Time and Control 167
evaluate events and update/develop an understanding of the situation (T
E
),
the time needed to choose or select an appropriate action (T
S
), and the time
window allowed for execution after the response has been selected (T
W
). (A
more complete version of the model also includes T
P
, the time needed to
perform an action.)
The other set comprises the points in time when things happen, denoted as
τ
O
and τ
LFT
. The former is the point in time when an action becomes
necessary, either due to an external event or due to the person (e.g., an
intention or recognised need to do something). The latter is the latest
finishing time (LFT) for the action that corresponds to what happened at τ
O
(cf. Allen, 1983). This means that the total time available to respond to an
event cannot be larger than T
A
= τ
LFT
- τ
O
. In practice, the time available to
carry out the response is T
W
rather than T
A
, where T
W
= T
A
– (T
E
∪T
S
). (We
use T
E
∪T
S
rather than T
E
+T
S
because T
E
and T
S
may be overlap.) According
to this line of reasoning, if (T
E
∪T
S
∪T
P
) > T
A
then the operator has too little
time to understand what is going on and to respond effectively, and control
therefore will sooner or later be lost. Conversely, when (T
E
∪T
S
∪T
P
) < T
A
,
then the operator has enough time to understand what is going on and to
choose and effectuate a response, and is therefore likely to remain in control
and possibly even able to plan ahead. (The simplicity of the model in Figure
8.2 suggests that T
E
and T
A
occur in a sequence and at discrete times. That is,
however, an artefact of the graphical representation. In reality, event
evaluation and action choice will usually be intermingled, although there is a
logical necessity that the former precede the latter. Figure 8.2 represents the
case of a single action of undefined scope. However, it is quite possible to
apply the same type of reasoning to more complex interactions and to
combined or aggregated actions although it becomes messier. The principles
are nevertheless easiest to illustrate with an idealised case.)
Research on human-computer interaction has generally referred to tasks
that were self-paced and where limited time therefore played only a minor
role. In contrast to that, tasks in most industrial domains are force-paced or
process-paced. The available time, T
A
, is determined by the speed of the
process and if (T
E
∪T
S
∪T
P
) exceeds T
A
, it puts severe constraints on the
operators’ possibility to evaluate events and select actions. Some processes,
such as steel rolling mills, electronic trading, or flying an airplane, require
rapid or even near instantaneous responses (cf. Table 8.2). Other processes
such as power generation, land-based transportation or surgery pose less
severe demands but still require that actions be taken within a limited time. It
would clearly be better if there were ample time, i.e., if (T
E
∪T
S
∪T
P
) were
less than T
A
, since the operator then would have time not only to respond in
the situation but also to refine the current understanding, to plan before
acting, hence to improve control of the situation. This can be achieved if the
time limitations can be relaxed, for instance by slowing down the process,
168 Joint Cognitive Systems
although that is possible in only very few cases. A more common approach is
to reduce either T
E
or T
S
by improving the system and interface design,
although this usually is done in a piecemeal fashion.
In cases where the process develops at a regular or stable speed and where
there are few unexpected events, the τ
LFT
of event
i
is predictable, since it
corresponds to τ
O
of the following event
i+1
. There are, however, many cases
where the progress of events is irregular or where unexpected events are
common. Unexpected events are not just due to possible failures of the
process, but equally and more often due to the unpredictability of the
environment (cf. the discussion in Chapter 7). If τ
LFT
is uncertain, then the
JCS in principle has two possible strategies. One is to stick to the task and
ignore unexpected events, unless they are very serious. (Note that this does
require some kind of quick evaluation or screening, similar to the problem in
selective attention.) The other is to try to complete the action at hand as
quickly as possible, i.e., to minimise T
E
, T
S
, and T
P
. This is consistent with
the ETTO principle, according to which people trade-off thoroughness for
efficiency in order to gain time. In this case the trade-off is, however, due to
uncertainty rather than to external production pressures.
Controller /
controlling
system
Events /
feedback
Action
Construct
τ
O
,
τ
LFT
External event /
disturbance
T
W
= time
window (time
allowed) for
execution
T
W
T
W
= time
window (time
allowed) for
execution
T
W
T
P
= estimated
performance time
T
P
T
S
= time to
select action
T
S
T
E
= time to
evaluate event
T
E
Figure 8.2: Time and control in COCOM.
Time and Control 169
Time and Control Modes
The predominant position of event evaluation and action selection in
COCOM provides an easy way of accounting for the coupling between time
and control and thereby to provide a little more detail to the description of
Figure 8.1. As described elsewhere (Bye, Hollnagel & Brendeford, 1999),
event evaluation and action selection can be carried out to various depths
depending on the control mode. This corresponds to the basic fact that
humans, unlike machines, may be more or less thorough in what they do
depending, among other things, on the available time. This provides both a
single way of expressing the time-control dependency conceptually and a
simple way of implementing it computationally.
To study this in more detail it is useful for a moment to depart from the
cyclical model and focus exclusively on the different time sets. The relations
can be laid out as shown in Figure 8.3, which depicts a situation where there
is only a single line of action. It is therefore allowable to treat the events as if
they were ordered sequentially. Figure 8.3 illustrates the problems that can
occur when the three prototypical parts of an action have to be
accommodated within an available time window, T
A
. In cases where the
available time is insufficient for event evaluation, action selection, and
execution together, the JCS will be incapable of responding before it is too
late. If this condition is prolonged, it will sooner or later lead to a loss of
control, hence to a degradation of the control mode. The relations can be
described as in Table 8.3.
Time needed for evaluation: T
E
Time needed for selection: T
S
Time
Event
i
occurs
Event
i+1
occurs
Performance time: T
P
τ
O
τ
LFT
Available time: T
A
= τ
LFT
- τ
O
Figure 8.3: Temporal relations at work.
The conditions listed in Table 8.3 are indicative rather than exhaustive.
This goes especially for the condition corresponding to the opportunistic
control mode. One obvious improvement would be to add a conditional
condition, for instance that (T
E
+ T
S
+ T
P
) < T
A
. Another would be to increase
the resolution by considering the preparation and the execution of the action
separately, e.g., as [(T
E
+ T
S
) < (T
A
–T
P
)] OR (T
P
< T
W
). Yet another
170 Joint Cognitive Systems
improvement, and an important one, would be to recognise that the various
times referred to here are approximate rather than precise, hence introducing,
for instance, Min(T
E
) and Max(T
E
) as the upper and lower boundaries for T
E
,
respectively. This could be even further refined by applying the
categorisation of time proposed by Allen (1983).
For the present purpose the descriptions and proposals above are,
however, sufficient to show that it is possible to speak about the temporal
characteristics of actions in an unambiguous manner, and that it furthermore
is possible to propose qualitative – and perhaps also quantitative – relations
between time and control. In reality the situation will always be more
complex than the examples used here, and it is an abstraction to speak about
time without in the same breath mentioning the conditions on which available
time depends. The main conditions are the type of process (process domain),
the state of the process (e.g., normal or disturbed), the frequency of
unexpected events in the process or environment, the appropriateness (or
quality) of previous actions and the anticipation of feedback, and the time
used for previous actions.
Table 8.3: Control Mode Dependency on Time
Condition Resulting control mode
(T
E
∪ T
S
∪ T
P
) > T
A
Scrambled
(T
E
> T
A
) XOR (T
S
> T
A
) XOR (T
P
> T
A
) Opportunistic
(T
E
∪ T
S
∪ T
P
) < T
A
Tactical
(T
E
∪ T
S
∪ T
P
) << T
A
Strategic
HOW TO ENHANCE CONTROL
A shortage of time is always a problem, but it is fortunately a problem to
which several solutions can be found. It is, indeed, possible to see many
common design features as ways of reducing temporal demands. The
strategies or tricks used to compensate for a shortage of time can
conveniently be discussed under two headings, one in terms of the technical
solutions available and the other in terms of the heuristics that people use to
cope with the temporal complexity.
In the following, COCOM will be used as a frame of reference to describe
a number of ways in which the problem of time, or rather the problem of a
shortage of time, is being addressed. This should not in any way be seen as a
validation of the model, but rather as an indication of the usefulness of it.
Since too little time is one of the reasons why control can be lost, the various
solutions can consequently be seen as ways in which control may be
facilitated.
Time and Control 171
Technological and Organisational Solutions
Control rooms broadly speaking, including aircraft cockpits and others, are
awash with technologies that in one way or the other are supposed to make
the operator’s life easier. Some of the commonly applied solutions are shown
in Figure 8.4.
Time to Evaluate Events (T
E
)
The time needed to evaluate events, (T
E
), can be reduced primarily by
improving the design of information presentation. This has for many years
been one of the dominating concerns in the field of human-machine
interaction, and a number of high-level design principles have been put
forward, including adaptive displays (Furukawa & Inagaki, 1999),
multimedia interfaces (Bergan & Alty, 1992), and ecological interface design
(Vicente et al., 1995). The lofty goal has been to provide the operator with
‘the right information, in the right format, at the right time’, although this in
most cases is easier to say than to do (cf. the discussion in Chapter 4).
Information presentation
selective filtering
adaptive grouping
Controller /
controlling
system
Events /
feedback
Action
Construct
τ
O
,
τ
LFT
External event /
disturbance
T
W
T
P
T
S
T
E
Automation,
amplification,
I&C design
Coupled
procedures
Model-based prediction
Evidence check
Decision support
Figure 8.4: Technological solutions to alleviate time shortage.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.