The most important model – Human Factors in ETCS design

The most important model – Human Factors in ETCS design

I can even link the model I am discussing to the concept of clickbait – expectancy and value are top-down (or knowledge-driven) factors which help you decide where to allocate your limited attentional resources. Which links us nicely to the model we are discussing, Christopher Wickens’ Model of Human Information Processing [1].

That’s the beautiful thing about this model, you can use it describe to human behaviour in so many situations. Furthermore, you can use this model to understand the role of the human in your system and make better design decisions. In this article I discuss one how the SYSTRA Human Factors team has used the Model of Human Information Processing to assess selective attention and perception for in-cab signalling solutions. In future articles we will discuss other aspects of the Model of Human Information Processing in a diverse range of rail projects.

MODEL OF HUMAN INFORMATION PROCESSING

For anyone with an interest in human factors I would highly recommend reading ‘An Introduction to Human Factors Engineering’ by Christopher Wickens [1]. It was my go-to textbook during university and has been a permanent feature on my desk ever since. In the following sections I will briefly introduce the model, however if I have piqued your interest, I suggest you refer to the far more comprehensive explanation therein.

The human information processing system is comprised of three distinct stages: perception of information, central processing and transformation of that information, and responding to that information. On the left of the model the senses gather information which can be perceived and turned into meaningful information on which we can take action.

Sometimes perception can lead directly to an action, such as slamming on the car brakes when you perceive a hazard ahead. This form of perception leading directly to action is synonymous with the System 1 thinking described by Daniel Kahneman [2]. You think fast, automatically, and with minimal effort. Other times your response is delayed, as information is manipulated within working memory in order to select the most appropriate response. This System 2, or Thinking Slow, in Kahneman’s parlance, is conscious and effortful. An example is trying to solve a complex math problem.

At the top of the model is attentional resources pool. This illustrates the finite pool of cognitive resources we have available to support each stage of the process. This limited pool of resources has a direct consequence on the selection of sensory channels (left of the model) for further information processing. Whilst not shown on the model it is worth noting that a persons’ mood or emotions may influence the available resources or directly bias one of the stages of information processing.

We could not possibly perceive all the information our senses provide us, as such we must select the sources to pay attention to. Wickens’ identifies four factors which influence our selection of channels, these are discussed within the following project example.

ANALYSING THE HUMAN INFORMATION PROCESSING FOR ETCS LEVEL TRANSITIONS

SYSTRA has extensive experience in the design of European Train Control Systems (ETCS) here in Australia and around the world. ETCS is the signalling and control component of the European Rail Traffic Management System (ERTMS). It is a sophisticated train control system that improves rail safety and efficiency. In Europe it is replacing many legacy train protection systems and provides interoperability between jurisdictions. Critically for this example, ETCS most commonly utilises in-cab signalling, rather than traditional lineside signalling.

ETCS is typically implemented in stages, rather than a whole network at once. This results in areas with different levels of supervision and control. For example, the busy inner section of the network may employ ETCS Level 2, whilst the outer regions may retain a legacy system, or a lower level of ETCS. When a train transitions to a region with a lower level of supervision (e.g. ETCS Level 2 to legacy unfitted section) the driver is required to acknowledge this transition on the in cab display, the Driver Machine Interface (DMI). See an example of the DMI indication that must be acknowledged in Figure 2. Note the icon will be flashing.

The SYSTRA Human Factors team sought to better understand human information processing associated with an ETCS level transitions and other features of the ETCS solution, to inform design decisions and to assess any safety and performance risks.

For ETCS level transitions the focus was on the selective attention and perception components of the model. We were interested to know whether drivers would identify the transition alerts and take action and to explore any potentially negative consequences associated with distraction.

Selective attention

A driver is alerted to an ETCS level transitions by two auditory alerts and visual cues on the DMI. One alert occurs in advance of the transition, and the second occurs as the transition is executed. The driver must acknowledge a flashing icon on the DMI within 5 seconds of the second alert (though other configurations are possible). Wickens’ framework is used to assess whether these cues are sufficient to ‘grab’ a drivers’ attention:

• Salience – this is a bottom-up process that relates to something’s ability to capture your attention. The audible alert and flashing yellow icon for level transition are considered a salient cue.
• Expectancy – are you expecting an event? Based on our prior knowledge (top-down process) people are more likely to pay attention where they expect to find information. As level transitions occur in the same fixed locations, a driver will be expecting this to happen when approaching that location.
• Value – how valuable is it to pay attention to a source, or how costly would it be to miss an event in this channel? Failure to acknowledge a level transition to a lower level of supervision will result in an automatic brake application, therefore there is value in a driver paying attention to this source.
• Effort – how much effort is required to attend to a channel? Our selective attention can be inhibited if it is effortful, this is an important design consideration in many contexts. In the case of ETCS level transitions, minimal effort is required.

Based on the four factors of selective attention, it was concluded that a driver is highly likely to pay attention to the level transition alerts and indications. Though as previously noted, we only have a limited pool of attentional resources. Our attention will be influenced by external factors, and any other demands being placed on us at that time. As such it is imperative that ETCS level transitions are positioned at locations with low driver workload. Placing a transition in a complex section of the network, where a drivers attentional resources are depleted, will reduce the likelihood of the level transition ‘channels’ being selected.

Selective attention does not always guarantee perception, a topic that is explored below.

Perception

Perception involves the extraction of meaning from information processed by the senses. Wickens’ defines three simultaneous and concurrent processes that influence ability to perceive. These are discussed below using the example of the ETCS level transition:

• Bottom-up feature analysis – refers to the analysis of the raw features of a stimuli or event. Designers should always seek to maximise bottom up processing. Human factors professionals employ various techniques to achieve this. Examples include the analysis of character heights, colours, font types to optimise text legibility; considering the size, shape and location of icons; and analysis of alarms distinguishability, and audibility above ambient noise level. In the case of the ECTS level transition, bottom-up feature analysis will be influenced by factors such as the icon design (a standardised ETCS feature), viewing distance to the DMI, the presence of glare or other visually degrading factors, and the ambient noise levels within the cab. As such these become important design considerations to optimise bottom-up feature analysis.
• Unitisation – occurs when a person becomes familiar with a set of features that occur together. Using their past experience a person is able to rapidly and automatically perceive sets of features with high unitisation. In the case of level transitions, the same set of stimuli occur at the same location, providing a high level of unitisation.
• Top-down processing – where bottom-up feature analysis is degraded, we are dependent upon top-down processing to ‘correctly guess’ what a stimulus or event is. In our example of level transitions, a drivers’ experience of transitions at that same location will support top-down processing and perception, even if the audible and visual cues are degraded. It is important to optimise top-down processing potential in a design, this can be achieved with simple principles such as maximising discriminating features, providing redundancy, and using a small and simple vocabulary.

This example does not go into detail on the central processing and response stages of the model. Acknowledging a level transition is likely to be an automatic response, based on the factors of selective attention and perception described above. Long term and working memory are fascinating topics where the application of good design and human factors principles can optimise the match between man made products and the human information processing system. We will explore these factors further in future articles.

This article provides an overview of the selective attention and perception components on the Human Information Processing Model. These tools are demonstrated for a relatively simple project example where existing design features and situational factors mean that perception and response are expected. However, for more complex situations, operator perception and action is not always guaranteed. SYSTRA has utilised the same approach to assesses selective attention and perception for a broad range of ETCS features, proving focused recommendations to optimise these processes, and ultimately improve the usability and safety of ETCS solutions.

For more information, please get in contact with Jamie Barton.

REFERENCES

• Wickens, C, An Introduction to Human Factors Engineering, Pearsons, Second Edition, 2003, https://www.amazon.com.au/Introduction-Human-Factors-Engineering/dp/0131837362
• Kahneman, D, Thinking, Fast and Slow, Penguin, 2012
• European Union Agency for Railways, ECTS Driver’s Handbook, Version 1.1.0, 2019