y The objective of this work-in-progress is to investigate the potentialities but also the limitations of traditional risks analysis tools especially in the context of emerging technologies and develop a method facilitating the early detection of scenarios of accidents. This is certainly a challenge particularly for new industrial fields since, in this case, very little or no lesson from past accidents is available. It is believed that such situations cannot be conveniently treated using traditional risk assessment methods (HAZOP, FMEA,...) and typical examples are given. The reason is that those methods rely heavily on past accidents and are therefore "trapped" in them so that they are largely "inductive". In terms of foreseeing the future, the shortcomings of inductive methods are recalled. The possibility to imagine the future with very little clues is then discussed on the ground of theoretical consideration and a way to do so is proposed (abduction, serendipity). Then on the basis of the observation of how the experts work and how discoveries are made, a potential new methodology is outlined. (C) 2016 Elsevier Ltd. All rights reserved.