The (Missed) Poetry of Climate Models

The Poetry of Science project aims to restore imagination as a legitimate epistemic dimension of mathematical modelling, especially of integrated assessment models (IAMs) for climate change^. Why IAMs? Because they are among the central technical infrastructures through which global climate governance makes possible futures visible and translates climate research into policy pathways for mitigation and adaptation. They do so by bringing together representations of energy systems, economies, land use, technologies, emissions and climate to explore how these systems may evolve together.

The policy inertia of global climate action made me wonder whether there was something inside these models that prevented them from serving the purpose for which they were created. Perhaps some would contend that models are the wrong target when it comes to discuss policy failures in climate action. In fact, I myself don’t blame the lack of climate action on models, nor would I ever say that the problem is that these models are not good enough. There is extensive sociological and political literature showing that the policy architecture of climate governance is in many ways dysfunctional while producing weak forms of international cooperation.

So why waste time looking inside models? The answer is that models participate in sustaining the current climate governance architecture. Their scenarios of emissions, socio-economic impacts, vulnerabilities, and mitigation or adaptation options, are coordinated through the work of the Intergovernmental Panel on Climate Change (IPCC), which forms the central scientific pillar of the United Nations Framework Convention on Climate Change (UNFCCC). This coordination has enabled an extraordinary global research community to produce important assessments and warnings. However, it has also normalised the idea that the boundary between what science can say and what policy should do is unmovable and should not be trespassed. The broader principle is that climate science should be policy-relevant but not policy-prescriptive: it should inform political judgment without replacing it. While this boundary is essential to refrain from technocratic decision-making, it has also generated a deep misunderstanding within the scientific community that their research should be value-free, emptied of subjective values while remaining factual and rigorous.

Decades of sociological and philosophical scholarship have demonstrated that science cannot be value-free. Epistemic, social and ethical values enter the framing of research questions, the selection of evidence, the treatment of uncertainty and the evaluation of consequences. This is all the more evident when research, based on modelling, is saturated with uncertainty and deals with futures that do not yet exist. A deep friction therefore exists between the epistemic warning for scientists to stay neutral and the modelling tools underpinning their work. Models require researchers to constantly make judgments in response to uncertainty. They need to decide what should be represented, which relationships should be formalised, which phenomena should remain outside the system and how incomplete knowledge should be treated. These judgments respond in part to practical computational needs and epistemic values such as accuracy, coherence, etc. They also respond to ethical and social values, which play an important role in determining whose vulnerability matters, how much weight should be given to future generations, how risks and benefits should be distributed, which forms of harm should be avoided, and what counts as a fair or acceptable transition.

Thus, my question: what would happen if climate modellers considered the role of values in their modelling practice? What consequences would this have for climate governance? Some modellers are already very conscious of the role that values play in their work and are seeking ways to solve this situation, for example by exploring ways to better communicate uncertainties and values to the users of their models. Nonetheless, when those users are policymakers, the good communication solution may fall short of describing the context at hand. While communication is possibly the most visible and tractable dimension in the relation between researchers and policymakers, it also tends to distract from the fact that policymaking does not only involve receiving and communicating knowledge, but also representing problems, allocating responsibility, distributing costs and benefits, designing options and deciding what kinds of futures are acceptable.

Modellers go through a comparable process when they conceptualise, frame, structure, implement and interpret models of coupled human and natural systems, which is why I see no reason why such invisible policy process should remain implicit rather than openly discussed. The scenarios produced by models are not automated outputs or neutral representations of climate and human futures. They are artefacts of future-making. They organise assumptions, evidence, values and imagination into an intelligible account of what might happen. This is a precious feature of model-based scenarios that is currently overlooked—if not wasted—in climate governance and that prevent scenarios to become really transformative of climate action.

With my research, I propose to explore the use of deliberative settings based on ‘mini-publics’ to activate this feature in the development of emission scenarios and the assessment of climate mitigation and adaptation policies.

Three Routes into the Invisible Deliberative Dimension of IAMs

Three approaches have progressively intersected in my research into the deliberative dimension of IAMs: sociology of science, philosophy of science, and literary theory, each supplementing the limits of the others.

Sociology of science has long examined how science and policy shape each other at specific sites of norm and decision-making and has pioneered the concept of co-production to describe the mutual constitution of knowledge and social order.

Philosophers of science have, for their part, examined the relation between normative and epistemic values. Their interaction is evident in many scientific and technological applications, especially where research is intended to guide action under conditions of uncertainty. What remains contested is when, how and under what procedural conditions social and ethical values may legitimately contribute to the validation, credibility and use of knowledge. In relation to complexity, some have called for a new epistemological sensibility supporting a more socially distributed and reflexive process of knowledge validation. They also have taken seriously the metaphorical characteristic of models, which help us speak about realities that are not yet present, are not directly visible or cannot be fully grasped. This attention to complexity and metaphor inspired me to get serious about the narrative nature of emission scenarios and climate policies by looking at literary theory.

From Framing to Storytelling

Narratives and stories are part of the IPCC’s official vocabulary for guiding (qualitatively and quantitatively) the development and exploration of possible interactions between the natural and socio-economic systems. Their role is usually understood in terms of framing capacity: setting the boundaries of what is to be modelled and what is not. These boundaries operate as grand narratives guiding IAMs towards plausible stories of technological progress, economic development and policy intervention. This understanding of narrative as a framing device contrasts with the critical and literary understanding of narrative as performative. To illustrate the difference, I draw from the work of the literary scholar and historical theorist Hayden White, who studied nineteenth-century historiography to show that narrative forms helped historians select, order and employ events, thereby shaping how the past became intelligible. Far from being neutral historical containers, narrative participated in the production of historical meaning.

There are at least two reasons why White’s take on narrative matters for future scenarios. First, historical events and futuristic events are not fundamentally dissimilar as they both concern a time that is “unpresent”, i.e. no longer present in the case of history, and not yet present in the case of the future. Both are also pursued for reasons of collective learning: through memory in history making and anticipation in future making. Second, complexity tells us that complex events are ‘historical’ events because while they are not pre-dictable, they are still retro-dictable, meaning that retrospectively they can always show a causal pathway.

What does it then mean to endorse Hayden White’s literary approach to history for rethinking emission scenarios? It means using their narrative form not just for bringing coherence to a certain projection, but to retrace the steps through which such projection appears as true or inevitable. It means giving imagination a useful role to identify how the passage from coherence to inevitability occurs, what it says about the conceptual and contextual setting in which the storyteller/modeller operates, and how this passage activate realities that would otherwise stay untold. It also means understanding imagination beyond its cognitive ability to construct possible futures, but also as an expression of the storyteller/modeller’s consciousness of the present they inhabit. Hayden White invites us to pay attention to the latent content of historical discourse, which lies in the norms, ideas, forms of authorities and political arrangements that a historian draws from the present when assembling an account of the past.

I take this as an invitation to look at the latent content of future-oriented climate discourse as they emerge through climate change research lines, models, and scenarios, and to pay attention to the necessary interrogation of the present. What present is being normalised when a scenario imagines a particular future? Whose present is being treated as the starting point? Which forms of life, political economy and social organisation are being carried forward as though they were simply the world as it is? The present is a contested field of scientific and political representation, not just because we all experience very different presents, but also because we are made to imagine very different futures that affect our present. We tend to think that the present is the starting point for imagining the future, but in fact there is much to reveal about whose present we are favouring, normalising and justifying when we enquire into the imaginations feeding into specific representations of the future. Contrary to what we generally read today, it is not true that we need to activate our imagination for a better world. Our imagination is already active; perhaps it is toxic; for sure, it is unevenly distributed. We should therefore pay attention to imagination as an active dimension of the epistemic infrastructure of climate governance, including IAMs.

 The Alternative Stories We Leave Out

Within the climate science community, this uneven distribution of imagination is receiving increasing attention, even if it is not always described in those terms. The scenario community has progressively recognised that some social values have been taken for granted or dismissed in the development and assessment of baseline, or reference, scenarios; these are scenarios used as counterfactuals against which mitigation interventions are compared. Among the values receiving increasing attention is social justice and equity. Critiques of modelling brings together concerns about the normative character of models, their treatment of uncertainty and the assumptions through which they define the system under study. For instance, many conventional global IAMs incorporate assumptions about continuing economic growth that set their storylines in the direction of cost-efficiency criteria and technologically mediated solutions for achieving an emissions or temperature target. When scenarios are explored and assessed through these assumptions, carbon dioxide removal often appears as a way to confidently delay emissions cuts while remaining compatible with ambitious climate targets. Similarly, solutions based on technological substitution and supply-side change can become more visible than pathways that stop fossil-fuel expansion, protect land for people and ecosystems, or treat carbon removal as limited by ecological and social conditions. Sufficiency and reduced demand can also become less central to addressing climate change compared to maintaining economic growth and minimising aggregate mitigation costs.

It is in this sense that alternative stories, values and ideas can become easily excluded or marginalised: stories that reject techno-optimism and distrust carbon removal techniques; stories that reject fossil-fuel extraction and prioritise land protection; stories that imagine sufficiency, redistribution or democratic control as central components of climate action. These omissions may occur through different modelling choices, be them over specific parameters,   the form of the objective function, or over the structure of the model.

These choices are not simply right or wrong. The modeller will never obtain a perfect representation a coupled human-natural system, nor a perfect projection of its future. Uncertainty is both a condition of knowledge and a condition of modelling practice. This is why the rightness or wrongness of their choices cannot be settled only technically. It needs to be deliberated. 

Models have a conceptual and argumentative function — not only a predictive one. They can help reconceptualise mechanisms and generate understanding even when they are not accurate representations in every respect. Storytelling can support this function by creating a shared language for examining how a model ‘reasons’ about the future through the layered chain of conceptual choices, formal assumptions, data, code, numerical procedures, and experimental framing. Storytelling will not solve the opacity of relevant modelling steps hidden in code, data transformations or undocumented settings, nor will it solve the opacity related to the many interacting components in a model and they casual paths. Storytelling, instead, could prevent these forms of opacity from disrupting broader discussions about climate and human futures.

Bringing Deliberation into Modelling

My research aims to explore the conditions for building and testing deliberative spaces around IAM scenarios. This is not a matter of just constructing more inclusive foresight practices, but a matter of embracing the state of uncertainty about the future as it is: dependent on the idea that we do not know many things, and that there is human choice. If the global future is in the realm of climate governance, then its uncertain state must be taken seriously. This means stopping the externalisation of human choice as an afterthought of expert knowledge and building human choice in the very construction of action-oriented knowledge. There are different deliberative formats that encourage human choice (e.g. citizen assemblies, citizen panels, citizen juries, citizen initiative reviews, to name a few). These are based on the idea that, under facilitated and safe conditions, ‘mini-publics’ composed of ordinary citizens and reflecting the diversity of a target population can develop forms of collective intelligence for addressing complex questions. In such deliberative settings, the climate community would experience a different relation with the policy dimension of its work. Modellers and citizens could scrutinise together what kind of imagination is at play, with its implications and meaning, and deliberate on the epistemic and political consequences of the process — for instance, whether specific modelling outputs could be held politically accountable.  

To the best of my knowledge, the idea of applying deliberative processes to IAMs scenarios has never been explored before, nor the idea of using storytelling as a central methodology in a deliberative process. Inspirational grounds come from both the system modelling and the deliberative democracy communities.

Participatory modelling already involves stakeholders in defining systems and exploring options, while participatory computational modelling applies similar principles to computational models and decision-support processes. My proposal is to bring their important foundations in conversation with deliberative ‘mini-publics’, which add a more structured process of learning, discussion and collective judgment.

On the side of deliberative democracy innovations, there is increasing interest in applying deliberative formats to issues that do not necessarily pertain to public policy issues. The relative success of citizen assemblies around the world over complex policy issues, from abortion to climate change or euthanasia, has reinvigorated the systemic turn in deliberative democracy that see deliberation expanding beyond the official sites of policymaking.

Imagination as Collective Epistemic Action

The direction of building deliberative settings for scenario modelling based on storytelling is not to replace evidence, modelling or technical expertise with imagination. It is to emphasise that climate governance already depends on imagination, whether it acknowledges this or not. Every scenario imagines a future. Every model embodies a view of what can change, what will remain stable and which relationships matter.

The issue is therefore not whether imagination should enter climate modelling. It is whether we are prepared to examine the imaginations already embedded there and make room for others.

Storytelling offers one way of doing this. It connects the modelled future to histories of inequality, present experiences of vulnerability and visions of a more just world. It gives participants a common language for exploring the consequences of different assumptions without reducing disagreement to a technical error.

The climate crisis demands more than better predictions. It demands distributed capacities to imagine, deliberate and act under conditions of uncertainty. IAMs can contribute to those capacities if they are treated not as machines for delivering policy answers, but as sites where possible futures and current presents can be explored, and their underlying commitments publicly examined. What I advocate for is an activist epistemology of climate modelling: to turn uncertainty from a barrier to knowledge into a condition for shared inquiry, to turn imagination from an invisible background operation into a democratic resource for climate action.