As an example of a theory-of-action consider Figure 3.1: an example of a learning
process that a manager of several regional call centres may go through during a DES
study. In his or her day to day management of the call centres, the manager might
reason that small groups of call operators can be specialised to increase the speed of
call handling and, at the same time, maximise their utilisation. These objectives are
an example of the governing variables (Argyris and Sch¨on, 1996) that make up the
theory (Argyris and Sch¨on, 1996) or image (Senge, 1990) that limits and controls
the managerial decisions made by the individual. Long term decision making will
attempt to satisfy these variables. Thus, the manager’s long term strategy uses
separate regional call centres for handling ‘customers’.
Assume now that the manager is then involved in a DES study to reduce cus-
tomer average wait time on the phone by x number of minutes. The study may show that combining all or even some of the call centres helps meet the modelling
objective i.e. queues are reduced. This, of course, also leads to larger, less spe-
cialised, call operator teams. The manager may even be surprised to find out that
overall the average utilisation of the call operators has increased slightly.
Figure 3.1: An illustration of single and double-loop learning
attitude towards combining the call centres and the larger call operator groups
for the business. This may lead to implementation of the study results and an
improvement in long term service level. This is defined as single-loop learning: a
correction of errors in the management of the specific business problem (Argyris
and Sch¨on, 1996). Figure 3.1 illustrates this concept through the feedback from
consequences of action, in this case highlighted by the model, to actions and strategy.
Single-loop learning is distinct from the manager’s understanding of why the
combination option outperforms his or her theory of small efficient call centres. A
second level of error correction is necessary to achieve this understanding (Argyris,
1992; Argyris and Sch¨on, 1996). This is defined as double-loop learning: a reflection
of the difference between the manager’s theory and the performance of the simulation
model. The outcome of a double-loop learning process is a change in the managers’
long term decision making behaviour.
As an example, consider now that same manager oversees a back office process
dealing with customer applications and record checking. This is also a queuing
process subject to variation in arrival rate and activity time that is split up by region.
If the manager has reflected on the results of the call centre simulation model and
the reasoning behind their own management decisions then they may realise that
a relationship exists between resource utilisation and performance. Thus they may
consider the option of combining the back office resources to speed up application processing. If, however, they have only undergone single-loop learning then they
would automatically apply the same governing variables as before and not consider
combining resources.
If the framework is now used to consider the Bakken et al. (1994) result, discussed
in Section 2.3.2, it can be seen that it is thetheory-of-action that decision makers use to learn from simulation that is key to double-loop learning. Figure 3.2 illustrates the learning systems used by the management and student participants respectively.
The managers used what Argyris and Sch¨on (1996) call a ‘win not lose’ governing
variable. The managers may have felt that they had much more face to save in
the experiment than the students i.e. they had experience in the domain and going
bankrupt frequently might be perceived as quite embarrassing. Thus, in general,
the management participants were not open to testing or reflecting on their theories.
They ‘knew how the system worked’ and did not attempt to refute this knowledge.
This resulted in only minor attitude change about what to do in the first simulation
model and low transfer of learning to the second model.
On the other hand the students may have the advantage of feeling that they had
less to lose; hence they were open to failure and reflected on why they performed
poorly. This enhanced their learning in the first simulation model and demonstrated
their understanding in the transfer model. Figure 3.2 illustrates this with the feed-
back from the consequences to problem governing variables.
Figure 3.2: Learning systems used by management and students participants in Bakken et al. (1994)
on methods to encourage double-loop learning (see Argyris, 1992), but little has
been done to measure it. The most relevant study, Bakken et al. (1994), measures
transfer of learning as an indicator of double-loop learning; however, it does not
consider single-loop learning and the potential benefits that can bring for decision
making.
The remainder of this chapter reviews the literature relevant for measuring single
and double-loop learning - namely attitude change and transfer of learning theory.