A model of human performance in information searching tasks.

Ha, Jun Su ; Seong, Poon Hyun

1. INTRODUCTION

The main role of operators in control rooms (CRs) of nuclear power plants (NPPs) is generally to supervise and operate the system. Operators' tasks are performed through a series of cognitive activities (Barriere et al., 2000). Among the activities, correct understanding of the situation is frequently considered a crucial key to improving performance and reducing error (Adams et al., 1995). For the accurate and timely understanding of the situation, monitoring and detection should be effectively accomplished. In other words, effective information searching is crucial for enhancing the safety of NPPs.

In this work, a model of human performance in information searching tasks is proposed in order to provide a framework that can be effectively used for evaluating HMI designs and/or developing a training program. Poor HMI design and poor mental model are modelled to most affect operator's performance. Hence effects of poor HMI design and poor mental model on operator's performance (perception success) are demonstrated through an experimental study. In Section 2, the proposed model is described. An experimental study regarding the model is presented in Section 3. Concluding remarks are maded in Section 4.

2. A MODEL OF INFORMATION SEARCHING

Performance in information searching tasks such as monitoring and detection in NPPs is affected by four factors such as salience, expectancy, value, and effort (Wickens & Hollands, 2000). Operators in NPPs may have expectancy and/or values of various information sources more clearly, as their mental models are getting well developed through experiences and training. This is the reason why experts show better performance than novice in information searching tasks (Wickens & Hollands, 2000; Ha & Seong, 2005). Salience and effort are matters to be considered during designing an HMI. Important information should be designed with appropriate level of salience and effort to access. If an information source cannot be distinguished clearly from adjacent information sources, sometimes the information source may be misunderstood by operators. In addition, if it is too difficult to find out an information source important to understand a situation, an operator eventually gives up finding out the information source. Hence, poor performance in monitoring and detection in NPPs is thought to be mainly caused by a poor mental model and/or a poor design of HMI.

[FIGURE 1 OMITTED]

In complex systems such as NPPs, there are numerous information sources that should be monitored but operators have only limited capacity of attention and memory. Operators selectively attend to important information sources based on their mental models to overcome the limitation. If poor performance in the selective attention is observed, there can be problems regarding the mental model and/or the HMI design. Poor performance in information searching tasks is usually coupled with difficulties such as poor situation awareness, frustration, excessively physical or/and mental load, and so on. If the difficulties are once reported, they are thought to be eliminated by training relevant operators (enhancing operator's mental model) and/or improving the HMI design.

The model can be used as a framework in applications regarding human performance (e.g., especially in evaluating HMI designs or developing a training program). For example, if poor performance results from the mixture of a poor mental model and a poor HMI design, it cannot be found out whether poor performance is originated from poor knowledge on the relevant system or deficiencies in the relevant HMI. Hence deficiencies in HMI can be assessed by evaluating difficulties coupled with poor performance given that the operator's mental model is well-constructed. Developing a training program can be considered in a similar way as well.

3. EXPERIMENTAL STUDY

3.1 Preparaton of the Experiments

The purpose of this experimental study is to investigate the effect of a poor design and/or poor metal model on operator's performance. Subjects conduct complex diagnostic tasks as operators with the FISA-2/PC real time micro-simulator, which simulates a PWR type NPP (Seok & No, 1994). A normal (or acceptable) and a faulty HMI designs are used for this study. The digit number indicators for levels of both loop-A steam generator (S/G (A)) and S/G (B) are intentionally removed from a normal GUI (only bar graph indicators are available), which was identified as a HMI deficiency from the previous study (Ha & Seong, 2007). Also subjects' mental models are controlled by training. The mental model before training is considered poor. The mental model after training is considered well-constructed. Subjects' mental models were scored with four questionnaires about the dynamics of the FISA-2 simulator before conducting experiments. The questionnaires have 100 point scales. The mental model score is then evaluated by averaging the scores of the four questionnaires.

Six tasks including loop-A steam generator tube rupture (SGTR (A)) and loop-B steam line break (SLB (B)) out of fourteen diagnostic tasks are randomly given to subjects. The SGTR (A) and the SLB (B) cases are conducted with the normal GUI and the faulty GUI, respectively.

Subjects were provided with the purpose and the procedure of the experiments and the explanation of fourteen diagnostic tasks before experiments. Subjects conduct diagnostic tasks before and after training. The training includes education in the dynamics of the FISA-2 simulator in the cases of the fourteen diagnostic tasks and exercises with the FISA-2 simulations. After an experiment, it is checked as operator's performance whether changes in the S/G (A) and (B) levels are perceived or not during the experiment.

Fifteen graduate students (fourteen males and one female) with nuclear engineering backgrounds of 5.2 years on average participated as operators in the experiments. They had normal or corrected-to-normal vision.

3.2 Results of the Experiment

[FIGURE 2 OMITTED]

Two S/Gs are equipped in the FISA simulator. Hence, there can be 2/2 success and 1/2 success of the S/G level perception, perception fail, and detection fail, as shown in Fig. 2. 2/2 success means that a subject have perceived the trends of a change in both S/G (A) and (B) levels. The detection fail represents that a subject do not figure out the existence of the change.

After training, 2/2 success were increased 15 % (43-28=15) and 1/2 success had no change. As the GUIs is changed from the faulty ane to the normal one, 2/2 success and 2/1 success were increased 50 % (61-11=50) and 14 % (18-4=14), respectively. Hence, the effect of poor HMI design is concluded to be stronger than that of poor mental model in this study.

4. CONCLUSIONS

In this work, a model of human performance in information searching tasks is proposed. Attention is affected by four factors such as salience, expectancy, value, and effort. Expectancy and value of information are determined based on the operator's mental model, whereas salience and effort are the matter of HMI design. Hence human performance in information searching tasks is modelled to be most affected by poor HMI design and poor mental model and an experimental study regarding this model shows poor HMI design affect more than poor mental model on operator's performance (perception success) this study. Hence much effort should be given to HMI design as well as rigorous training programs in order to improve human performance (eventually system safety such as in nuclear power plants). The proposed model is also expected to be helpfully used as a framework in evaluating HMI designs or developing a training program.

5. REFERENCES

Adams, M.; Tenney, Y. & Pew, R. (1995). Situation awareness and cognitive management of complex system. Human Factors, Vol. 37, No. 1, pp. 85-104.

Barriere, M.; Bley, D.; Cooper, S.; Forester, J.; Kolaczkowski, A.; Luckas, W.; Parry, G.; Ramey-smith, A.; Thompson, C.; Whitehead, D, & Wreathall, J. (2000). Technical Basis and Implementation Guidelines for a Technique for Human Event Analysis (ATHEANA), Rev. 01, NUREG-1624, US NRC.

Ha, J. & Seong, P. (2005). An experimental study: EEG analysis with eye fixation data during complex diagnostic tasks in nuclear power plants. In: Proc. of International Symposium On the Future I&C for NPPs (ISOFIC), Chungmu, Republic of Korea.

Ha, J. & Seong, P. (2007). Attentional Resources Effectiveness Measures during Complex Diagnostic Tasks in NPPs. In: Proc. of International Symposium on Symbiotic Nuclear Power Systems for 21st Century (ISSNP), Fukui, Japan.

Seok, H. & No, H. (1994). Development of a Software for the Micro Simulator for Kori Nuclear Power Plant Unit 2. Nuclear Technology, Vol. 106, No. 3, pp. 384-396.

Wickens, C. & Hollands, J. (2000). Engineering Psychology and Human Performance, 3rd Edition, New Jersey, Prentice Hall.