About MoDyS

The MoDyS research group evaluates and develops methods for modeling operator/user-behavior in dynamic human-machine-systems. Especially quantitative models which describe interaction processes are matter of particular scientific interest. Areas of application include operator support systems, training systems and educational software as well as tools for the human centered design of interaction in dynamic human-machine-systems.

To accomplish this task MoDyS brings together ideas, methods and expertise from different research areas:

• cognitive psychology
• artificial intelligence
• computational intelligence
• human-machine-system studies
• software engineering
• control theory
• first-principles based dynamic modeling of technical systems
• control and information systems engineering

Methods and Equipment

The methods used by the MoDyS research group include performance tests on laboratory multi-task platforms, field studies with chemical plant controllers as expert participants and computational modeling of human performance. Additional techniques involve measures of behavioural indicators for attention and mental workload, primarily by measuring of eye movements, and psychometrically-validated questionnaires.

	Experimental devices
	Measurement methods
	Software

Research on human cognition in dynamic human-machine systems has to cope with a multitude of difficulties that stem from the domain. Complexity and dynamics of the task are nearly unlimited sources of variance that can be hardly controlled in real live situations. Experimentation with higly artifical stimuli that can be easily provided in the lab, on the other hand, suffers from a lack of ecological validity and results are hardly transferable back to the domain.

Microworlds and simulators are nice and promissing experimentational tools, that fill the gap between the before-mentioned extremes. Looking at microworlds from the point of real systems, they appear as reduced and simplified model environments, that have some aspects in common with the real world. From an experimental point of view they might be interpreted as very rich stimuli.

Scalable Multitasking Micro-worlds for Process Control This scalable simulation tool that emulates an industrial process control systems operator console allow to create different tasks envirnoments in the area of - supervisory control - manual control - alarm management - navigation - information selection and - error diagnosis Examples can be found here	Fig 1 Screenshot of a complex multitasking enviroment for tasks in process control
Vibroaccoustic Driving Simulator This experimental device provides a high fidelity driving simulation with a high immersion potential. The simulator consist of a Volkswagen Bora, a VR System including video, sound and vibration and a nelaborated recording suite for car and performance data.	Fig 2 Psycho-physiological measurement in the driving simulator

 

Gaze Tracking Gaze Tracking enables us to measure attention allocation to different tasks and to find fixations on particular points of interest. These measurements are valuable indicators for cognitive processes. Our head mounted device (SMI iView) provides high spacial resolution, the remote systems (Smart Eye) low intrusion.	Fig 3 Course of Attention Allocation to Different Task Elements
Subjective Assessment of Workload : Rating Techniques A multitude of validated rating techniques are used for workload assessment	NASA-TLX (Hart & Staveland 1988) A multi-dimensional rating procedure which allows to assess subjective workload Bedford Scale (Roscoe & Ellis 1990, Huntley 1993) The scale is based on judgements of spare information processing capacity and workload.
Psychophysiological Assessment of Workload: ECG (Electrocardiogram) and EDA (electrodermal activity) Heart-rate (HR) measures are used to assess workload in multitask settings. A variety of information processing activities relate cardiac activity (Wilson & Eggemeier, 1991) The magnitude of EDA provides a reliable index of processing demands in laboratory tasks (see Kramer 1991).
Thinking Aloud Protocols Recording of subjective reflections on present cognitive processes are used to elicitate high level cognitive strategies. This method is conducted concurrently with the task or retrospectively by means of video confrontation.	Fig 4 Clipping of a transcript of a think aloud protocol
Behavioural Videoanalysis Observations about system and operator behaviour are classified by means of a predefined category system to derive quantitative data.	Fig 4 Cutouts of video recordings of an operator interacting with a process control system
Performance Data We have developed a large library of reusable functions to partly automate logfile analysis	Fig 5 Cutout of a data flow defintion for the calculation of a scalar performance measure from time rows

 

ATLAS.ti!	Qualitative analysis of large bodies of textual, graphical, audio and video data
INTERACT	Transcription, coding and processing of observational data
SPSS	Statistical analysis
Clementine	Flow oriented data mining workbench for preprocessing, statistical analysis and descriptive modelling (Neuronal networks, Decision Trees, Clustering and Classification Methods)
MATLAB	Mathematical computation, analysis, visualization, and algorithm development
ACT-R	Cognitive Architecture for low-level user modelling of goal-oriented information processing
Together Control Center	Highly integrated object oriented software model-build-deploy enviroment