Tasks Decomposition of System Models for Human-Machine Interaction Analysis

Maudoux, Guillaume;Combéfis, Sébastien;Pecheur, Charles
(2015) Workshop on Formal Methods in Human Computer Interaction (FoMHCI), in conjunction with ACM SIGCHI Symposium on Enginnering Interactive Computing Systems (EICS) 2015 — Location: Duisburg, Germany (23.June.2015)

Files

FoMHCI2015_maudoux.pdf
  • Open Access
  • Adobe PDF
  • 174.92 KB
Download
FoMHCI2015proceedings.pdf
  • Open Access
  • Adobe PDF
  • 4.01 MB
Download

Details

Authors
  • Maudoux, Guillaumeorcid-logoUCLouvain
    Author
  • Combéfis, SébastienUCLouvain
    Author
  • Pecheur, Charlesorcid-logoUCLouvain
    Author
Abstract
This paper is concerned with the problem of learning how to interact safely with complex automated systems. With large systems, human-machine interaction errors like automation surprises are more likely to happen. Previous works have introduced the notion of full-control mental models for operators. These are formal system abstractions embedding the required information to control a system completely and without surprises. Full-control mental models can be used as training material but are ineffective as their control over a system is only guaranteed when fully learned. Operators that have learnt all the possible behaviours of a system have built a full-control mental model. Such models have been defined in [4] and techniques to build minimal ones have been described in [3] and [2]. These models allow to control safely all the features of a system. However, learning full-control mental-models is impractical as it implies to learn all the features of a system in one big step. This means that newly hired operators are useless before they master the full complexity of the system. Large systems might even be too complex for one operator to manage. In that case, the system must be split in tasks dedicated to different operators. This work investigates the problem of decomposing full-control mental models into smaller independent tasks. These tasks each allow to control a subset of the system and can be learned incrementally to control more and more features of the system. This paper proposes an operator that describes how two mental models are merged when learned sequentially. With that operator, we show how to generate a set of small tasks with the required properties.
Affiliations

Citations

  • APA
  • Chicago
  • FWB

Maudoux, G., Combéfis, S., & Pecheur, C. (2015). Tasks Decomposition of System Models for Human-Machine Interaction Analysis. Workshop on Formal Methods in Human Computer Interaction (FoMHCI), in conjunction with ACM SIGCHI Symposium on Enginnering Interactive Computing Systems (EICS) 2015, Duisburg, Germany. https://hdl.handle.net/2078.5/231545