Coordinating Tasks in a Simplified Driving Environment Modelled with ACT-R and Threaded Cognition
Master's thesis of Hannah Kosanke
On the highway drivers are not only engaged in lane keeping, but perform an array of other tasks such as having a conversation, route planning or texting. Therefore driving is a prime example for multitasking in an applied context. A modelling approach to realize multitasking in general is “Threaded Cognition” first introduced by Salvucci and Taatgen in 2008. “Threaded Cognition” uses no explicit resource allocation but chooses the task, which has not been executed longest during conflicts and allows for parallel processing of non-conflicting tasks. For exploring the limits of “Threaded Cognition” in an applied setting, modelling approaches in ACT-R were realized and compared with data from a laboratory multitasking experiment realized within the master thesis of Marika Nürnberg  (2015). In Nürnberg’s experiment, 45 Participants dealt with tasks related to driving in a single modus and in double and triple combinations. The experimental tasks chosen for the paradigm were the critical tracking task (CTT) simulating steering, a two-level Working Memory Updating task (WMU) mirroring traffic complexity and the Peripheral Detection Task (PDT) corresponding to irregularly occurring events in traffic, which require the driver’s attention. The experimental data showed that the performance in the CTT-task decreased with the number of executed tasks. In the WMU and PDT task the reaction times as well as the error rates increased with the number of tasks performed simultaneously. The modelling approaches realized in this master thesis match the general trends of the experimental data but show a more distinctive drop in performance for the CTT task and a less drastic increase in the percentage of errors during the difficult WMU over the number of performed tasks. One explanation could be an insufficient resource allocation mechanism offered by the “Threaded Cognition” approach, which could be improved with an explicit prioritization of the CTT task over the other two tasks. Subsequently, a possible next step would be the development more refined task prioritization modelling to allow for a better approximation of the experimental data.