Traffic Psychology and Sustained Mobility Literature summary
LECTURE 1
LEWIS-EVANS (2012)
Michon’s review of driving behaviour models (1989) approached these models by characterising them in terms
of what made them ‘move’. He divided theories of driver behaviour into those which move ‘magically’ based on
curve fitting, and those rational or functional models which are driven by explicitly stated concepts and rules.
Michon saw the most promise in the latter category (described as models which can adapt and learn).
The category of functional models was further split into those which Michon described as motivational and
those which are rule based.
The motivational models include:
- Taylor’s Risk-Speed Compensation Model (Taylor, 1964)
- RHT (Wilde, 1976)
o Critique: it contains circular explanations
- Threat Avoidance Theory (Fuller, 1984)
o Critique: it doesn’t handle nested behaviour
- Zero-Risk Theory (Näätänen & Summala, 1974) (only mentioned in Michon’s early review)
Michon critiqued these motivational models for being over concentrated on risk and road safety, rather than
explaining the normal everyday basis of driving. Furthermore, Michon views the explanations and mechanisms
of the motivational models as not sufficiently identifying the underlying cognitive mechanisms that they rely
on.
Instead, Michon promotes rule based theories, which are based on the influence of cognitive science and the
idea of schema or scenarios. This means that rule based theories state that behaviour operates along the lines
of finding a correct rule, and then applying it in a “IF THEN” fashion. For example, IF I feel that my wheels are
slipping on the road THEN I begin to drive cautiously. As an example of the rule based type of theoretical
approach, Michon referred to the SOAR cognitive architecture (Laird, Newell, & Rosenbloom, 1987).
Like Michon, Ranney (1994) praises the motivational models for incorporating motivational aspects into their
accounts of driver behaviour. However he raises concerns that motivational models have been somewhat
unclear as to how they operate, leading to a lack of testable hypotheses, and therefore, a lack of productive
research into the operation of these theories. Ranney also comments, as Michon did, that there has been an
over emphasis on the idea of risk as a controlling variable in behaviour. He also adds that a quite protracted
debate over the total risk compensation mechanism proposed by RHT had stalled progress in the area of model
development.
Ranney also sees more potential in the use of information processing theories to explain driver behaviour, but
notes that cognitive information processing theories had not made much of an impact on driver behaviour
models at the time. Ranney therefore introduces the idea of hierarchical control models, particularly through
the use of the Skill-Rule-Knowledge framework (Ramussen, 1987a), the Generic Error-Modelling System (GEMS)
proposed by Reason (1990) and Michon’s Three Level Control hierarchy (Michon, 1985).
In association with the idea of automaticity Ranney praises information processing models as moving away
from a concentration on risk, and through the use of hierarchies, a move towards modelling driver behaviour in
the frame of a more complex systems view.
Huguenin & Rumar (2001) state that the models which have been put forward have tended to be either so
broad as to be unusable for generating useful predictions, or so specific as to only explain certain small parts of
,Traffic Psychology and Sustained Mobility Literature summary
the driving task. They do however present a classification of models of driver behaviour as either driver task
related, functional control, or motivational (Rothengatter, 1997) but then reject it as being overly simplistic and
not capturing the full complexity of driver behaviour.
Huguenin and Rumar (2001) discuss risk in terms of objective versus subjective risk comparisons by drivers, and
the occurrence of behavioural adaptation. Behavioural adaptation is the idea that drivers tend to adapt their
behaviour to changes in the environment and may therefore end up reducing or negating any potential safety
gains which these changes may have brought about.
Huguenin and Rumar have also described what they class as motivational models. They see motivational
models as models that state that there are expectancies for action, and then certain valences (or benefits)
which are applied to different acts. Therefore, under this classification, behaviour is determined in a kind of
utility fashion where the expected results of behaviour are weighted with the expected benefits and then the
most beneficial option is taken. The upshot of this is that due to the relatively low accident likelihood
associated with most on-road behaviours, and the high valence of unsafe behaviour, that unsafe behaviours
(for example speeding) may be chosen more often than objectively safer alternatives.
Rothengatter is generally positive towards the use of attitudinal theories (e.g. the Theory of Planned
Behaviour), although he does point out that, methodologically speaking, their use within driver research has
been problematic in terms of an over reliance on self-report. He concludes, much like Huguenin and Rumar
(2001), that the focal point for traffic psychology, and therefore for the development of theories of driver
behaviour, should be where performance theories and social psychology theories meet.
These 4 reviews (Michon, Ranney, Huguenin & Rumar, and Rothengatter) all seem to agree that older trait and
skill based models of driver behaviour will not get the field any further in terms of understanding normal
everyday driving behaviour. They also all put forward, in one way or another, the most common argument
against the models of driver behaviour that existed at the time, the argument of unfalsifiability. This argument
states that many of the existing models are too vague in their underlying mechanisms and, therefore, it is
difficult to form useful predictions from them and to then test these predictions.
As mentioned above, the most common criticism of models of driver behaviour is that they do not produce
testable hypotheses and that they are descriptive rather than predictive. In an attempt to correct this situation,
Fuller (2000) proposed the model of Task-Capability Interface (TCI) as a representation of the driving task, and
Task Difficulty Homeostasis (TDH) as an explanation of driver behaviour.
The TCI model contains two components: Capability and Task Demands.
According to this model, drivers possess a certain level of capability, which is constructed in a hierarchical
fashion. The combination of the learnt and physiological components is referred to as the driver’s competence,
as can be seen below in Figure 2.1.
,Traffic Psychology and Sustained Mobility Literature summary
Given the resource limited nature of human performance, it is unrealistic to expect people to operate at their
full level of competence all of the time, so the final step in the creation of capability is to subtract human
factors from competence. The term ‘human factors’ refers to situation-dependent attributes such as
motivations, attitudes, and state-conditions like intoxication, fatigue, and emotional state. Human factors are
relatively unstable, unlike the components of capability that come before it.
The result of this hierarchical process is capability, which reflects the current situational ability of the driver to
respond to task demands. This is not simply the amount of effort a driver has available at any particular
moment. Rather, it is the amount of effort available plus the knowledge and experience of how to best use and
apply this effort, minus any relevant situational-dependent factors.
Task demands sit on the other side of the TCI and reflect the environment in which the driver is currently
operating. It is a combination of environmental factors, including: the road environment, behaviour of other
road users, weather conditions, vehicle characteristics, road position, and the trajectory of the driver’s car and
its travelling speed. The speed and trajectory of the driver are also linked to their capability.
Within TCI, task difficulty is produced by the interaction of task demands and capability. Task demands can be
seen as representing the minimum amount of capability required to retain control of the vehicle. Therefore if
task demands exceed capability then loss of control occurs. Furthermore, as task demand approaches
capability then performance loss begins to occur.
However, in the end, TCI is just a description of the driving task and does not really generate many testable
hypotheses. It also does not detail how drivers determine that their choices will not place them in a situation
where their capability cannot match the demands of the task, and does not particularly help to explain
behavioural adaptation. To address these issues, Fuller proposed Task Difficulty Homeostasis theory.
Task Difficulty Homeostasis theory (TDH) states that people have a set range of experienced task difficulty at
which they prefer to operate at. This range of task difficulty is not static, but rather determined by a driver’s
perceived capability, effort motivation, and trip goals, and can therefore vary between, and even within, trips.
Perceived task difficulty, a product of the interaction between perceived capability and task demand, is then
compared with preferred task difficulty and action is taken if the currently perceived task difficulty falls outside
of the target range. This leads to task difficulty being constantly maintained in a homeostatic way. This means
that behavioural adaptation should occur whenever drivers detect that they are operating at a level of task
difficulty outside of their preferred range.
The model did not stand still, though, and continued to evolve eventually incorporating nearly the entire TCI
model within TDH itself. The incorporation of TCI into TDH was mainly carried out through the addition of distal
, Traffic Psychology and Sustained Mobility Literature summary
and proximal determinants. These determinants act on different parts of the core comparison between
perceived task difficulty and preferred task difficulty, and are intended to make the model more
comprehensive. See figure 2.2 below.
Unlike in TCI, perceived task demand now also appears to be constructed in a hierarchical fashion.
Another part of TDH is the idea of a risk threshold. This risk threshold was conceptualized as being similar to
that already proposed by earlier theories, such as zero-risk theory, and triggers when drivers are near the limit
of their preferred range of task difficulty. In this way it can act as a warning, in addition to the constant
comparison of perceived and preferred task difficulty, that the preferred range of task difficulty is about to be
exceeded (Fuller, 2005; Fuller et al., 2008). Importantly, the risk threshold within TDH does not refer to crash
risk as detailed in earlier theories of driver behaviour, such as Risk Homeostasis theory (Wilde, 1976), but
rather to a ‘feeling of risk’.
Risk as a variable of import within TDH did not stop with the risk threshold, however. Rather in 2008, Fuller
reconceptualized TDH into Risk Allostasis Theory (RAT). This new form of the model is summarized in Figure
2.3.
Allostasis refers to a process of achieving stability in a system through making changes. This is in contrast to
homeostasis which refers to achieving stability in a system by staying the same.
