Reçu le 28 août 1989; accepté le 2 mai 1990) Résumé — L'aide à la décision technique ne consiste pas seulement à informer le décideur, à diagnostiquer l'état des cultures ou prévoir leur évolution, mais aussi à accompagner l'agriculteur dans sa propre démarche d'intégration de ces différents éléments. Un système informatique d'aide à la décision doit alors contenir un modèle des raisonnements mis en œuvre par l'agriculteur. À partir d'observations par entretiens de quelques processus de

more »

... , on relève quelques caractéristiques structurelles communes à ces raisonnements. On en conclut qu'il est possible de modéliser les processus de décision. De tels modèles sont utilisables dans un but cognitif (pour représenter ce qu'on connaît d'un processus de décision), prévisionnel (pour étudier les propriétés de tel ou tel processus de décision), ou décisionnel (comme module d'un système informatique). L'apport potentiel de l'intelligence artificielle en la matière est souligné. agriculture / aide à la décision / modélisation / raisonnement / système informatique Summary — Models of agricultural reasoning and consequences on the design of decision support systems. Agricultural technical decision support should be used not only to inform the decision-maker, diagnose the state of the crops, or forecast their evolution, but also to accompany the farmer in his own process of integration of these elements. A computer-based decision support system should therefore contain a model of the reasoning carried out by the farmer. The present work attempts to describe and analyse this reasoning. It is based on a series of interviews with 18 farmers (the characteristics of the corresponding cropping systems are in table I, the interview framework in table II). The first section of the paper reports on the 2 types of behavior observed when the farmer is confronted with a technical problem (ie damaged soil structure): he either tries to find out and eliminate the cause of the emergence of the problem or only treats its observed consequences (fig 1) . Following this, some common structural characteristics of the reasoning are outlined, leading to the feasibility of modelling decision-making processes. Examples of such models are analysed in 2 particular areas: weeds infestation (fig 2) and the annual crop-plot assignment (fig 3) . Some important consequences of the above-mentioned analysis on the designing of decision support systems (DSS) are the necessity for a DSS: to help the farmer in jointly using information processing, diagnosis and solution design facilities; to take into account the temporal and spatial dimensions of the considered decision-making process. In addition, we stress the need for specific research on decision-making processes modeling: 3 classes of models are now considered (fig 4) . Decision-making process models may serve 3 purposes: cognitive (to represent what we know about a given decision making process), predictive (to investigate the properties of such and such decision process) and decisive (as part of a DSS). The potential contribution of artificial intelligence in this area is outlined. agriculture / decision support / modeling / reasoning / computer system