Citation

Gaymard S, Tiplica T. Int. J. Inj. Control Safe. Promot. 2015; 22(4): 340-351.

Affiliation

Laboratoire de Psychologie des Pays de la Loire (LPPL) , LUNAM Université, University of Angers , UPRES EA 4638, Maison des Sciences Humaines, 5 bis boulevard Lavoisier 49045 Angers cedex 01, France.

Copyright

(Copyright © 2015, Informa - Taylor and Francis Group)

DOI

10.1080/17457300.2014.909500

PMID

24787234

Abstract

The conditional script questionnaire (CSQ) makes possible to study the conditions under which drivers find it legitimate to transgress the Highway Code. In this paper, we propose to analyse conditional respect towards the pedestrian with a new methodology based on Bayesian networks (BN). This methodology is designed to give a useful decision support tool for the analyst. Starting from data encoded in the CSQ, we use structure learning algorithms in order to build a BN. Then, we exploit it for two purposes: to extract new knowledge about the main topics expressed in the CSQ and to make inferences. This methodology helps to better understand the behaviour of drivers interacting with pedestrians and what might be the cause of their decisions of legitimate transgressions. The efficiency of the methodology proposed here is illustrated and a context-dependent 'mapping' of the legitimate transgressions established.

Language: en