When and why design choices can be modelled as uncertain factors in prospective LCA

In prior work I have argued that open design choices in prospective LCA, such as material choices for a novel solar cell, can be modelled as uncertain factors, with a probability that further research can update. Over the years I have often been asked (and challenged on) whether such choices constitute a form of uncertainty that can be propagated and subjected to global sensitivity analysis. In this commentary I give the answer I have settled on, and the decision theory and philosophy behind it: it depends on the question and the moment. When the choice is being made now, it is a decision, best informed by a separate LCA for each option, as decision analysis holds (Morgan and Henrion; Gelman et al.). For any other question asked while the choice is still open, it is an uncertainty for everyone, developers included. The choice belongs to a later stage of a multistage decision, after tests whose results nobody can predict (Popper). Even Levi, for whom deliberation crowds out prediction, allows probabilities for one’s own acts that are not being deliberated now. I also discuss what a probability of adoption cannot capture, from lock-in among competing research groups (Arthur) to options nobody has considered yet. Finally, I explain why probabilities remain the practical way to screen many compounding design choices at once, so that only the few that matter need their own LCAs, in line with Stirling’s call for plural and conditional advice.