MoDyS-Project OpTiHuMS

Operators' Timing in Human-Machine Systems

Summary: Temporal reasoning is part of system state diagnosis in supervisory control. As the estimation of durations is supposed to be a prerequisite of this kind of diagnostic reasoning, temporal errors are a main issue of human-machine reliability. Different hypothesis about the cognitive nature of temporal errors are tested in several simualtion experiments. The results provide clear evidence that temporal errors influence operators' picture of the system state and their decisions.

Temporal Errors in Diagnostic Reasoning

Temporal reasoning is part of system state diagnosis in supervisory control (Hoc, Amalberti, & Boreham, 1995, Decortis, 1988, Zakay, Block, & Tsal 1999), and thus deserves to be treated as an important aspect of human reliability (Decortis, de Keyser, Cacciabue & Volta 1991). As the estimation of durations is supposed to be a prerequisite of this kind of diagnostic reasoning (Decortis et al 1991), especially the observation of temporal errors that occur during supervisory control should call the attention of reliability analysts.

Some temporal errors that were observed in human-machine interaction (Decortis et al. 1991) are ascribed to misestimations of interval durations due to objective (i.e., characteristics of the situation like the number of events that occur during the interval) and subjective factors (i.e., the operators occupation with these events). The phenomenon of duration experience has been under close experimental scrutiny, not at least because time is supposed to be an important aspect of environmental adapation (Block & Zakay 1997). Thus, duration over- and underestimation as a function of concurrent workload is a well-known phenomenon to cognitive psychology (e.g. Brown & Boltz 2002), and several models compete to explain these effects on the level of cognitive mechanisms (for a review, see Mangels & Ivry 2000).

To explain the temporal experience of external duration in the range of seconds and minutes, the psychology of time draws a fundamental distinction between prospective and retrospective duration judgements (James 1890). In the prospective paradigm, participants know in advance that they will be asked to judge the duration of a time period. In the retrospective paradigm, participants do not know until after a time period that they are being asked to judge its duration (see Block & Zakay 1997).

Decortis et al. (1991) report that in human-machine interaction intervals filled with task engagement are tend to be overestimated, whereas empty intervals tend to be underestimated. Based on their observational data the authors conclude in accordance with the psychological evidence that temporal reasoning in human-machine interaction should be based on durations estimations that are made in retrospect (Decortis et al. 1991, p. 67).

Note, however, that retrospective estimations in cognitive psychology designate judgements of participants who are naive about the upcoming temporal task during the presentation of a target interval. As an operator in process control is supposed to have awareness about the relevance of state durations, these situations of duration evaluations described by Decortis et al (1991) have to be assigned to the prospective duration estimation paradigm by definition. For prospective duration estimations there is converging evidence that durations are underestimated with increasing non-temporal task demand during interval presentation, which is contrary evidence to the observations made by Decortis et al. (1991).

It is an open question if the assumptions of attentional models of duration estimation, which are postulated to explain duration-estimation behaviour up to the minute range (Block & Zakay 1997) are sufficient to predict time-based error detection behaviour, not at least because they are based on evidence with low ecological validity (compare Hoc, Amalberti & Boreham 1995).

Method

The influence of different non-temporal information processing demand on a duration estimation task are tested. However, to make predictions about duration judgements in the context of supervisory control, the duration judgement task under scrutiny, as well as the conditions created to produce different non-temporal information processing load have to show a high degree of ecological validity. To assess ecological validity, a microworld is used that simulates the processes in a chemical process plant.

Four psychological simulation experiments were conducted to test for temporal errors in diagnostic reasoning.. Within the microworld framework, a duration-based error detection task is created, which is similar to the situation observed by Schulze-Kissing, van der Meer & Urbas (2001). Participants had to wait for a signal to occur with a fixed delay after scenario onset. The task is to differenciate between scenarios with signal and scenarios without signal. The signal is displayed as a fluid filling a tank in the microworld framework. In a coverstory, the event is introduced as a systems response to an ongoing process. A lack of event after the fixed interval of 60s is introduced as a symptom for a system failure. Thus the only criterion to make a decision is the duration of the system-response delay itself. This error-detection task is instructed as the primary task. The importance of this error-detection task was underpinned by the instruction, that a system failure has to be evaluated as dangerous. A decision about a system failure has to be given by pressing one of two optional buttons, one symbolizing the initiation of a shut-down procedure, the other symbolizing a positive response to correct system performance. For clocks are removed, decisions have to be taken on the basis of duration estimations.

Results

The experimental effects converge and provide clear evidence for the existence of temporal errors in human diagnosis. However, the direction is not the one that is expected by Decortis et al. (1991). The time judgement data show the classic interference effect in timing. Error-detection delays increase as a function of concurrent-task demand. The general underestimation of elapsed duration with increased concurrent task difficulty, which is indicated by significant delays in error detection, shows that attentional models of prospective duration estimation do better account for the diagnostic behaviour.

Resume

The results of the current studies are still relatively preliminary in terms of potential generalizability to real-world diagnosis. However, the main implication of these results is that temporal errors do occur in diagnostic reasoning, especially under multi-tasking conditions, and thus have the potential to influence operators' picture of the system state and their decisions. Temporal errors are an issue of human-machine reliability, which is still treated as Cinderella. We hope to have shown that an experimental approach with microworlds is a method to make considerable progress in explaining the nature of temporal errors. This, as we think, bears the opportunity to predict situations where temporal errors occur within human-machine interaction, and to better understand a reason why accidents happen.

Literature

Schulze-Kissing, D., van der Meer, E. & Urbas, L. (2004). A Psychological Analysis of Temporal Errors in Human-Machine-Systems. Proceedings of the IFAC Symposium: Analysis, Design and Evaluation of Human-Machine-Systems. Atlanta, USA, 07.-09. September 2004.

Schulze-Kissing, D., van der Meer, E. & Urbas, L. (2004). Zeitpsychologische Ableitung eines Gestaltungskriteriums für die Mensch-Maschine-Schnittstelle in komplexen Systemen. In M. Thüring, L. Urbas und C. Steffens (Ed.), Entwerfen und Gestalten. 5. Berliner Werkstatt Mensch-Maschine-Systeme (Vol. 14, pp. 104-126). Düsseldorf: VDI Verlag.

Schulze-Kissing, D., van der Meer, E. & Urbas, L. (2003). The Effect of Temporal Distortions on the Error Diagnosis in the Control of a Complex Technical System. In Th. Bajo & J. Lupiánez (Eds.), XIII Conference of the European Society of Cognitive Psychology ESCOP 2003 (pp. 295-296). Granada: Actas.

Schulze-Kissing, D., van der Meer, E. & Urbas, L. (2003). Zeitliche Informationsverarbeitung in Mensch-Maschine-Systemen am Beispiel einer simulierten Prozesstechnischen Anlage. In J. Golz, F. Faul, & R. Mausfeld (Eds.), Experimentelle Psychologie. Abstracts der 45. Tagung experimentell arbeitender Psychologen. Christian-Albrechts-Universität zu Kiel, 24-26.03.2003 (pp. 139). Lengerich: Pabst.

Schulze-Kissing, D., Oldenbourg, R. & Urbas, L. (2003). Blickbewegungsmessung als Indikator für Zeitempfinden bei der Steuerung einer prozesstechnischen Simulation - Eine explorative Studie. MMI Interaktiv, 6, 65-75.

Schulze-Kissing, D., Urbas, L., van der Meer, E., & Leuchter, S. (2002). Der Einfluss kognitiver Beanspruchung auf zeitgesteuertes Verhalten in Mensch-Maschine-Systemen. In E. van der Meer, H. Hagendorf, R. Beyer, F. Krüger, A. Nuthmann, S. Schulz (Ed.), 43. Kongress der Deutschen Gesellschaft für Psychologie. Humboldt Universität zu Berlin: 22.-26. September 2002. Programm, Abstracts (pp. 136-137; TS0231). Lengerich: Pabst.