1. Introduction
A reasonable abstraction of causal concepts, and the underlying dynamics of an induced startle, is made by establishing the key drivers of such behaviour, as well as exploring the possible connections that exist amongst them. This paper aims to evidence (as well as justify) the use of fuzzy cognitive maps (FCM) to establish startle causality objectively.
For some context, assume that the human decision-making process during in-flight operations is a fluid activity [1]–[3]. When an unexpected event occurs in flight, causing what is called an aircraft upset, it has been evidenced in the literature that a loss of control (LOC) is a potential result of this situation [4]. In the last 10 years, the LOC category of incidents has contributed to a significant level of all fatal incidents [5]–[8] and is currently a key focal point for improving aviation safety. Furthermore, recent studies have suggested that startle effects have an impact on pilot performance and subsequently leading to a LOC [9], [10]. The consensus on the impact of startle on LOC situations is that such decision making is significantly degraded in circumstances that are unexpected and potentially startling to the pilot in control. Furthermore, the issue of startle potentiated loss of control has been discovered to affect even highly experienced pilots with devastating consequences [9], [11]. For this reason, it has become vital to improving pilot training in particular as a key part of the industry wide strategy to mitigating LOC incidents [5]–[7]. One such way of achieving this is by modern flight simulation technology, to provide high fidelity and quality training programs which can help pilots hone skills that are transferrable to real-life operations.
Rationale
Interestingly, most efforts in this domain , although transferrable across categories, have been heavily focused on commercial and transport categories of the aviation sector. This has generated much impetus to cater for the general aviation (GA) sector as well [12]–[14]. This paper -as part of a larger PhD research focusing on general aviation novice pilots’, aims to support filling the gap by exploring the nuanced assessment of startle propagation – causality, in order to support the eventual development of appropriate training protocols in the flight simulator training paradigm. However, it is clear to see that an understanding of startle as an emotional reaction of humans from a psychophysiological perspective cuts across all category and experience level of pilot operations. Primarily, because the notion of startle, in terms of pilot performances under pressure, is strongly related to the situational awareness (SA) construct [2], [11], [15], [16], and other factors such as the interconnectivity of the pilot’s mental model in relation to the aircraft state, influenced by the environmental factors. As such, it is anticipated that this discussion may serve to extend the current discussion in the domain. The issue of situation awareness specifically has been studied extensively as in [17]–[20] and outside the scope of this paper. The interested reader is encouraged to look to the mentioned works for insights on this topic. However, for context, the widely accepted model of situation awareness consists of three main levels - Perception, Comprehension and Projection of future state status given a dynamic scenario [19]. According to [19], [17], [21] as well as the model presented by [11], this paper takes the position that startle would be prevalent at the level 1 (Perception) stage of SA. It is plausible that the fast appraisal of a situation constrained by the fuzzy nature of a “knee jerk” emotive response to a stimulus, would most likely be of impact at the perception stage of decision making the ultimate goal in a dynamic situation with the potential for LOC.0
Contribution and Organisation of Paper
Considering the above discussion, this paper posits that, in the search for deciphering startle impacts on performance, answers may be found by interrogating the most correlative influencers of such human behaviour – Human factors. In this case, we attempt this through extrapolating from the widely accepted human factors analysis and classification system (HFACS) by [22]. Section II provides a brief discussion on related works around the modelling of human-machine interactions, pertinent to the aim of this research. Such as modelling cognitive based behaviours in operational settings and the notion of attentional resource allocation and indeed all in line with the decision-making process.
Section III explores the FCM as a construct and makes a case for its application to the central principle for studying startle impact on performance. We suggest that the FCM as a complimentary framework to the distillation of human factors, affords a representation, as closely related to reality as possible. Section IV presents the application of FCMs in this present work of analysing startle from Human factors. This represents a unique opportunity to evaluate human functional factors, which are inherent in the performance of a flight task, and which may ultimately influence the pilot’s reaction to a startling event. Section V provides some final thoughts and suggests opportunities going forward.
2. RELATED WORK
In literature, various research [23]–[26] have established a basis of cognitively modelling and engineering the human factors of decision making, safety limitations and time pressures. Several tools have also been devised for these purposes. For example, Man-Machine Integration Design and Analysis System MIDAS [8] [27]–[30], and the Integrated Safety Assessment Model ISAM. [3], [31] in their research, used ISAM to evaluate general aviation safety in the National Airspace System (NAS) from a context of unsuccessful aircraft manoeuvring resulting in loss of control (LOC). The ISAM works by utilising event sequence diagrams (ESD) with fault trees that present parameters of the subject of interest. It is a recognised causal risk model and enjoys much use in the industry. [32] also references the use of ISAM for studying opportunities to mitigate runway safety operations.
The ACT-R framework [17], is also another abstraction that theorises how human cognition works based on psychological experiments. Principally, this method of cognitive analysis seeks to develop a model-based implementation of algorithms, devised on general assumptions of human cognition, with assumptions made about a domain. This process is known as knowledge representations. The model developed, can then be used for comparing actual tasks, which adopt the traditional measures of cognitive psychology. These measures include accuracy of tasks; time that is taken to perform tasks, as well as neurological measures such as MRI outputs. However, this framework, albeit quite successful in its use, is only considered for its inspirational value. It was discovered to be quite instructive and complementary in philosophy to the presented choice of FCM as a framework for understanding startle causal factors and organising these factors into a hierarchy. The FCM is preferred for its potential to capture the theoretical and practical aspects of startle causality easily. In addition to this, the FCM Expert tool [33] presents an exciting prospect for the use of machine learning (ML) algorithms in the expression of innovative decision making, where more esoteric understanding is sought. Its use for this research, however, seeks to expand on the work that has been done in recent decades on improving safety through a multidisciplinary understanding of human factor effects in the aviation industry. Such endeavours are currently a matter of urgency in the General Aviation domain. As mentioned earlier, the urgency stems from the fact that the GA (Part 91) operations have been significantly underrepresented in the existing discourse on LOC [12], [14], [34], [35]. Indeed, this case is more compelling when considering issues such as a loss of situation awareness and even startle, which occurs at the first level of the situational awareness framework [16], [19]–[21]. Given the renewed drive in the industry to close such gaps, this paper seeks to help bridge the gap in the unique way proposed in this paper – Using FCMs. For this work, the Human Factors Analysis and Classification system (HFACS) presents a vital foundation.
The HFACS framework [22], is useful across various disciplines [36]–[40] and applied extensively, giving confidence in its applicability. However, it is applied in this research to the notion of understanding startle causality, in the context of enabling flight simulation and pilot startle resilience training. In the subsequent sections, startle causality is considered in terms of a reasoned probabilistic view on the cause and effect conundrum. Based on a case based reasoning of the uncertainty and unpredictability, a fuzzy representation may be attained to reflect the modality of abruptness in an evolving in-flight situation, capable of causing startle.