David D. Woods
Cognitive Systems Engineering Laboratory
Institute for Ergonomics
The Ohio State University
Columbus, Ohio USA 43210
Lawrence G. Shattuck
Department of Behavioral Sciences and
Leadership
United States Military Academy
West Point, New York USA 10996
Cognition, Technology and Work, 2, 86—96, 2000
Process control
(Woods, O’Brien, and Hanes, 1987), military command and control (van
Creveld, 1985; Shattuck and Woods, 1997), and air traffic management (Smith et
al., 1998) are examples of distributed supervisory control systems. Distributed supervisory control systems
are hierarchical and cooperative.
They include remote supervisors who work through intelligent local
actors to control some process.
With this
framework, human supervisors, designers, and procedure writers can all be
viewed as remote supervisors who provide
plans and procedures to multiple local actors. The distant supervisors have a
broader scope and a better understanding of the overarching goals and
constraints for the larger distributed system. The local actors have privileged access to the monitored
process and what is actually happening “on the ground” within their
field of view and narrower scope.
These plans and
procedures often are inadequate to cope with the potential for surprise in specific situations (Woods and Roth,
1988; Woods et al., 1994). Given the potential for surprise and privileged
access to data about the evolving situation, local actors must adapt the plans
and procedures to the situation based on their understanding of the intent and
goals behind the detailed steps in the plan (this is often referred to as the
intent behind the plan).
The potential for
surprise in a field of practice, as well as other factors, lead plans and
procedures to be underspecified (Suchman,
1990; Woods, Roth and Bennett, 1990).
For example in one study of device troubleshooting, a rule based expert
system’s directions to human technicians functioned as a kind of plan
linking the expectations, analyses, and heuristics of a remote designer to the
local situation—an actual broken system to be restored to service. Inevitably, in both anticipated and
unanticipated ways, complicating factors arose which challenged execution of
the troubleshooting plan. For
example, impasses arose where a diagnostic test requested by the machine expert
could not be carried out given other circumstances. The plans and procedures were underspecified in the face of
the potential for surprise, requiring technicians to “supply knowledge
and act outside of the scope and direction” of the expert system (Roth,
et. al., 1987). As Suchman (1987) summarized, “instructions must be
interpreted with respect to a collection of actions and circumstances that they
never fully specify;” in other words, “plans are resources for
action.”
Woods (1984)
found in studies of simulated and actual nuclear power plant emergency
operations that “good operations require more than rote rule
following.” Two types of
failures can occur when “events demanded a relatively variable sequence
of component actions and extensive feedback from the environment in order to
adapt to unpredictable constraints or disturbances” (Woods,
O’Brien, and Hanes, 1987):
·
“Type A problems where rote rule
following persisted in the face of changing circumstances that demanded
adaptable responses.”
·
“Type B problems where adaptation
to unanticipated conditions was attempted without the complete knowledge or
guidance needed to manage resources successfully to meet recovery goals.”
In these studies,
either local actors failed to adapt plans and procedures to local conditions,
often because they failed to understand that the plans might not fit actual
circumstances, or they adapted plans and procedures without considering the
larger goals and constraints in the situation. In the latter type B problems
the failures to adapt often involved missing side effects of the changes in the replanning process
(Woods, 1988; Woods et al., 1994).
Shattuck and Woods (1997) found the same pattern in a study
of how local actors adapted when surprises occurred in simulated command and
control scenarios and how they used their commander’s statement of intent
behind the plan in adapting to unexpected events. At one extreme, practitioners would rotely follow the
original plans as described by the supervisor with no regard for the local
complicating factors. At the other
extreme, practitioners would act completely autonomously, leaving their
supervisors ‘out of the loop’ and failing to coordinate with other
local actors toward an organizational target. The results demonstrate the need to strike a cooperative
balance between remote supervisors and local actors, where local actors have
the knowledge and authority that they need to respond to unanticipated local
situations in ways that support achieving higher level goals.
The very high
potential for surprise in military operations leads organizations to develop a
means to support skill at adapting to surprise within the context of larger
plans. In command and control, supervisor-local actors teams practice
communicating commander’s intent and using intent information to develop
skill at adapting to surprise (Klein. 1993; Shattuck and Woods, 1997). Similarly, the potential for surprise
is high in space mission operations, and here too we see an organization that
has developed a means to balance distant supervision with local adaptation.
Watts-Perotti (2000) observed that space shuttle mission
control teams wrestle with the relationship between contingency plans and the
unique characteristics of anomalous situations. During a space shuttle mission
an anomaly occurred during the ascent phase. The anomaly disrupted plans for
the mission. Among the
reverberations of the disrupting event, a previously written contingency plan
directed the controllers to drastically shorten the mission. Was this plan relevant to the situation
they faced? What was the intent and assumptions behind that rule? How did this situation relate to that
intent and assumed situation? How
did this contingency plan apply to the tradeoffs and risks in the situation
they faced? How should the plans be adapted for this mission, given the
anomaly, and how does this situation teach us to revise the plans for future
missions?
Watts-Perotti analyzed the cooperative interactions across
multiple teams in the replanning process as they coped with the consequences of
the disrupting event. The teams
engaged in a sophisticated process of considering the implications of the
anomaly for future plans, evaluating possible contingencies in light of the
anomaly, and revising mission plans using the previous procedures as a
resource. The study revealed a
variety of mechanisms by which mission control balanced distant plans as a
guide to action with the need to adapt to surprising events.
These findings,
from observations across multiple domains, illustrate an inherent and
fundamental tradeoff in the relationship between remote supervision and local
action in establishing the framework for adaptation (Hollnagel, 1993; Ashby,
1956). We will call skill at this
tradeoff the resilience function
of a distributed system.
Supervisors and
the larger organizational context must determine the latitude or flexibility
they will give actors to adapt plans and procedures to local situations given
the potential for surprise in that field of activity (Hirschhorn, 1993). Supervision that establishes
centralized control inhibit local actors’ adaptations to variability,
increasing the risk of Type A failures.
At the other extreme is supervision that provides local actors complete
autonomy. In the latter case, the
goals and constraints important in remote supervisors’ scope are disconnected
from the activity and decision making of local actors. As a result, the
response across multiple local actors may not be coordinated and synchronized
properly, increasing the risk of Type B failures. Skill is a resilience process in distributed cognition that
balances the risks across the two types of failure on either side of the
tradeoff function (Shattuck and Woods, 1997; Woods and Patterson, 2000).