Performance Analysis of Stimulus Rich Reactive Interfaces

Organisation:	Centre of Cognitive Neuroscience and Cognitive Systems, University of Kent, Canterbury (UNITED KINGDOM) MRC Cognition and Brain Sciences Unit, Cambridge (UNITED KINGDOM)
Method:	LOTOS
Tools used:	CADP (Construction and Analysis of Distributed Processes)
Domain:	Cognitive Systems.
Period:	2004 - 2008
Size:	1,453 lines of LOTOS code 564 lines of C code
WS
Description:	General cognitive frameworks, such as ICS (Interacting Cognitive Subsystems) can be applied to analyze interfaces for human-computer interaction. This case study considers SRRIs (Stimulus Rich Reactive Interfaces), which present stimuli rapidly to a human operator, as happens for instance in an airplane cockpit. Salient stimuli, called targets, are prioritized and processed more extensively than others, called background items. In SRRIs, stimuli may arrive so rapidly that they may interfere with each other, so that human operators could potentially miss some highly salient stimuli. A large number of psychological experiments have shown a temporal attention limitation, known as attentional blink: Humans may miss a second target following a previous one after a few hundred milliseconds. This case study is a result of the EPSRC funded project Computational Modeling of Salience Sensitive Control in Humans and in Artificial Systems, under the AIBACS programme (Adaptive and Interactive Behaviour of Animals and Computational Systems), which developed a spectrum of computational models of temporal attention. These models explain how meaning and emotion direct attention in time by reproducing a number of human behaviours, and they also account for the interaction between cognition and emotion. The case-study presents an ICS model of attentional capture, represented as a "box and arrow diagram", as it is usual for ICS and other psychological theories. This ICS model represents cognition as a 2-stage pipelined process, the first stage determining if an incoming item is salient and the second stage more precisely evaluating the items selected by the first stage. Attentional blink occurs because attention can be engaged at one stage at a time. From the ICS model a LOTOS specification has been derived, based on the idea that ICS boxes correspond to LOTOS processes and ICS arrows correspond to LOTOS communication links. Using the EXEC/CAESAR environment for rapid prototyping, this LOTOS specification has been transformed into an executable C program which allowed to simulate the human behavior efficiently. Psychological simulations often obtain good results from the average of thousands of runs. The number obtained by taking average over thousands of simulation runs based on EXEC/CAESAR closely fits the numbers obtained by human participants. In a second step, the LOTOS model has been combined with different interfaces of an SRRI so as to evaluate the probability of reporting target stimuli. These experiments allowed to estimate the conditions under which attentional blink aware interfaces (which ensure that targets are always separated by a certain number of less salient stimuli) improve the performance (i.e., increase the number of reported targets without introducing too much delay).
Conclusions:	A strength of formal methods is that they are abstract and the resulting specifications of the operator are general purpose, ensuring the broadly applicability of results. Thus, new modeling techniques from computer science can also be employed in computational modeling of the mind. These would complement existing techniques, being specifically targeted at psychological level modeling, in which it is advantageous to directly represent the distribution of control. In [Barnard-Bowman-04] was illustrated the potential of process algebra, and LOTOS in particular, to implement modular mental architectures of wide scope, in which control is distributed rather than centralised. Drawing on the ICS mental architecture, [Barnard-Bowman-04] and [Su-Bowman-Barnard-07] present models of the attentional blink effect implemented using LOTOS. Moreover, the models have been extended to simulate human performance in the context of human-computer interaction [Su-Bowman-Barnard-08], which shows that these models are applicable across a range of practical settings when human attentional limitations come into play. In this respect, the model described in [Su-Bowman-Barnard-Wyble-08] also simulates human electrophysiological data, such as electroencephalogram recordings, which can be compared to real electrophysiological data recorded from human participants. This model has been used to evaluate the performance trade-offs that would arise from varying key parameters and applying either a constructive or a reactive approach to improving interactive systems in a stimulus rich environment. The EXEC/CAESAR environment of CADP enables the use of LOTOS to formally specify and simulate box-and-arrow diagrams of psychological theories, enabling the comparison of psychological models with the behavior of humans as well as the evaluation of different human-machine interfaces.
Publications:	[Barnard-Bowman-04] Philip J. Barnard and Howard Bowman. "Rendering Information Processing Models of Cognition and Affect Computationally Explicit: Distributed Executive Control and the Deployment of Attention". In Cognitive Science Quarterly, Hermes Science Publications, vol. 3, no. 3, pp 297-328, April 2004. Available on-line at: http://www.cs.kent.ac.uk/pubs/2004/1558/content.pdf [Su-Bowman-Barnard-07] Li Su, Howard Bowman, and Philip J. Barnard. "Attentional capture by meaning: A multi-level modelling study". In Proceedings of the 29th Annual Cognitive Science Society (Austin, Texas, USA), pages 1521-1526, 2007. Available on-line at: http://www.cogsci.rpi.edu/csjarchive/proceedings/2007/docs/p1521.pdf [Su-Bowman-Barnard-08] Li Su, Howard Bowman, and Philip Barnard. "Performance of Reactive Interfaces in Stimulus Rich Environments, Applying Formal Methods and Cognitive Frameworks". In Antonio Cerone and Paul Curzon, editors, Proceedings of the 2nd International Workshop on Formal Methods for Interactive Systems FMIS'2007 (Lancaster, UK), Electronic Notes in Theoretical Computer Science, volume 208, pp. 95-111, April 2008. Full version available on-line at: http://www.cs.kent.ac.uk/pubs/2007/2595/content.pdf or from the CADP Web site in PDF or PostScript [Su-Bowman-Barnard-Wyble-08] Li Su, Howard Bowman, Philip Barnard, and Brad Wyble. "Process algebraic modelling of attentional capture and human electrophysiology in interactive systems". In Antonio Cerone, Paul Curzon, and D. A. Duce, editors, Formal Aspects of Computing Volume 21, number 6, pp.513-539, December 2009. Available on line at: http://www.cs.kent.ac.uk/pubs/2009/2818/content.pdf or from the CADP Web site in PDF or PostScript
Contact:	Li Su Cognitive Neuropsychiatry Department of Psychological Medicine Institute of Psychiatry at King¹s College London Box P068 De Crespigny Park London SE5 8AF United Kingdom Tel: +44 (0)20 7848 5134 Email: Li.Su@iop.kcl.ac.uk Web: http://www.iop.kcl.ac.uk/staff/profile/default.aspx?go=12060 and http://www.cs.kent.ac.uk/people/rpg/ls68/index.html
Further remarks:	This case study, amongst others, is described on the CADP Web site: http://cadp.inria.fr/case-studies

Back to the CADP case studies page