Energy models compute, i.e., optimize or simulate results based on data and assumptions. If model assumptions are uncertain, so are model results. The dissertation investigates whether a quantitative uncertainty assessment is desirable, possible and meaningful. And consequently, if – and if yes, which – legitimate inferences or recommendations can or should be based on uncertain model results. The project is therefore extended to epistemological questions and decision theory: how can one justify beliefs in form of assumptions for an energy model? Which kinds of “knowledge” are assumption-based energy model results? Should a rational agent base (political) decisions on uncertain energy scenarios? Especially if model results are used for policy advice and decision support an analysis of uncertainties seems relevant.
In economics a distinction between risk and uncertainty is popular. Knightian uncertainty refers to immeasurable uncertainty, in contrast to risk which can in principle be measured, though may be unknown. Classical uncertainty quantification (UC) techniques as used in natural sciences and engineering distinguish between epistemic and aleatoric uncertainty. UQ aims at a quantification of aleatoric uncertainty by different mathematical approaches and a reformulation of epistemic uncertainty in computable aleatoric terms (variability). In the dissertation project the uncertainty quantification is based on statistical methods. First results suggest that at least some assumptions in energy models are highly uncertain. Beliefs and consequently assumptions reflecting economic, social, technical, and energy political considerations did not render energy models with predictive power in the past and statistical analyses indicate highly complex or unstable relations in the target system. Qualitative approaches of uncertainty assessment or evaluation use scales or classes. Classes can be binary, thus uncertain or not, or gradual, for example significant, of medium impact, insignificant, or hybrid forms of uncertainty assessment.
The project investigates which epistemic approaches can depict the belief update or change processes based on uncertain energy model results. Concepts as for example the Stability Theory of Belief, probabilistic dynamic belief revision, or revision by comparison are considered in the context of uncertain energy model results. Preliminary results suggest that both qualitative and quantitative approaches may be suitable to depict the process – dependent on the uncertainty analysis. A probability-based explication of uncertainty is developed in this part of the project. The question if energy model results can – and if yes, under which conditions – lead to belief change for a rational agent is addressed.
In the project normative decision theories are considered, for example utility theory, or decision principles such as maximin, minimax, for decisions based on belief generated by uncertain energy model results. The aim is the "application" of the previous considerations. A decision is considered as expression or result of (1) the recognition (perhaps even quantification) of uncertainty present and (2) a conscious reflection of uncertainty integration in belief change or belief updating.
The dissertation project is funded by the Helmholtz Research School on Energy Scenarios.
|Supervisor:||Jun.-Prof. Dr. habil. Gregor Betz|
|Advisor:||Prof. Dr. Armin Grunwald|
|Related projects:||Shared Research Group "Limits and Objectivity of Scientific Foreknowledge: The Case of Energy Outlooks" (LOBSTER)|
|Doctoral students at ITAS||see Doctoral studies at ITAS|
Karlsruhe Institute of Technology (KIT)
Institute for Technology Assessment and Systems Analysis (ITAS)
P.O. Box 3640
Tel.: +49 721 608-22501 (ITAS-Sekretariat)