3 Values in Climate Science

Pulkkinen

Values play a role in the construction of climate change information.

Science has its own values, including openness, objectivity and evidence-based thinking. However social values — fundamental views on what is good, right and important — guide a number of decisions in the construction, assessment and communication of information.

This marks a departure from the traditional ‘value-free ideal’ of science, according to which social values should have a limited role in scientific research, while values that are epistemic (for example, precision and accuracy) are seen as legitimately influencing research.

Developing an acute awareness of how methodological choices and broader aims advantage different interests forms the first step in effectively managing the influence of values.

There is no neutral way of framing information.

Scientific research cannot be value-free, and climate science is no exception.

The value of values in climate science

To date, values are not widely acknowledged or discussed within physical climate science. Yet, effective management of values in physical climate science is required for the benefit of both science and society.

Values in multimodel-based assessments

A great number of research questions in climate science are answered by combining results from global climate model simulations within a multimodel framework and/or by their integration with observations. Winsberg20 argues that an opaque, inscrutable tapestry of values lies behind such results, due to the models’ size and complexity, distributed epistemic agency and generative entrenchment of methodological choices. Any multimodel-based assessment must moreover deal with the questions of which models to include, and how to combine them. The extremes range from including all available models, for example, in a Coupled Model Intercomparison Project context, and applying a one-model-one-vote principle, to selecting a single or very few flagship models. The underlying question of what is a good (enough) model is made explicit in model selection and implicit in model weighting, and relies on value-laden choices of metrics that may favour one spatial scale or region over another, one process over another or one stakeholder interest over another. This applies also to the AR6 approach of using a constrained ensemble of emulators for future projections, where the constraints are chosen to be based on simulation of past warming, equilibrium climate sensitivity and transient climate response.

Values in event attribution

Event attribution in its broadest sense is the evaluation of the contribution of causal factors to observed events. Two different methodological approaches to event attribution in climate science have been at times fiercely debated: the so-called probabilistic approach and the storylines approach, which occupy different positions on a spectrum of what level of conditioning on the meteorological circumstances is appropriate. A focus of debate has been the treatment of uncertainty in the dynamic response to anthropogenic forcing, given that uncertainty in the thermodynamic response is generally much lower. It has been argued that the two sides fundamentally disagree about risk preferences. The proponents of the storylines approach are more concerned with false negatives (that is, falsely rejecting or underestimating anthropogenic influence on an event), and their methodology is supposedly less prone to this type of error, while it is the opposite for the probabilistic approach and its proponents. Both risk preferences, and hence preference for either methodology, are argued by Winsberg et al. to be motivated by values, in particular by the balance between valuing epistemic confidence and informativeness.

Values in climate services

Climate services provide climate information to assist decision-making, aiming to support adaptation, mitigation and risk management decisions. This can be influenced by the values of all parties involved. Maximizing the fit of the information provided to the needs of the service users includes, in particular, the consideration of the users’ value system. Parker and Lusk argue that a significant and feasible component is to match the risk preferences of the analysis to those of the users. This can be done by learning which types of errors the users find particularly undesirable; recognizing methodological choices that differ in the risk of these errors; and making those choices in consultation with the users. For on-demand climate services, the authors suggest the use of clear warnings about product limitations and uncertainties in anticipation of various risk preferences, which allow for user customization at the point of service. Otherwise, they propose the prioritization of those user groups that might suffer especially severe harms and have limited access to climate information, and call for clear communication of which choices are influenced by values and how.

Pulkinen (2022) The value of values in climate science (pdf sharedit)