1 5 T H an Nu Al Ac T - R W or Ks Ho P July 18 - 20 , 2008 Adamson Wing 136 a - Baker Hall

A model of prospective time-estimation is introduced which explains the interplay of working memory demands on duration estimation. The approach is integrated into ACT-R and tested by estimating the duration of a task that varied coordinative and sequential demands on working memory. The comparison with experimental data shows that the model is able to simulate the influence of these demands on human time-estimation. Introduction The cognitive ability to be aware of the passage of time is beneficial in dynamic environments. Time-judgments are important to stay tuned to this environment, to plan steps in a task, and to identify problems (e.g. after an expected duration of booting a computer the monitor stays blank). In the context of human-machine interaction, the knowledge of temporal dependencies is of great interest. For example, in order to drive safely, drivers need to divide their visual attention in a reasonable way between traffic and secondary tasks such as In-Vehicle-Information-Systems. Operators can deduce a malfunction from the system’s temporal behavior in comparison to the temporal properties of a functioning system (Schulze-Kissing, 2007). Here we introduce a computational model of timeestimation that shows how a demanding task disrupts the ability to judge time. In this model, the need to maintain and update information (e.g. a number in arithmetic) during a task distorts the construction of time representation during this period. The approach is integrated into ACT-R (atomic components of thought – rational analysis; Anderson et al., 2004). In this way the influence of cognitive processes and demands on the construction of time representations can be explored in a cognitive context. For cognitive architectures, it is valuable to have an integrated component that simulates temporal human behavior. This is especially important for modeling switching tasks, multitasking and tasks under time-pressure. The integrated timing-model is tested within a counting task (Dutke, 1997) with varying demands to compare human data to the performance of the model. Psychological Models of Time-Estimation The research field of human time-estimation explains differences in estimates on a number of factors such as the duration of the interval, the kind of instruction given to the subjects, when and how an interval is estimated (production, reproduction), or the number of incidents experienced during a given interval. It is generally found that a demanding task affects timeestimation. Time-estimates are shorter when compared to less demanding conditions (Zakay, 1993; Dutke, 1997; Brown, 1997). A number of authors (e.g. Block & Zakay, 1996; Brown & West, 1990) assume that attentionallocation is the responsible factor for the interference between task and time-estimates. A number of other authors assume a strong influence of working memory on timeestimation. Quantitative Time-Estimation Model The proposed model of prospective time estimation consists of four parts: a pacemaker that generates pulses, an accumulator which collects pulses for short durations, a process of construction which updates the time representation, and a procedure which finally estimates time, e.g. by comparing an old time representation with a new interval as in the reproduction task. The first two parts are modeled by adding a timing-module to the architecture. The third and forth parts of the approach integrate the output of the new timing-module with already existing processes and modules of the cognitive architecture. Discussion The approach under discussion explains the way working memory demands effect duration estimation. For a timeestimation, the temporal representation during an interval has to be updated continuously. In order to do this the latest representation has to be maintained in working memory. A task that calls upon working memory mechanisms interferes with the working memory mechanism of maintaining the latest time-representation. Compared to other theoretical accounts of duration estimation, this model is more parsimonious in that no additional elements like an attentional gate or processes of a central executive are necessary to explain observable distortions of the estimation process. Acknowledgments This work was sponsored by VolkswagenStiftung (Research Group Modelling of User Behaviour in Dynamic Systems) and Deutsche Forschungsgemeinschaft (DFG Research Training Group Prospective Design of Human Technology Interaction, GRK 1013). 15TH ANNUAL ACT-R WORKSHOP 28 References Anderson, J. R., Bothell, D., Byrne, M. D., Douglass, S., Lebiere, C., & Qin, Y. (2004). An integrated theory of the mind. Psychological Review, 111, 1036–1060. Baddeley, A. (1986). Working memory. Oxford, UK: Clarendon Press. Block, R.A., & Zakay, D. (1996). Models of psychological time revisited. In H. Helfrich (Eds.), Time and mind. Proceedings of the International Symposium on Time and Mind (pp. 171-195). Kirkland, WA: Hogrefe & Huber. Brown, S.W. (1997). Attentional resources in timing: Interference effects in concurrent temporal and nontemporal working memory tasks. Perception & Psychophysics, 59, 1118-1140. Brown, S.W., & West, A.N. (1990). Multiple timing and the allocation of attention. Acta Psychologica, 75, 103-121. Dutke, S. (1997). Erinnern der Dauer: Zur zeitlichen Rekonstruktion von Handlungen und Ereignissen [Remembering duration: The temporal reconstruction of actions and events]. Lengerich, Germany: Pabst Science Publishers. Dutke, S. (2005). Remembered duration: Working memory and the reproduction of intervals. Perception & Psychophysics, 67, 1404-1422. Schulze-Kissing, D. (2007). Eine Mikroanalyse des zeitlichen Fehlers in der Mensch-Maschine-Interaktion. VDM Verlag. Taatgen, N.A., van Rijn, H., & Anderson, J.R. (2007). An Integrated Theory of Prospective Time Interval Estimation: The Role of Cognition, Attention and Learning. Psychological Review, 114(3), 577-598. Zakay, D., (1993). Relative and absolute duration judgments under prospective and retrospective paradigms. Perception & Psychophysics 54, pp. 656–664. 15TH ANNUAL ACT-R WORKSHOP 29 The Influence of Task Demands in a Model of Time Estimation Nele Pape, Technische Universität Berlin 2 19.07.2008 Nele Pape ACT-R Workshop 2008 Time-judgments are important to: stay tuned to the environment to plan steps in a task to detect delays, i.e. absence of events • In the context of human-machine interaction, the knowledge of temporal dependencies is of great interest, e.g.: in order to drive safely to deduce a malfunction from the system’s temporal behavior 3 19.07.2008 Nele Pape ACT-R Workshop 2008 Why is time estimation important for cognitive architectures? • It is valuable to have an integrated component that simulates temporal human behavior: switching tasks multitasking tasks under time-pressure (subjective time on task / remaining time) timing of action without external cues 4 19.07.2008 Nele Pape ACT-R Workshop 2008 Psychological Model: Attentional-Gate Problems: What is attention-to-time? What corresponds to this in ACT-R? [Block & Zakay 1996; Zakay & Block 1997] 5 19.07.2008 Nele Pape ACT-R Workshop 2008 Models of prospective time-estimation... ... are already implemented in ACT-R (Byrne, 2006; Taatgen, van Rijn & Anderson, 2007) • Byrne: Attentional Gate Attention can either be devoted to time (which opens the “gate,” allowing pulses to be counted), or to other tasks In ACT-R terms, this means: a production has to fire to increment the counter pulses can be missed, producing underestimates • Taatgen: Slow down of pacemaker (duration < 15 sec) If a task is very demanding people forget to estimate time and have to restart their estimation no account what kind of task demands are responsible for distortions in time-estimates 6 19.07.2008 Nele Pape ACT-R Workshop 2008 The ‘Task Sensitive Time Estimation Module’ (TaSTE Module) • Target: Prospective time estimation for longer durations • Foundations: Empirical findings of influence of task demands on time estimation (Brown, 1997; Dutke, 2005) Time Traces in Episodic Memory (Baddely, 2002) • No Magic Doctrine: Use regular memory processes such as Spreading Activation and the concept of Working Memory already implemented in ACT-R to „Emerge“ human time estimation effects (Lovett, Reder & Lebiere, 1999) 15TH ANNUAL ACT-R WORKSHOP 30 7 19.07.2008 Nele Pape ACT-R Workshop 2008 The ‘Task Sensitive Time Estimation Module’ (TaSTE Module) • Pacemaker • Accumulator • Reconstruction Process • Reproduction or comparison Declarative Memory First time chunk 0 Very old time chunk 8 Pacemaker Accumulated pulses: Timing Module