be met, but a policy-by-policy quantification of expected delivery with explicit disclosure of delivery risk.
The revised Carbon Budget Delivery Plan was struck down in 2024 on the same grounds. The High Court found it unlawful for the Secretary of State to have adopted a plan when officials had "very low confidence" or "low confidence" in the delivery of approximately half the required emissions reductions, without disclosing that assessment. The legal standard under sections 13 and 14 of the Climate Change Act 2008 requires the government to adopt policies it considers "will enable" carbon budgets to be met — a standard the court interpreted as demanding honest, evidence-based probabilistic assessment of delivery risk, not deterministic assertion of success.
These rulings are not merely compatible with probabilistic governance. They are judicial enforcement of it. The courts are holding that a government cannot satisfy its statutory obligations by asserting a deterministic outcome without disclosing the risk distribution underlying that assertion. The Climate Change Act, as interpreted by the courts, is a probabilistic governance instrument in statutory form.
ClientEarth v Shell's Board [2023] EWHC 1137 (Ch) — The Limits of Judicial Reach
The limits of the judicial approach are equally instructive. ClientEarth's derivative action against Shell's directors — arguing that the board had breached sections 172 and 174 of the Companies Act by adopting an inadequate climate strategy — was struck out at the permission stage. Trower J found that ClientEarth had not demonstrated that the directors' decisions fell below the required standard, applying a margin of judgment to the board's strategic choices on climate risk management.
The reasoning reflects a genuine institutional boundary. Courts can enforce procedural obligations — assess Scope 3 emissions; quantify delivery risk; disclose uncertainty to Parliament. They cannot, as currently constituted, impose substantive liability on boards for contributing to a probability distribution of aggregate harm. The institutional architecture of legal causation — built around specific actors, specific decisions, specific outcomes — does not map cleanly onto the distributed, non-linear causation of climate systems.
Courts can insist that probability be honestly assessed and disclosed. They cannot yet assign liability for each actor's contribution to aggregate expected harm.
That institutional gap — between the legal obligation to disclose probability and the absence of liability for distributional choices — is precisely where fiduciary reform, institutional redesign, and capital allocation governance must operate. Climate litigation, at its most effective, does not resolve the governance problem. It creates the information environment and the procedural accountability within which probabilistic governance becomes enforceable. The rulings are powerful. They are not sufficient.
Three serious objections to probabilistic governance deserve acknowledgment.
The first is the technocracy risk. If governance is the art of shifting probability distributions through complex adaptive systems, does this not require technical expertise inaccessible to democratic deliberation? Probability frameworks can produce pseudo-precision — scenario analyses that give the appearance of rigour while embedding contestable assumptions at every structural choice point. The answer is not to abandon probabilistic governance but to demand transparency in the construction of probability distributions, democratic input into the scenarios that anchor them, and accountability for the assumptions embedded in institutional forecasting.
The second is the accountability objection. If outcomes are distributed across a complex system, the causal chain from decision to outcome becomes attenuated. This is a real risk. It is addressed by locating responsibility at the level of structural choices — the design of incentive frameworks, the calibration of instruments, the governance of institutions — where probability distributions are genuinely shaped. Responsibility attaches to the architecture of the distribution, not merely to individual positions within it. In short: diffuse causation does not dissolve responsibility. It relocates it.
The third is the dilution-of-individual-responsibility objection. If the ethics of climate action is probabilistic rather than personal, does this provide comfortable cover for inaction? The answer lies in the proportionality principle embedded in the framework: every actor is required to shift probabilities in proportion to their capacity to do so. For a pension fund managing hundreds of billions, that capacity is substantial. For a national government, it is enormous. Probabilistic ethics does not dissolve individual responsibility. It scales it.
The practical implications of this framework are concrete.
It requires that states abandon the linear causation model — impose rule, obtain outcome — in favour of designing interventions that alter the incentive architecture governing decisions across the relevant system. The UK's Contract for Difference auctions, which created long-term revenue certainty for offshore wind developers and produced one of the most dramatic cost reductions in energy history, are a model of probability-shifting design. The EU ETS in its early phases, which created a price signal too weak and too uncertain to alter capital decisions, is a model of distributional calibration failure.
It requires that boards adopt scenario analysis as a genuine governance tool — one that maps the range of outcomes under different physical and policy trajectories and integrates that range into capital allocation decisions in real time. The Taskforce on Climate-related Financial Disclosures framework was a necessary first step. Disclosure of scenarios without consequential governance response leaves the underlying incentive architecture of capital allocation unchanged.
It requires that fiduciaries expand their concept of prudence to encompass system-level risk contribution, not merely system-level risk exposure. The distinction is not philosophical. It is the difference between a passive allocation to a high-carbon sector and an active choice whose aggregate systemic effects the fiduciary has considered and accepted.
It requires, most fundamentally, that governance frameworks themselves be protected against capture by actors whose interests lie in probability management toward delay. The Carbon Budget litigation has established that courts will enforce this procedurally. The substantive obligation falls on institutions.
There is a particular kind of courage required to govern well under genuine uncertainty — not the courage of certainty, which is merely the confidence of those who have not looked closely enough, but the courage of acting responsibly in a system where outcomes are neither guaranteed nor foreclosed.
The climate transition sits precisely in this space. It is constrained by physical laws, capital structures, political economies, and path dependencies that cannot be wished away. Within those constraints, the space of possibility remains genuinely open. Multiple futures remain physically and politically accessible. The difference between them is not fate. It is governance.
Governance that mistakes constraints for determination will produce the outcome it predicts — not because the outcome was fixed, but because the prediction became self-fulfilling. Governance that understands the difference — that operates through probability rather than command, acknowledges the role of power in shaping distributions, and holds itself to the ethics of responsibility rather than the comfort of certainty — can do something determinism cannot.
It can change what is probable.
The courts, as ‘Finch’ and the Carbon Budget cases demonstrate, can enforce the procedural obligations of probabilistic governance: the obligation to assess, disclose, and honestly characterise the distribution of expected outcomes. That is not a small thing. It means that governments and developers can no longer assert deterministic success without evidential foundation. It means that Scope 3 emissions — the downstream consequences that every producer has preferred to externalise — are now a mandatory consideration in development decisions. The information environment for governance has changed.
But enforcing disclosure is not the same as governing. The gap between what courts can mandate and what governance requires is where institutional design, capital allocation reform, fiduciary reorientation, and democratic innovation must operate. The law has moved. Governance must now move with it.
Climate governance fails when it mistakes probabilistic systems for deterministic ones. The transition is neither mechanically inevitable nor infinitely controllable. It is a structured field of evolving probabilities shaped by institutions, incentives, infrastructure, power — and, increasingly, by the courts. Governing it well begins with understanding what it is.
illuminem Voices is a democratic space presenting the thoughts of leading Sustainability & Energy writers, their opinions do not necessarily represent those of illuminem.
Curious how major companies measure up on climate? On illuminem’s Data Hub™, explore verified emissions data, net‑zero targets, and sustainability performance of thousands of firms — from industry leaders to emerging innovators.
illuminem briefings
Climate Change · Effects
Diego Balverde
Energy Management & Efficiency · Energy
Praveen Gupta
Insurance · Adaptation
The Economist
Climate Change · Effects
The Guardian
Climate Change · Food
World Economic Forum
Climate Change · Environmental Sustainability