Many thanks to Serge Galam (Ecole Polytechnique, CREA Centre de Recherche en Epistémologie Appliquée) for his invitation. Here is the abstract of my talk.
Title : When bees are dying, words are counted.
Modelling translation of scientific knowledge in public debate
This talk throws light on translation dynamics connected to the explanation of the unusual death of bees in France through the lexical evolution in the media. After collecting all papers (more than 1000 papers in French) dealing with this ecological and disturbing fact, a statistical analysis of textual data is performed on the lexicon used by journalists to describe, report associated information, and judge the main causes of the bees' death.
Three main results have been found.
First, information is modified throw each iteration i.e., each time a paper is published on this subject. The main subject of the public debate slides from analytic and scientific controversy to a synthetic and moral issue. The vocabulary evolves from doubt to certitude with major negative words.
Second, at start a neutral description is obtained which shows a complex phenomenon. But it is gradually shifted towards selecting one single cause to the bees death. During this change the density of words per paper decreases drastically after some critical value.
Third, after the precautionary principle was decided (against pesticides), new scientific studies have appeared showing that multi factors could explain bees' death. Then, the judgment shifts gradually from one factor (the pesticides) to different causes. Therefore the debate is balancing between multi and unifactorial explanation of the phenomenon.
As a conclusion, the data are analyzed using the Galam sequential probabilistic model of opinion dynamics. The robustness of the analysis is discussed in particular its dependence with respect to the words chosen for the analysis. Extension to other problems is considered.
keywords: sociophysic, lexical and statistical analysis, media, ecology