Episode Summary
Executive Summary: Erica Thompson argues that mathematical models are useful but intrinsically limited tools: they simplify reality, embed values, and require human judgment to interpret. In climate policy especially, model outputs can create false certainty, obscure ethical trade-offs, and shape the futures they claim merely to predict. The conversation calls for epistemic humility, more diverse models and modelers, and explicit discussion of values alongside science.
Main Topics: What models are and why they matter (Priority: 5/5): Thompson defines models broadly—from climate simulations and spreadsheets to maps, photos, and even fiction—as simplified representations used to understand and act in the world. Uncertainty, risk, and judgment (Priority: 5/5): The discussion distinguishes uncertainty inside model land from the harder problem of translating model output into real-world decisions, where expert judgment remains unavoidable. Adequacy for purpose (Priority: 5/5): A model cannot be judged in the abstract; it is only good or bad relative to the specific decision or use case it is meant to inform. Performativity and conviction narratives (Priority: 4/5): Models do not just describe reality; they influence behavior, policy, and expectations, sometimes becoming self-fulfilling or counterproductive through feedback effects. Diversity of models and modelers (Priority: 5/5): Rather than converging on one 'best' model, Thompson argues for a wider range of plausible models and more diverse perspectives to expose blind spots and challenge groupthink. Climate economics, values, and false objectivity (Priority: 5/5): Integrated assessment models and damage functions can smuggle in major ethical assumptions—such as assigning dollar values to vastly unequal harms—while presenting them as objective. Geoengineering, futures, and the need for explicit values (Priority: 4/5): The conversation warns that model structures can normalize high-risk technological fixes like stratospheric aerosol injection unless policy debates explicitly confront values and desired futures.
Key Arguments: Models are not neutral mirrors of reality; they are selective representations shaped by assumptions, data choices, and human judgment. There is a crucial difference between uncertainty within a model and uncertainty about whether the model is appropriate for the real-world task. Past model success does not guarantee future validity, especially when systems change or models are used in extrapolatory contexts. The only meaningful standard for a model is adequacy for purpose, which varies by user, decision, and political context. Performativity means models influence the systems they model, so forecasting is never purely observational. A single consensus model is less informative than a diverse ensemble of plausible models that tests the boundaries of possibility. Climate and economic models often embed major value judgments—about growth, equity, species loss, and acceptable risk—while pretending to be purely technical. Scientific output cannot by itself determine what society ought to do; values must be made explicit rather than hidden behind “follow the science.” Diversity of modelers matters because people with different backgrounds and experiences will notice different assumptions, blind spots, and priorities. Fiction, narrative, and other nontraditional forms of scenario-building can sometimes be as useful as mathematical models for thinking about futures and consequences.
Data Points: Year of original Grist essay about discount rates: 2005 - Roberts references a long-ago piece that framed discount rates as an ethical dispute hidden inside science. Time since that essay: about 15-18 years - The conversation notes the essay dates back to Roberts's early writing career. North Atlantic storm-model review period: 10–15 years ago - Thompson describes her PhD-era literature review on climate impacts on North Atlantic storms. Number of test flights before Challenger failure: several test flights - Used to illustrate how the same data can support opposite judgments about model validity and risk. Economic growth assumption in climate-economics models: 2% per year forever - Roberts criticizes integrated assessment models for assuming steady global growth as a baseline. Climate damage estimate mentioned for Africa: about 80% GDP reduction - Thompson discusses papers where climate change is modeled as causing catastrophic regional GDP losses with limited global feedback. Mitigation target referenced: below 2 degrees global mean temperature - Used to discuss the framing that makes geoengineering appear increasingly inevitable in models. Threshold example for carbon capture: $2,000 per tonne CO2 - At this assumed cost, carbon capture would not be used in model pathways. Threshold example for carbon capture: $2 per tonne CO2 - At this assumed cost, carbon capture becomes a dominant solution in model pathways. Climate policy horizon discussed: 2100 - Used in the discussion of desired futures, geoengineering, and long-term policy visioning.
Pivotal Quotes: "we know nothing for certain, but we don’t know nothing" — Erica Thompson: A concise expression of epistemic humility and confidence-with-limits in model-based knowledge. "If the target of climate policy remains couched in the terms of global average temperature, then stratospheric aerosol geoengineering seems to me now to be an almost unavoidable consequence" — Erica Thompson: A warning that narrow climate targets can implicitly steer policy toward geoengineering. "science does not tell you what to do" — David Roberts: Roberts summarizes a core theme: science can inform choices, but it cannot supply values or decisions.
Implications: Listeners should treat models as decision aids, not truth machines. Better climate policy requires more diverse modeling, explicit values, and greater humility about uncertainty—especially to avoid locking in technocratic or ethically distorted futures.