Modeling human decision making associated with discrete and continuous tasks

P.A. Wieringa¹ ,M.E. Vriezen¹ and F. Vanderhaegen²

¹Man-Machine Systems Group, Delft University of Technology, Delft, The Netherlands
²Laboratoire d’Automatique, de Mécaniques et d’Informatique Industrielles et Humaines, University of Valenciennes, Valenciennes, France

The goal of this research is to develop a format for obtaining data to related to the human decision making. The operators perform a monitoring task of a complex system which includes discrete and continuous control tasks.

Method & Results

The basic idea is that the decision making is mainly based on personal weighting of the costs and benefits of the execution (model developed at the University of Valenciennes). The model consists of two main parts: the criteria definition and the decision making. The first part de.nes the increase of certain criteria (e.g. safety, productivity and workload) as a function of a situation S and a certain action A which can be performed in that situation. The second part of the model calculates a so-called preference for each possible action (Act*), as function of the situation and based on the differences between the criteria (i.e. .Safety(S,A)). These preferences are compared with the actual performed action (Act).

Data to fill out the model are collected during experiments with 9 participants. Data are recorded while a subject controls the a simulation program called NewTranspall. Data are collected in four 15-minutes sessions (two sessions on a complex system and two sessions on a less complex system). The first dataset on each system is used to .t the decision making of the model.

The subject based data are collected during a verbal protocol in which a human performs a rating scale questionnaire on the different criteria. A rules set is formed and validated which is the base of the criteria definition.

Conclusion

Preliminary conclusions are that

• The set of rules is probably sufficient, though not optimal, accurate to serve as input for the human decision making.
• The exponential trade-off has better performance modeling the decision making, than the fuzzy model using 1 antecedent, 1 consequent rules. This could be due to the fact that fuzzy model does not explicitly introduce trade-offs between the different criteria.

Paper presented at Measuring Behavior 2005 , 5^th International Conference on Methods and Techniques in Behavioral Research, 30 August - 2 September 2005, Wageningen, The Netherlands.

© 2005 Noldus Information Technology bv